Best Hardware for GPU Rendering in Octane – Redshift – Vray (Updated)

by Alex ⋮ Last updated on April 3rd, 2019 ⋮ 178 comments

CGDirector is Reader-supported. When you buy through our links, we may earn an affiliate commission.

Graphics Card (GPU) based render engines such as Redhift3D, Octane or VRAY-RT have matured quite a bit over the last years and are starting to overtake CPU-based Render-Engines.

But what hardware gives the best-bang-for-the-buck and what do you have to keep in mind when building your GPU-Workstation compared to a CPU Rendering Workstation? Building a 3D Modeling and CPU Rendering Workstation can be somewhat straightforward, but highly optimizing for GPU Rendering is a whole other story.

So what are the best Hardware-Components and best GPU for rendering with Octane, Redhsift3D or VRAY-RT, that also is affordable? Let’s take a look:

TABLE OF CONTENTS

Best Hardware for GPU Rendering

Processor

Since GPU-Render Engines use the GPU to render, technically you should go for a max-core-clock CPU like the Intel i9 9900K that clocks at 3,6GHz (5Ghz Turbo) or the Intel i7 8700K that clocks at 3,7Ghz (4,7Ghz Turbo).

That said though, there is another factor to consider when choosing a CPU: PCIe-Lanes.

GPUs are attached to the CPU via PCIe-Lanes on the motherboard. Different CPUs support different amounts of PCIe-Lanes and Top-tier GPUs usually need 16x PCIe 3.0 Lanes to run at full performance.

The i9 9900K/8700K have 16 PCIe-Lanes, meaning you could use only one GPU at full speed with these type of CPUs.

If you want more than one GPU at full speed you would need a different CPU that supports more PCIe-Lanes like the AMD Threadripper CPUs, that have 64 PCIe-Lanes (e.g. the AMD Threadripper 2950X), the i9 7800X (28 PCIe-Lanes) or the i9 7900X Series CPUs that support 44 PCIe-Lanes.

GPUs, though, can also run in lower speed modes such as 8x PCIe 3.0 Speeds and then also use up fewer PCIe-Lanes (8x). Usually, there is a negligible difference in Rendering Speed when having GPUs run in 8x mode instead of 16x mode.

This would mean you could run 2x GPUs on an i9 9900K in 8x PCIe mode, 3x GPUs on an i9 7800X and 5x GPUs on an i9 7900X. (Given the Mainboard supports this configuration)

CPUs that have a high number of PCIe-Lanes usually fall into the HEDT Platform range and are usually also great for CPU Rendering as they tend to have more cores and therefore better multi-core performance.

When actively rendering and your scene fits nicely into the GPUs VRAM, the speed of GPU Render Engines is of course mainly dependent on GPU performance.

Some processes though that happen before and during rendering rely heavily on the performance of the CPU, Hard-Drive, and network.

For example extracting and preparing Mesh Data to be used by the GPU, loading textures from your Hard-Drive and preparing the scene data.

In very complex scenes, these processing stages will take lots of time and can bottleneck the overall rendering speed, if a low-end CPU, Disk, and RAM are employed.

If your scene is too large for your GPU memory, the GPU Render Engine will need to access your System RAM or even swap to disk, which will considerably slow down the rendering.

Best Memory (RAM) for GPU Rendering

Different kinds of RAM won’t speed up your GPU Rendering all that much. You do have to make sure, that you have enough RAM though, or else your System will crawl to a halt.

I recommend keeping the following rules in mind to optimize performance as much as possible:

To be safe, your RAM size should be at least 2x your combined VRAM size.
You can benefit from higher Memory Clocks, but usually at a larger price premium
You can benefit from more Memory Channels on certain Systems
Look for lower Latency RAM. CL14, for example, could be slightly faster in some workloads than CL16.

Best Graphics Card for Rendering

To use Octane and Redshift you will need a GPU that has CUDA-Cores, meaning you will need an NVIDIA GPU. VRAY-RT additionally supports OpenCL meaning you could use an AMD card here.

The best bang-for-the-buck NVIDIA cards are the NVIDIA 2070 RTX (2304 CUDA Cores, 8GB VRAM), 2080 RTX (2944 CUDA Cores, 8GB VRAM) and the 2080 Ti (4352 CUDA Cores, 11GB VRAM).

On the high-end, the currently highest possible performance is offered by the NVIDIA Titan V and Titan RTX, that also comes with 24GB of Video RAM.

These Cards though have worse Performance per Dollar as they are targeted at a different audience and VRAM is very expensive but not necessarily needed in such high capacities for GPU Rendering.

GPUs, that have 12GB of Video RAM and more, can handle high-poly scenes with over 200 million unique objects best. Take a look at the performance per dollar tables below, though, to get an overview of how expensive some of these cards can get without offering that much more performance.

GPU Cooling

Founders Edition Blower Style Cooler

PRO: Better Cooling when stacking more than one card
CON: Louder than Open-Air Cooling

Open-Air Cooling

PRO: Quieter than Blower Style, Cheaper
CON: Worse Cooling when stacking cards

Hybrid AiO Cooling (All-in-One Watercooling Loop with Fans)

PRO: Best All-In-One Cooling for stacking cards
CON: More Expensive, needs room for radiators in Case

Watercooling

PRO: Best temps when stacking cards, Quiet, can use only single slot height
CON: Needs lots of extra room in the case for tank and radiators, More Expensive

NVIDIA GPUs have a Boosting Technology, that automatically overclocks your GPU to a certain degree, as long as it stays within predefined temperature and power limits. So making sure a GPU stays as cool as possible, will improve the performance.

You can see this effect especially in Laptops, where there is usually not much room for cooling, and the GPUs tend to get very hot and loud and throttle very early. So if you are thinking of Rendering on a Laptop, keep this in mind.

Power Supply

Be sure to get a strong enough Power supply for your system. Most Cards have a Power Draw of around 180-250W. CPU of around 100W and any additional Hardware in your case.

I Recommend a 500W for a Single-GPU-Build. Add 250W for every additional GPU. Good PSU manufacturers to look out for are Corsair, beQuiet, Seasonic, and Coolermaster but you might prefer others.

There is a Wattage-Calculator here that lets you Calculate how strong your PSU will have to be.

Mainboard & PCIe-Lanes

Make sure the Mainboard has the desired amount of PCIe-Lanes and does not share Lanes with SATA or M.2 slots. Also, be careful what PCI-E Configurations the Motherboard supports. Some have 3 or 4 PCI-E Slots but only support one x16 PCI-E Card.

This can get quite confusing. Check the Motherboard manufacturers Website to be sure the Card configuration you are aiming for is supported. Here is what you should be looking for in the Motherboard specifications:

Image-Source: Asus

In the above example, you would be able to use (with a 40 PCIe Lane CPU) 1 GPU in x16 mode. OR 2 GPUs in both x16 mode OR 3 GPUs one in x16 mode and two of those in x8 mode and so on. Beware that 28-PCIe Lane-CPUs in this example would support different GPU configurations than the 40 lane CPU.

Currently, the AMD Threadripper CPUs will give you 64 PCIe Lanes to hook your GPUs up to, if you want more you will have to go the multi-CPU route with Intel Xeons.

To confuse things a bit more, some Mainboards do offer four x16 GPUs (needs 64 PCIe-Lanes) on CPUs with only 44 PCIe Lanes. How is this even possible?

Enter PLX Chips. On some motherboards these chips serve as a type of switch, managing your PCIe-Lanes and leads the CPU to believe fewer Lanes are being used. This way, you can use e.g. 32 PCIe-Lanes with a 16 PCIe-Lane CPU or 64 PCIe-Lanes on a 44-Lane CPU. Beware though, only a few Motherboards have these PLX Chips. The Asus WS X299 Sage is one of them, allowing up to 7 GPUs to be used at 8x speed with a 44-Lane CPU, or even 4 x16 GPUs on a 44 Lanes CPU.

This screenshot of the Asus WS X299 Sage Manual clearly states what type of GPU-Configurations are supported (Always check the manual before buying expensive stuff):

Image-Source: Asus Mainboard Manual

PCIe-Lane Conclusion: For Multi-GPU Setups, having a CPU with lots of PCIe-Lanes is important, unless you have a Mainboard that comes with PLX chips. Having GPUs run in x8 Mode instead of x16, will only marginally slow down the performance. (Note though, the PLX Chips won’t increase your GPU bandwidth to the CPU, just make it possible to have more cards run in higher modes)

Best GPU Performance / Dollar

Ok so here it is. The Lists everyone should be looking at when choosing the right GPU to buy. The best performing GPU per Dollar!

GPU Benchmark Comparison: Octane

This List is based on OctaneBench 4.00.

(It’s quite difficult to get an average Price for some of these cards since crypto-currency mining is so popular right now, so I used MSRP)

GPU Name	VRAM	OctaneBench	Price $ MSRP	Performance/Dollar
RTX 2070	8	210	550	0.381
GTX 1070	8	133	400	0.333
GTX 1070 Ti	8	153	450	0.340
GTX 1060	8	94	300	0.313
GTX 1080 Ti	11	222	700	0.317
GTX 1080	8	148	550	0.269
RTX 2080	8	226	799	0.282
RTX 2080 Ti	11	304	1199	0.253
TITAN XP	12	250	1300	0.192
RTX Titan	24	326	2700	0.120
Titan V	12	396	3000	0.132
GTX TITAN Z	12	189	2999	0.063
Quadro P6000	24	139	3849	0.036
Quadro GP100	16	284	7000	0.040
RTX 2060	6	170	350	0.485

Source: Complete OctaneBench Benchmark List

GPU Benchmark Comparison: Redshift

The Redshift Render Engine has its own Benchmark and here is a List based off of Redshift Bench. Note how the cards scale (1080TI) [RedshiftBench Mark (Time [min], shorter is better)]:

GPU Name	VRAM	RedshiftBench	Price $ MSRP	Performance/Dollar
RTX 2070	8	11.35	550	1.601
GTX 1070	8	17.11	400	1.461
GTX 1080 Ti	11	11.44	700	1.248
RTX 2080	8	10.59	799	1.181
4x GTX 1080 Ti	11	3.07	2800	1.163
2x GTX 1080 Ti	11	6.15	1400	1.161
8x GTX 1080 Ti	11	1.57	5600	1.137
GTX 1080	8	16.00	550	1.136
RTX 2080 Ti	11	8.38	1199	0.995
TITAN XP	12	10.54	1300	0.729
Titan V	12	8.50	3000	0.392
Quadro P6000	24	11.31	3849	0.229
Quadro GP100	16	9.57	7000	0.149
4x RTX 2080 Ti	11	2.28	4796	0.914

Source: Complete Redshift Benchmark Results List

GPU Benchmark Comparison: VRAY-RT

And here is a List based off of VRAY-RT Bench. Note how the GTX 1080 interestingly seems to perform worse than the GTX 1070 in this benchmark:

GPU Name	VRAM	VRAY-Bench	Price $ MSRP	Performance/Dollar
GTX 1070	8	1:25 min	400	2.941
RTX 2070	8	1:05 min	550	2.797
GTX 1080 TI	11	1:00 min	700	2.380
2x GTX 1080 TI	11	0:32 min	1400	2.232
GTX 1080	8	1:27 min	550	2.089
4x GTX 1080 TI	11	0:19 min	2800	1.879
TITAN XP	12	0:53 min	1300	1.451
8x GTX 1080 TI	11	0:16 min	5600	1.116
TITAN V	12	0:41 min	3000	0.813
Quadro P6000	24	1:04 min	3849	0.405

Source: VRAY Benchmark List

Speed up your Multi-GPU Rendertimes

So, unfortunately, GPUs don’t scale linearly. 2 GPUs render an Image about 1,8 times faster. Having 4 GPUs will only render about 3,5x faster. This is quite a bummer, isn’t it? Having multiple GPUs communicate with each other to render the same task, costs so much performance, that one GPU in a 4-GPU rig is mainly just managing decisions.

One solution could be the following: When final rendering image sequences, use as few GPUs as possible per task.

Let’s make an example:

What we usually do in a multi-GPU rig is, have all GPUs work on the same task. A single task, in this case, would be an image in our image sequence.

4 GPUs together render one Image and then move on to the next Image in the Image sequence until the entire sequence has been rendered.

We can speed up preparation time per GPU (when the GPUs sit idly waiting for a task to start) and bypass some of the multi-GPU slow-downs when we have each GPU render on its own task.

Now 4 GPUs are simultaneously rendering 4 Images, but each GPU is rendering its own Image.

Some 3D-Software might have this feature built-in, if not, it is best to use some kind of Render Manager, such as Thinkbox Deadline (Free for up to 2 Nodes/Computers).

Beware though, that you might have to increase your System RAM a bit and have a strong CPU since every GPU-Task needs its amount of RAM and CPU performance.

Buying GPUs

NVIDIA and AMD GPUs are both hard to get by for a reasonable price nowadays since mining is so popular. Graphics Cards are mostly Out of Stock and even when they are available, they are nowhere near MSRP. There is a site called nowinstock that gives you an overview of the most popular places to buy GPUs in your country and notifies you as soon as cards are available.

I put together some Builds in different Price ranges here for you to get a head start in configuring your own dream-build.

Redshift vs. Octane

Another thing I am asked often is if one should go with the Redshift Render Engine or Octane.

As I myself have used both extensively, in my experience, thanks to the Shader Graph Editor and the vast Multi-Pass Manager of Redshift, I like to use the Redshift Render Engine more for doing work that needs complex Material Setups and heavy Compositing.

Octane is great if you want results fast, as it is slightly easier to learn for beginners. But this, of course, is a personal opinion and I would love to hear yours!

Custom PC-Builder

If you want to get the best parts within your budget you should have a look at the Web-Base PC-Builder Tool that I’ve created.

Select the main purpose that you’ll use the computer for and adjust your budget to create the perfect PC with part recommendations that will fit within your budget.

CGDirector PC-Builder Tool

What Hardware do you want to buy?

178

Comments

James

Hi,

Another question!

I have one of these https://amfeltec.com/products/multi-gpu-cluster/ with 2 x 1080ti, can this be used on the GPU max part picker, so 3 x 2080ti and the 4th slot would have the amfeltec card – so total of 5 GPUs.
Hope that makes sense!

Thanks

Alex from CGDirector

Hey James,
That should work, yes. Beware though, that the Cluster (with 3 or 4 gpus) will plug into your pc on only one pcie slot, meaning the bandwidth from the system and cpu to the external gpus, will be heavily limited. This might not make a big impact on small scenes that don’t need much out of core performance but can slow things down when rendering more complex frames.

Cheers,
Alex

Thanks for the info! I will use the Amfeltec machine as the 2nd machine background renderer.
Cheers!
James

Hi Guys, what case could be used for 4 x Gigabyte AORUS GeForce RTX 2080 Ti XTREME WATERFORCE 11G
The hybrid fans are double in size from last years 1080ti!

Thanks!

Hey James,
I am not sure this is even possible with mainstream cases. With double hybrid fans, I would guess the mounting of the cooling at some point is so far away, that the hoses become too short.

Let me know if you find a case that supports this, would be interested myself!

Hi Alex,

Thanks.

This could be the case but very $$$

Thermaltake Core W200 Super Tower XL ATX Case

What do you reckon?

Hey James,
Hard to tell from the spec sheet. The specs say it has a bunch of room for radiators even up to 600mm, or even 4x 240mm at the top, but I don’t think all of the GPUs hoses will be long enough for top mounting of the radiators. Maybe two top one bottom and one right though..

Alex

Thanks Alex!

I’ll have to try and test it, hope all 4 GPU card hoses will be long enough.

Aakash

Hello Sir, After looking at your PC build post I build my PC in December 2018….
Spec
CPU AMD Ryzen 5 2600
RAM Corsair Vengeance 3000mhz 8GB
GPU Zotac GTX 1060 6gb Amp edition
POWER Corsair 650w SMPS
MOTHERBOARD Gigabyte Aorus B450 Elite
CASE Aorus Mid Tower

My question is…I have been using only Arnold and Redshift Renderes
But now I want to use Octane 3/4
How will it affect performance of my PC
should i increase RAM only say i have only 8gb….if i add 16gb 3000mhz
How processing time of Rendering will affect
and
Aorus b450 has 4 slots for ram expendable upto 64gb
I already have 8gb installed and If i add 16gb on another slot will it be fine, or another 8gb will do the work for me?
I have watched a video of “Andrey lebrov” on youtube…and he tells that Ocatne has some limitation that it cannot go beyond a certain VRAM …does increasing my PC RAM will affects Render time or not?

Please Help me in this regard
Thanks
Aakash

Hey Aakash,
You have to differentiate between VRAM and RAM. VRAM is on your GPU(s) and that is where your scene gets loaded into when rendering on the GPU, such as in redshift or octane.

If your Scene is too big for your VRAM it gets outsourced to your System RAM. This though will slow down the rendering, because VRAM access is much faster for the GPU than RAM access.

That said, you should still have a certain amount of RAM for every GB of VRAM you have, because there are things that are loaded into your RAM even if your Scene fits into your VRAM. Usually it is advised to get 2x the amount of RAM of the combined amount of VRAM you have. So in your case at least 12 GB of RAM.

The thing with Ryzen CPUs though is, they are very picky about RAM. Intel you can more or less just dump in any kind of RAM and timings or combinations of RAM and in most cases it will work just fine. With Ryzen though you have to take special care of RAM combinations and it is best to only put complete RAM kits in, and not combinations of 8GB single kit and 16gb dual kit and so on.

So what you can do is get 16gb of RAM (say 2x 8gb) and add that to your already existing RAM but you have to note that it might not work in combination. You will then have to take out your single 8gb stick and run with only the 16gb (2x 8gb) kit which will definitely work on its own.

I wouldn’t though expect a steep performance increase as RAM only is a performance factor if there is too little. If you have too little things start slowing down, but it does not matter if your have enough or 2x enough 🙂

The best improvement for octane and redshift of course would be to get a second gpu, which will boost your performance 100%. You will need more ram then too, maybe 32GB and a second 1060 GTX.

Thank you Sir, for replying so quick…..Your guidance has cleared my doubts

Deepti

Hi I m from india.. i am looking for a PC for 3d Max and Vray.. looking for intel not AMD or threadripper. Can u please Suggest Two Best Configuration… my Budget is around $2000.

Hey Deepti,
Go take a look at the following PC-Builder Tool where you can play around with the budget a bit and are recommended excellent proven pc-builds for your specific use-cases:
https://www.cgdirector.com/pc-builder/

Hope this helps,
Alex

Thanks for the reply Alex! will definitely check.

Sebastien

Thank you for all this information!!

I’m trying to build my first PC, using mainly Cinema 4D with Octane Render (sometimes SketchUP/ Vray).
I would like to start with 1 GPU (RTX 2070) but in the future maybe add at least one more.
I was wondering if a build with AMD Ryzen 2100X CPU and 32 GB or RAM would be good enough for Octane renderings (4K most of the time), I have a limited budget of CAD$ 2000 because I also want to save money for a good 4k monitor.
Do you think I should go for an AMD threadripper 1st gen or I9 build?

Thank you in advance!

Hey Sebastien,
Going with a Ryzen 2700X can give you good performance for Octane. You can add up to two GPUs on that CPU (at x8 pcie) which is fine for Octane.

The AM4 (x470 …) platform is also very future proof, with the next gen ryzen being able to run on it just fine. So I think with your limited budget it is a fine choice. Going Threadripper or X-Intel i9 build would be somewhat more expensive initially but could of course be expanded more in the future (up to 4 GPUs)

Hey Alex,
Thank you so much the quick answer, it comforts me on my initial idea.
Do you think I should wait on the 3rd generation of ryzen?
Also any recommendation for a build?

I was thinking of something like that:

CPU AMD – Ryzen 7 2700X 3.7 GHz 8-Core Processor $404.95
CPU Cooler Cooler Master – MasterLiquid ML240L RGB 66.7 CFM Liquid CPU Cooler $89.99
Motherboard Asus – Prime X470-Pro ATX AM4 Motherboard $193.99
Memory Corsair – Vengeance LPX 32 GB (4 x 8 GB) DDR4-3200 Memory $268.99
Storage Western Digital – Caviar Blue 1 TB 3.5″ 7200RPM Internal Hard Drive $53.50
Storage Western Digital – Blue 1 TB M.2-2280 Solid State Drive $139.99
Video Card Gigabyte – GeForce RTX 2070 8 GB GAMING OC Video Card $699.00
Case NZXT – H500 (White) ATX Mid Tower Case $97.98
Power Supply Corsair – CXM 750 W 80+ Bronze Certified Semi-Modular ATX Power Supply $79.99
Total: $2028.38

Again, thank you for your help!

Hey Sebastien,
That’s a solid build already! Some notes: Consider going 2x 16gb of RAM as you can expand lateron if you want to. Ryzen only has dual Channel max, so no need to go with 4 Ram modules right away.

The Price of the Gigabyte GeForce RTX 2070 8 GB GAMING OC seems a bit steep, should be around 550$ and the 2700X should be around 300$. What vendor are you looking at?

The Rumors of Ryzen 3rd gen sure are quite nice, but if you need a new PC now, get one now, if not, wait 🙂

Thank you for the precisions, I gave it a thought and I think I’ll go with a Threaripper build instead, I don’t want to wait for the third generation lol.
Knowing that I will probably invest in an another GPU later on.

AMD – Threadripper 1920X 3.5 GHz 12-Core Processor
Price
$450.55
CPU Cooler NZXT – Kraken X52 Rev 2 73.11 CFM Liquid CPU Cooler
Price
$179.90
Motherboard Gigabyte – X399 AORUS PRO ATX TR4 Motherboard
Price
$359.99
Memory G.Skill – Sniper X 32 GB (4 x 8 GB) DDR4-3600 Memory
Price
$256.99
Storage Western Digital – Caviar Blue 1 TB 3.5″ 7200RPM Internal Hard Drive
Price
$53.50
Storage Western Digital – Blue 1 TB M.2-2280 Solid State Drive
Price
$149.99
Video Card Gigabyte – GeForce RTX 2070 8 GB WINDFORCE Video Card
Price
$634.00
Case NZXT – H500 (White) ATX Mid Tower Case
Price
$99.99
Power Supply EVGA – 750 W ATX Power Supply
Price
$70.9

Hey Sebastien,
Looks like a solid build!! I see no issues 🙂

Paul

Hi Alex, I apologise if this has been covered but I am trying to work out the exact ratio of cost vs convenience vs performance with liquid cooling the GPU’s. I am thinking about a 4 x RTX 2080ti workstation for rendering in redshift and compute tasks in Houdini and I am put off by the hassle of dealing with a full liquid cooled PC. I was wondering what the performance hit would be if i went for a hybrid cooled or blower cooled PC? Do you have experience with this? Thanks in advance, the comments section alone is worth the blog page.

paul

Ha, I just read below, sorry. You answered my question

OK, I have a valid question… I hope!! The X399 and TR4 motherboards.. There seems to be a big price difference with the boards. Are there any options to look out for? or, why would toy spend $600 on a MB rather than $300? Thanks Alex.

Hey Paul,
The differences are marginal and paying 600 instead of 300 is definitely not warranted. Some differences include, # of usb, # of gigabit lan or even 10gbe, onboard cooling solution (heatpipes/cooling blocks), quality of vrms and other onboard parts, design, size, audio chip, # of fan connectors, m.2 cooling plate on all m.2 slots and so on. Very minor details, unless of course you can make good use of some of these.

The important features will all be the same. I usually recommend going with one of the cheapest TR4 Boards that support 4 GPUs on double slot height and only upgrade if you need some very specific features. (The only thing I might make use of in the future is 10GBit Ethernet… )

paul silcox

That’s great Alex. I went for the MSI AMD Threadripper X399 SLI PLUS TR4 ATX Motherboard and duplicated the components (as closely as I could) from your build below. Very useful information and thank you very much for your time.

Full Build:
MSI AMD Threadripper X399 SLI PLUS TR4 ATX Motherboard
AMD 2nd Gen 16 Core Ryzen Threadripper 2 2950X Unlocked Processor
Corsair 64GB DDR4 Vengeance LPX 3000MHz Memory Kit 4x16GB
Corsair AX1600i 1600W Titanium Modular Digital ATX PSU/Power Supply
Samsung 860 EVO 500GB mSATA SSD/Solid State Drive
be quiet! Dark Rock Pro TR4 Dual Fan CPU Cooler
2 x Gigabyte NVIDIA GeForce RTX 2080 Ti 11GB TURBO Turing Graphics Card
Corsair Obsidian 750D Airflow Edition Full Tower PC Case
£3724.61 + VAT

Plus 2 x Nvidia Titan X which I already own.

Awesome! 🙂

John

Hi Alex, I keep reading your amazing blog and it helped me so much. I’m leaning toward GPU rendering, with a set up including a i9 9900K CPU with 2 x RTX 2070 running on 8x PCI-e lanes. I understand running this way is not a problem for rendering efficiency. But is there any drawback using PCIe lanes at 8x instead of 16x, especially regarding reactivity while tweaking the scene when doing test renders?
Thanks for your help,
Best,
John

Hey John,
No you will not see any drawbacks in using 8x vs 16x in reactivity inside your viewport and interactive previews. I’ll do some more tests on this, but so far I have not been able to measure any differences.

Matt

Hi Alex,
A few months ago I posted here with a build similar to the following which you were kind enough to point out a few additional things to consider (I believe now taken care of in this spec). I’m just dropping a line here again as I’m looking at going for a 4GPU setup and caught between the options of a liquid CPU cooler and additional case fans or the option of a full liquid cooler system. I understand that with 4 cards there could be trouble with overheating and so the full liquid route could be better… though more expensive! I wondered if you had happily built systems with four cards and not needed full liquid cooling? any advice would be hugely appreceiated. Ive copied the two approx specs below
Thanks again,
Matt

Case
THERMALTAKE CORE X71 TEMPERED GLASS EDITION GAMING CASE
Custom Liquid Cooling Kit
Liquid Series High Kit – EK
Graphics Card Cooling
EK FC Nickel & Backplate – For Two Identical Graphics Cards!
Overclocked CPU
Overclocked AMD Threadripper 2970WX 24 Core (3.0GHz @ up to 4.0GHz)
Motherboard
ASUS® ROG ZENITH EXTREME ALPHA (DDR4, WIFI, CrossFireX/SLI) – RGB Ready!
Memory (RAM)
128GB Corsair VENGEANCE DDR4 2933MHz ~ (8 x 16GB)
Graphics Card
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
1st Storage Drive
2TB Samsung 860 QVO 2.5″ SSD, SATA 6Gb/s (upto 550MB/sR | 520MB/sW)
1st M.2 SSD Drive
1TB SAMSUNG 970 EVO PLUS M.2, PCIe NVMe (up to 3500MB/R, 3300MB/W)
Power Supply
CORSAIR 1600W PRO SERIES™ TITANIUM AX1600i DIGITAL MODULAR PSU

Or…

Case
THERMALTAKE CORE X71 TEMPERED GLASS EDITION GAMING CASE
Overclocked CPU
Overclocked AMD Threadripper 2970WX 24 Core (3.0GHz @ up to 4.0GHz)
Motherboard
ASUS® ROG ZENITH EXTREME ALPHA (DDR4, WIFI, CrossFireX/SLI) – RGB Ready!
Memory (RAM)
128GB Corsair VENGEANCE DDR4 2933MHz ~ (8 x 16GB)
Graphics Card
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
1st Storage Drive
2TB Samsung 860 QVO 2.5″ SSD, SATA 6Gb/s (upto 550MB/sR | 520MB/sW)
1st M.2 SSD Drive
1TB SAMSUNG 970 EVO PLUS M.2, PCIe NVMe (up to 3500MB/R, 3300MB/W)
Power Supply
CORSAIR 1600W PRO SERIES™ TITANIUM AX1600i DIGITAL MODULAR PSU
Processor Cooling
Corsair H150i PRO Hydro Series High Performance CPU Cooler
Extra Case Fans
4x CoolerMaster MasterFan Pro 120 RGB + Controller Kit
Sound Card

Hey Matt,
Yes, as a matter of fact I am currently final Rendering a Project on 2 nodes with each having 4x1080Ti and 4x2080Ti respectively, all on Air. GPUs and CPU. Closed Case no external cooling, 2 additional case fans at front one at the back.

Overheating is usually not a Problem with GPUs as they only Boost until they get towards their temp limit and start clocking down slightly to stay within temperatures limits.

The great thing about watercooling of course is that your GPUs can boost longer or all the time, as the Cooling is much better. But you have the added costs, that is for sure.

I thought about going the watercooling route and did some tests, also because of the possibility of cooling the GPUs on single-slot height and getting up to 7 gpus into one system. (on intel and sage board)

But I found that I don’t have the time to set up and maintain a watercooling system, and also don’t want to spend that much more for a few additional % of performance.

I think it depends a bit on how much of a premium you are paying for the watercooling solution, as I wouldn’t expect more than 5 – 10% extra performance running 4 GPUs on water than on air (blower style).

The GPUs probably last longer, but then again I have always had to upgrade to new Gen GPUs before my old ones broke, so I can’t really say that longevity is my most important consideration.

My GPU Render Nodes are nothing special, but they get the job done. These are the specs:

CPU AMD Threadripper 2950X
CPU Cooler be quiet! Dark Rock Pro 4 TR4
Graphics Card #1 NVIDIA RTX2080TI – Asus Turbo
Graphics Card #2 NVIDIA RTX2080TI – Asus Turbo
Graphics Card #3 NVIDIA RTX2080TI – Asus Turbo
Graphics Card #4 NVIDIA RTX2080TI – Asus Turbo
Memory(RAM) 64GB Corsair Vengeance LPX Kit DDR4-2933, CL16-18-18-35
SSD Samsung SSD 860 EVO 500GB SATA
Mainboard MSI X399 Gaming Pro Carbon AC ATX TR4
Power Supply Corsair AX1600i 1600W ATX 2.4
Case Corsair 600q

At Roughly: $7180.79

Matt Rowley

Thanks for the quick reply Alex..

This sounds very encouraging.. Just out of interest, are the Asus cards you are stacking there the single fan blower style or more like the founders edition style with two fans.. I also read there is a marked difference between these types when stacking them?.. I may have to go with the two fan variety.

One other question.. If I only have 4 cards on this motherboard will I be able to stack them more than single slot distance apart? (Excuse my lack of build knowledge!)

Hey Matt,
The Asus Turbo are the single Fan Blower style GPUs. The dual Fan rtx ones from Nvidia unfortunately are not blower style. So you might want to go with a board Partner card.

Just to make things clear: the single blower style cooling is better for stacking than the dual open air fan cooling from nvidia directly.

The MSI Carbon Pro AC, and many others, have the 4 pcie-slots spaced exactly 2 slot-heights from each other, so you can build 4 double slot gpus in. You can usually tell by looking at pictures from the Mainboard you want to buy, as there really only is single slot distance (which is really close) or double.

Thanks once again Alex. In the end I swapped out the board for an Aorus Extreme X399 board for better spacing and a 4 x Zotac single fan blower card setup. Looking forward to getting up and running now. cheers!

robob3ar

so heavily researching amd 2990wx or i9 9980xe – I’m primarily rendering on octane, have 3 gpus so far, 2080ti 1080ti and 980ti – wanna get another 2080ti (got the asus strix that is 3x the size) – so multiple problems there
– I want to render, but I also wanna build water/smoke sims etc. like phoenix fd, and getting into houdini more, and also using unreal editor .. and also I wouldn’t mind having a go with cpu/gpu renderers such as vray next and (isn’t redshift using both as well) ..

– I wanna keep all options available, since I stayed away from liquid simulations and most of the cpu intensive tasks – it’s just damn frustrating trying to learn something when the system is unresponsive enough (that’s why octane is my favorite, the response is quite fast in most stuff) ..

.. I undestand the downfall of NUMA architecture on threadrippers, as phoenix fd tests showed that less cores on an intel gives better results, 22cores on old xeon beats by a few minutes 32 cores on 2990wx.. and there are a bunch of other situations where you either still want better single core performance, and you want better memory managament that comes with intel – and mostly where amd excells is the pure rendering and whatever paralel task is better – becuase of the core numbers ..

..so I was totally hyped on amd, managed to convince myself to invest into 2950x and then said fuck it – gona buy 2990xe (cause YOLO), and then, I started digging more into intel vs amd tests. and got a bit disapointed with amd, and now I feel intel is gonna run more stable, a bit faster single core, and a bit more reliable results – BUT we still have the number of pci lanes issue – 4 gpus – that’s 64 lanes

– and I also wanted to go mad and want to get 2 nvme m.2 disks, 1tb or 500gb for system, and 2tb for apps and storage of work files and simulations (would it be insane to have 3 ssds, one for system, one for apps, one for storage? – would I feel it? )
..this would total out the pci lanes with additional 8 or how many nvme uses?

SOooo getting back to the question what really happens if you have like 2 lanes per device (that’s 2gb.s afaik)..
– and in this case, should i sacrifice the stability of intel for better pci device support – here I can get intel o amd for very similar price, so it’s just a matter of choice ..
..also wanted to plug in the 128gigs of ram – and no one sells those in a pack, but as I’ve read here, I’ll look for some more – also regarding amd NUMA thing, the memory speed is greatly affected and causes lag, so I guess the faster the ram here would prove more benifical than on intel cpus..

What would your reccomendation be for that ultimate cpu and mainboard – for a 3d generalist, with a render performance in first place (I do as well modeling, simulations, UE, texturing (substance), zbrush, and after effects, photoshop, and cad tools) – sounds like intel would be my best bet, but what about the damn pci lanes?

Hey robob3ar,
I hear you! This is exactly the problem we have and the main decision we have to make as there is no one platform that fits all. It usually comes down to single-core speed vs HEDT platform and its benefits.

The easy solution for when money is not important, of course is to get two or more PCs. One high-end single core beast for active work and one pc that has loads of gpus and cpu-cores. Hook them up over 10Gbit LAN and get a Render Manager such as Thinkbox Deadline that is free for two nodes and you are good to go. This is how I do it currently for my Client work.

Multiple PCs also have the benefit, that your render beast can be in another room and not disturb you while you work or talk on the phone or listen to music.

If multple pcs is not an option for your then the closest thing you get to an “alround” pc for all kinds of tasks is an i9 9990XE (or 9980XE because the 99990XE is not publicly available) Expensive, I know!

You will need a SAGE mainboard for those additional pcie lanes too.

thanks, yeah I figured the same, so I already ordered the 9988k .. the i9-9990XE is impossible to get apparently, on auction only last thing I read..
..and yeah I might as well get the mega core machine in the following year if I determine there’s still a real need for it.. for what I’m using currently a really fast single core is a must, 3dsmax and rendering on 2x2080ti – so I don’t really need all those cores.. also on the upside some specific plugins such as phoenix fd for simulating water. smoke etc. seems to not like a lot of cores, where dual core xeons and amds seem to not to scale a lot, so this cpu seems like a best bet.. I mean the 2080ti is priced like a 24 core threadripper

andi bufi

hello people.
i just bought a amd ati radeon rx 580 8gb and i’m trying to use the latest vray rendering, but im having some problems to use, cuda,
im reading that i can’t use amd for rendering with vray next because i can use only nvidia graphic cards has the cuda tecnology. is it true??

Hey Andi,
Yes this seems to be the case unfortunately:
https://docs.chaosgroup.com/display/VRAY4MAX/Set+Up+V-Ray+GPU

Fikru

Hi Alex first honored me for your answers to your followers
This is my PC Build
HP z800 Xeon x5670 2.93 GHz, 48 GB DDR3 of Ram,
256 GB of SSD, 2x quadro 5000 2.5 GB SLI ,850 Watt PS.
But you know in max 2018 I am used Thanking Particles so when its simulating the scene it takes much time for 1 frame, the other thing is on Lumion 8 it is also to much slower on viewport. So what I want to ask is which part of my computer is updated to fix this very low performance issues.
Note : I am working mostly on Particles simulation in differing Software’s
Thank you

Hey Fikru,
First you should check your Taskmager to see how your hardware is utilized when you do what you usually do. Simulate some Particles and at the same time check in the Taskmanager if you CPU is being utilized fully (100% on all cores) or just a single core. That will already tell you if you particle Simulation needs high Core counts or high single core clock.

For Lumion you might want to check in a Lumion forum as I don’t use this program. I know it renders on the GPU mainly so RTX cards would be adviseable here, but the Viewport I would suspect depends mainly on the cpu core clock. So an i9 9900k COULD be best here.

Thank you very much for answering, but I do not know what the HP z800 CPU and GPU will accept, but I know 100% of the task manager manager is working on both CPU coordinates.
For general how can I upgrade my PC to the best CPU thenk you

Hey Fikru,
I fear I don’t fully understand what you are trying to say. As mentioned above there is no ONE cpu that is best for everything.

A good Multi-Core Choice is the Threadripper 2950X, a good active working cpu is the i9 9900k. That should be the base for you to start.

Check out the PC-Builder Tool that will help you find the right parts for your purpose here:

https://www.cgdirector.com/pc-builder/

Mau Borba

Hi Alex! Beautiful article. I’m not sure if someone already asked this, but what’s a good motherboard for working with i9 9900x with 4 GPUs. I’m aware the i9 doesn’t support 4 GPU’s at full speed, but in my experience, that’s not a big of a let down if you get the secondary lanes at 8x.

Hey Mau,
The ASUS WS X299 Sage is an excellent board that has 7 pcie-slots. As it has a PLX Chip you can run I think up to 7 GPUs at x8. Also the ASUS ROG Rampage VI Apex or similar boards with 4 evenly spaced pcie-slots with room for 4x dual-slot gpus are excellent.

Thank you Alex. 😀

Giuseppe Ferrillo

Hello Alex!!! First of all, I really appreciated your articles, so in depth and well explained.
Well, here you have my upcoming configuration, my main tool are c4d, octane/redshift and Ae.
In the future I would like to expand it more with one or two rtx 84 in total), let me know you think about it:, thanks a lot really!
Giuseppe

________

WS X299 SAGE Dual M.2 DDR4 Socket 2066

N. 2 RTX 2080 Ti Strix OC 11GB

RTX NVLINK Bridge 3-SLOT

AX1600i Digital 1600W Modular 80 Plus Titanium

HYDRO SERIES H150i Pro RGB 360mm x Socket 2066

2x32GB Kit TridentZ RGB DDR4-3000 CL14

Intel CORE i9 9920X 3.5Ghz 12 Core HT Socket 2066

N.2 512GB 970 Pro SSD NVMe

4TB WD RED – SATA 6Gb/s 64MB

The Tower 900 Extended-ATX NR 1 242,00 -2% 237,16

Monitor G 34WK95U UltraWide 5K

Hey Guiseppe,
Excellent build! I see no issues. What kind of RAM though is that exactly? 2 Kits with 2x 16GB? Or one kit with 2x 32GB?

Thanks,
Alex

hello Ales! thanks for the reply! Yes, there 2 kits with 2x16gb. I just a bit undecided about put a water cooled system right now with 2 gpu setup or put it in the future when the whole setup will be upgraded with 1 or 2 more RTX gpu’s.
thanks again!
Giuseppe

Hey Guiseppe,
Be sure to get one RAM kit with 4x 16GB not two with 2x 16gb as the full kits are tested together and will work together well.

It IS possible to have 4 GPUs air-cooled in a case, though keep in mind that these gpus should have blower style cooling as they are stacked very closely to each other. The STRIX GPUs will only cool well if they have space between each other.

Water Cooling can be the solution here of course, but it adds a good chunk of cost and maintenance to the system.

zivmil

Hi alex, After some animation render tests we did on our files in 3dsmax me and my partner decided we’d better go with a workstation computer based two video cards like the 2080TI in nvlink and not to buy and render on a processor-based computer like the 2990WX.
Every test i render with my 1080TI was away away faster then to render the same scenes with my 5820K @ 4100Mhz, so i think i go the GPU based render workstation system.
That’s the spec I’ve been thinking about

GA-X399-AORUS-XTREME – Gigabyte
AMD Ryzen Threadripper 2 2950X / 2920x
64GB (4x16GB) 3200MHz DDR4 CL16 DIMM XMP HyperX Predator RGB
Enthoo Evolv X Glass or Dark base 900 REV2
1200W HX PSU CORSAIR
NVMe ADATA XPG SX8200 Pro 1TB
2X GeForce RTX2080Ti XLR8 OC Triple Fan
AORUS NVLINK™ BRIDGE 4-Slot
NH-U14S-TR4-SP3 – Noctua

but my dilemmas are do I need a 16-core processor if in most cases I’m an going to render on a GPU based engine like VRAY NEXT GPU?
Or is it better to purchase a cheaper 12 or 8 cores TR processor or maybe go on Intel with fewer cores fewer PCI-E lanes but a higher clock speed and cheaper too like the 9900K?
should i go for the evolv x? or with 2 2080Ti the full tower darkbase 900 the cards will have more space?
thank you!

Not Victor :)

I’m guessing if you’re only putting in 2 graphics cards and you aren’t going to be rendering with something like Arnold, then go lower core count. With the money you save you can put it towards another GPU 🙂

Hey Zivmil,
GPU render engines like high cpu clocks and don’t need lots of cpu cores. So going with an i9 9900k for example would give you more performance than a 2950x. IF you stay with only 2 GPUs. If you want more you should get a CPU with more pcie lanes such as 2950X. x8 pcielanes per gpu on the i9 9900k is sufficient.

Also note, that if you want more than 2 GPUs in the future you should be getting blower-style gpus from the start (such as asus turbo), as they cool better when stacked on top of each other.

If 2 GPUs is the maximum number you want to get, the Enthoo Case is sufficient. As far as I know that case even has a riser cable to mount a fourth gpu vertically, so it should even be sufficient for a 4-gpu rig.

Wow, thanks for creating and maintaining this page! It’s a very helpful and rare resource.

I currently work with Redshift and have an opportunity to acquire 6 gpu’s. I’ve been posting to all sorts of hardware forums trying to figure out the best way to build a system that would support it, but nobody seems to have experience to give a clear direction.

Currently I’m looking at the Asus WS x299 Sage motherboard paired with the i7-9800X, but I also have a Threadripper 19650x system already.

According to this article, out of 64 total lanes there are enough to run 7 gpu’s @ 8x speed, but I haven’t seen a motherboard that supports this.
https://www.techpowerup.com/236017/amd-ryzen-threadripper-memory-and-pcie-detailed-what-an-mcm-entails

Not sure the best way to proceed focused on budget. Should I convert the threadripper system into a render rig and create a second work focused station? Or just go the whole 9 yards with a new HEDT system to house the gpu’s? I’m also assuming I’ll want at least 128GB ram?

I guess right now my main concern is how much RAM is enough to run that many cards efficiently. If I could get away with 64GB that would be great, 128GB would be the most expensive line item after all the graphics cards.

Hey Victor,
Thanks for the questions, very interesting 🙂

There are only 2 TR Mainboards that support 5 x16 mechanical pcie-slots. The Gigabyte x399 Aorus Series. That said though, only 4 of those slots support x8 or higher speeds. So lets put the Threadripper platform aside, as this wont work.

x299 X-Series cpu Intel platform is next. As you already said, the Sage Mainboard that have PLX chips allow up to 7x pcie cards. Single slot! You’ll need a bunch of riser cards to fit double height gpus on that mainboard, so keep that in mind. If you do have single slot gpus, this would be the way to go to have all the gpus in one system.

On RAM: It very much depends on the type of scenes you are rendering and how much concurrent tasks you are rendering at the same time. Say you are splitting up your animation frames to render 4 frames at the same time, 2 gpus each (or the last one would render alone), then you will need a lot more RAM than when you render one frame on all gpus. (so just one task at a time)

I would say 64 GB of RAM CAN be enough. Quick test to do right now. Go render a typical scene on your 1 gpu rig and see what ram usage your renderer has (c4d or whatever proramm you are using). Then multiply that by the number of gpus you will have.

Then there is such a thing as preparation time per frame. If you render fairly simple scenes that render fast, you will loose lots of performance when going single-pc with lots of gpus. When preparing a scene and the gpus arent utilized, 6 gpus sit there waiting idly until they can start rendering. If you have 2 pcs that time already is cut in half, if you have one pc per gpu that time is even further reduced.

Quick Maths: You have 6 frames to render & each frame has 30s preparation time and 30s bucket render time. On one pc with 6 gpus this takes 6 min. On 6 pcs with 1 gpu each this takes 3min30s.

Why? Because on 6 pcs with 1 gpu each, 6 frames can prepare simultaneously, having less idle time for the gpus. Of course this can be ignored the longer the bucket render time becomes and the shorter the preparation time is. So if you render frames that prepare 30s and render 12h, ignore what I just said 🙂

Very helpful, Thank you so much for taking the time to explain it. Yes, actually that confirms a suspicion I had, which was the setup time before renders. I have had scenes, especially with a lot of subdivision on deforming geometry, where there was a lot of pre-render time. I could imagine scenes like that taking a lot longer on a distributed load.

I was looking at getting a lot of 6x GTX 1070’s for $220 each which is such a good deal, but the overall the budget for the system was ballooning to $4k.

Someone from another site mentioned when GPU renderers start taking advantage of RTX, one 2070 card could run 2-3x the speed of a 1070 devaluing the rig pretty quickly.

Taken into consideration with what you’ve outlined that performance wouldn’t be guaranteed to scale up linearly, I decided to hold off on the idea for now.

It’s a shame with all the mining rigs out there being sold off at greatly reduced prices that there isn’t a practical way to leverage a system with a lot of older cards.

Yes you are right. The RTX Cards might get a good performance increase soon. Octane already has some 2-4x speedup benchmarks on simple scenes on the 2019 octane, and redshift seems to be getting some improvements in this area too soon.

The thing with mining cards is, yes they are nice and cheap but you also don’t know how long they will last as they have been used 24/7 for quite some time. But you can be lucky of course!

Chhers,
Alex

Hi, Alex! Which of these cards is better to buy now for GPU render? 2xRTX 2070 or 2xGTX 1080 Ti

Hey Alex,
The RTX 2070 has a higher value than the 1080Ti, also the 1080Ti is rather difficult to come by at reasonable prices. The only thing is, that the 1080Ti of course has 11gb of VRAM and the RTX 2070 “only” 8. If your scenes aren’t all that complex go with the 2070s.

Chris_N

Hi Alex, your Blog is really awesome and helpfull!
I’m a bit new in building my own system and hope you’ll find a seccond to check my pc-build-idea.
I’ll use it most of all for C4D, GPU Rendering (Octane) and Film editing.

What do you think?

Corsair Obsidian 750D mit Sichtfenster Big Tower ohne Netzteil schwarz
MSI MEG X399 CREATION AMD X399 So.TR4 Quad Channel DDR4 EATX Retail
16GB G.Skill Trident Z RGB DDR4-3200 DIMM CL16 Dual Kit
Enermax LiqTech TR4 II 360mm CPU-Kühler
1200 Watt be quiet! Dark Power Pro 11 Modular 80+ Platinum (for more than one card in the near future)
1000GB Samsung 860 Evo 2.5″ (6.4cm) SATA 6Gb/s 3D-NAND TLC (MZ-76E1T0B/EU)
Intel Core i9 9900K 8x 3.60GHz So.1151 WOF
1(x) 11GB KFA2 GeForce RTX 2080 Ti Dual Black OC Aktiv PCIe 3.0 x16 (Retail)

The plan for the future is to build a 2nd RTX in the system
Do I miss something?

Really thnx for your time!

Hey Chris,
The Mainboard and CPU don’t match. You neither need a Threadripper CPU or a LGA 1151 v2 Mainboard for the i9 9900k.

Other than that, looks nice!

Also check out the pc-builder tool on this site that can help you find the right components for your needs:
https://www.cgdirector.com/pc-builder/

Gianluca

Hi Alex, your blog is very helpful ☺️ I would ask you a thing: does Redshift 2.5.46/48 work with new RTX graphic cards like 2080, 2080ti, etc? Or is it necessary upgrade to the last release (Redshift 2.6)?

Hey Gianluca,
As far as I know you will need redshift 2.6 or newer to be able to use RTX cards. This question has been asked a few times on the redshift forum over here if you want to take a look at it:
https://www.redshift3d.com/forums/viewthread/22174/

Twen

Thanks for the helpful information Alex!
I’m building a new machine for GPU rendering (Octane/Cinema4D) so I need a strong single core performance and as much as possible PCI lanes for the GPU.

I’d like to hear your thoughts about the Build I made below.
How it can be improved? and where can I save some $$?
and since it’s going to be used for GPU rendering which cooling GPU/CPU setup do you suggest?

Motherboard: EVGA – X299 DARK EATX LGA2066 Motherboard ($248.90 @ OutletPC)
CPU: Intel – Core i7-9800X 3.8 GHz 8-Core Processor ($649.99 @ B&H)
Memory: Corsair – Vengeance LPX 64 GB (4 x 16 GB) DDR4-2400 Memory ($389.99 @ Newegg)
Storage: Samsung – 970 Evo 500 GB M.2-2280 Solid State Drive ($129.99 @ Amazon)
Storage: Samsung – 860 Pro 1 TB 2.5″ Solid State Drive ($297.99 @ Amazon)
Video Card: EVGA – 2x GeForce RTX 2070 8 GB XC ULTRA GAMING Video Card ($1100.00)
Power Supply: EVGA – SuperNOVA G2 1000 W 80+ Gold Certified Fully-Modular ATX Power Supply ($135.98 @ Newegg)
Optical Drive: Asus – DRW-24B1ST/BLK/B/AS DVD/CD Writer ($19.99 @ Amazon)

Hey Twen,
The 9800X has a very strong single core performance and 44 pcie-lanes so more than enough for you to run 2 GPUs. You could even add up to 4 GPUs at x8 without a problem.

Note that the GPUs have open Air coolers, so if you were to stack them right on top of each other OR if you were to get more GPUs in the future and then have to stack them, the cooling is not that great. For multi gpu setups blower style fans are recommended as they can handle gpu stacking much better. Asus Turbo 2070RTX comes to mind here as an example.

Other than that, great build!

Another Victor

Hey Alex, thanks for doing this thorough explanation, I want to start doing GPU render and I have been a Mac User for a long time.. Now I want to buy a PC and I don’t want to make a mistake choosing a wrong part.
Does this make sense for you?

-Intel® Core™ i7-8700K Processor (6x 3.70GHz/12MB L3 Cache)
-DEEPCOOL Captain 360EX 360mm Liquid Cooling System
-32 GB [16 GB x2] DDR4-3200 Memory Module – Corsair Vengeance-LPX
-NVIDIA GeForce RTX 2060 – 6GB (GDDR6) (VR-Ready)
-ASUS TUF Z390-PRO GAMING — RGB, Gb LAN, USB 3.1 (6 Rear, 4 Front), ASUS TUF Protection
– 1000 Watt – Corsair RM1000X Power Supply – 80 PLUS Gold, Full Modular
– M.2/PCI-E SSD Card: 250 GB Samsung 970 EVO M.2 PCIe NVMe SSD — Read: 3400MB/s; Write: 1500MB/s
-Primary Hard Drive: 256 GB Samsung 860 PRO SSD — Read: 560MB/s, Write: 530MB/s – Single Drive
-3D Premium Surround Sound Onboard
-Intel® Wireless-AC 3165 Wireless Network Card 802.11ac Dual Band (2.4GHz/5GHz) up to 433Gbps + Bluetooth 4.2 with PCI-E Adapter
– Windows 10 Home
-Tuniq TX-2 High Performance Thermal Compound

I’m thinking to put more GPU in the future..

Thanks for your time!

Hey Victor,
Your build is spot on! I see nothing that you have to change. Beware though that you will only be able to have 2 GPUs running at x8 (pcie speeds) maximum on that cpu and board. If you want even more, say 4 GPUs in the future you should go HEDT Platform such as Threadripper or Intel X-Series CPUs.

But to start this is an excellent mainstream platform build. You just cant upgrade it as far as an HEDT build.

Another Vicor

Thanks Alex,
And what about this one, I changed what you said..

Case iBUYPOWER Trace Tempered Glass RGB Gaming Case
Default Case Fan
Processor AMD Ryzen Threadripper 2950X Processor (16x 3.5GHZ/32MB L3 Cache)
Asetek 591LX 360mm Ultimate Liquid Cooling System
Memory 32 GB [16 GB x2] DDR4-3200 Memory Module – Certified Major Brand Gaming Memory
Video Card NVIDIA GeForce RTX 2070 – 8GB (GDDR6) (VR-Ready)
Motherboard MSI X399 GAMING PRO CARBON AC — RGB, 802.11ac WiFi, USB 3.1 (8 Rear, 4 Front), Unique MSI Gaming Tools
Power Supply 1000 Watt – Corsair RM1000X Power Supply – 80 PLUS Gold, Full Modular
Advance Cabling Options Standard Default Cables
M.2/PCI-E SSD Card 250 GB Samsung 970 EVO M.2 PCIe NVMe SSD — Read: 3400MB/s; Write: 1500MB/s
Primary Hard Drive 250 GB Samsung 860 EVO SSD — Read: 550MB/s, Write: 520MB/s – Single Drive
Sound Card 3D Premium Surround Sound Onboard
Network Card Intel® Wireless-AC 3165 Wireless Network Card 802.11ac Dual Band (2.4GHz/5GHz) up to 433Gbps + Bluetooth 4.2 with PCI-E Adapter
Operating System Windows 10 Home + Office 365 Trial [FREE 30 Day Trial] – (64-bit)
Keyboard iBUYPOWER Standard Gaming Keyboard
Mouse iBUYPOWER Gaming Optical Mouse – Multi-Color LED Lighting
Advanced Build Options – Thermal Paste Tuniq TX-2 High Performance Thermal Compound

thanks again…

Hey Victor,
That’s an excellent build that will let you run up to 4 GPUs. The 2950X too is an excellent CPU for fast CPU Rendering, so I’d say looks great, go for it! 🙂

Peter Hugh

” If you have to render 4 images and have 4 GPUs, let every GPU render one image instead of having 4 GPUs render on every image.” WHAT ?!?

You mean to say:

Have all GPUs Render One Image. Rather than ALL GPUS Rendering on ALL Images simultaneously.

If you’re going to write these Articles please “brush-Up” on your English. Less is More. Verbosity sucks.

Hey Peter,
Thank you for your feedback! I don’t think you understood what I meant in that passage though. I updated that bit and hope it is easier to understand now! 🙂

hans

If you are going to comment on someone’s writing then at least give an example that clearly demonstrates your mastery of the subject.

Scott

Hi Alex, great info. Can I ask what formula you are using to work out the redshift performance / dollar? It doesn’t seem to calculate against the time for me.

Hey Scott,
Sure here is the formula: ((1/Price) / Score).

Victor

Great article and it is very much like-minded to me. There is an outstanding experience expressed on your website, which you could be proud of! Would you be interested to test my PC 7x 1080ti for your research purpose and have few notes about my pc?

It would be great to hear back from you!

Thanks
Victor

Hey Victor,
Wow 7 GPUs in one build? I am very much interested what components you used, yes!

Hey Alex,
In the past, I used Asus Z10PE-8D + 2 Xeons with low frequency per single core and recently upgraded to Asus X299 Sage with i7 Cpu. The significant increase in speed because of 2-3 times faster-loading speed. Also, it has a water cooling loop

Hey Victor,
Thanks for the info!

So how many pcie-lanes do each of the gpus get? How much RAM do you have. Are you rendering one frame across all GPUs or splitting up and rendering multiple frames at once? Are you using a render manager?

What i7 CPU are you using?

Thanks for your info, this is exciting!!
Alex

Hi Alex,
I hope you had a great weekend. I use i7-9800x CPU, the cheapest processor in the lane which has 44 PCIe-lanes, as I don’t see any reason to overpay for additional cores which don’t use in rendering. Asus X299 Sage provides this scheme for PCIe-lanes – seven at x16/x8/x8/x8/x8/x8/x8. As a render manager, I use Deadline and render each frame with all GPUs. I’ve never thought about splitting them into multiple tasks. Probably it will be the next test
Thanks
Victor

In one of the articles you mentioned, that GPU is not scaling proportionally in Vray. In real life scenes, it works absolutely fine. VrayBenchmark is not optimized for multi-GPU systems and way too short.

Thanks I’ll have to do some more real-live tests apart from the benchmarks!

Hey Victor,
Awesome! Yes going fairly low# of cores and high-core clock is better for gpu as far as I have tested.

Having multiple concurrent tasks can speed up rendertimes expecially when you have short frame render durations, as lots of times the gpu are waiting for the mesh processing and so on to finish and are just waiting in idle. Better to have 4 or so process simultaneously but of course you need a good amount of RAM then for each rendertask. Deadline is an excellent Rendermanager, I use it daily. 🙂

Hey Alex,
Yesterday I tried to render with multiple concurrent tasks and want to share some thoughts. I didn’t manage to run more than 2 tasks at the same time. Probably because I have only 2 render nodes licenses or lack of RAM (32gb current setup). I’m sure it’s a win-win solution for “slow” Xeons with a good amount of memory on the board, but very controversial for me. I run my cards with 2100Mhz boost and quicker they finish the image more often they have a “break for cooling”, With water cooling loop they don’t exceed 60-65C degrees under full load over a long run, but still prefer to give them short tasks rather than long ones. I can assume, that in server racks graphics cards much quicker reach the throttling mode with long tasks
Thanks
Victor

Hey Victor!
I think deadline does not limit on a single machine, the license works with unlimited concurrent tasks. Did you limit the # of GPUs per task as you don’t want all gpus rendering multiple concurrent tasks.

Yes 32 GB is not much for Concurrent tasks, so more than 2 is probably not possible. I see what you mean with the watercooling, thats fabulous! 2100MHz continually is pretty great! 🙂

O don’t think the difference will be that great with concurrent tasks though when you render very short frames (<2min per frame) where the preparation time is a large percentage of the total render time it might come in handy. But I am sure you have it optimized well already!

Thanks!
Alex

Dan Goldstein

Hi Alex, great article! I’m looking to purchase an RTX 2080 Ti card in the near future to use with Maya 2018 and Redshift or Arnold (when they finally release the Arnold GPU version). Question – When deciding between different variations with say EVGA in terms of clock speeds and over-clocking, is that more or less important than the amount of CUDA cores for both the Maya viewport 2.0 and for a GPU renderer?

For example there is the EVGA 2080 Ti FTW3 card which seems to be a lot faster and a lot more expensive than the slower EVGA 2080 Ti Black Edition. So if clock speeds aren’t really the factor and the CUDA cores are then maybe I should go with the cheaper card? Thank you!

Hey Dan,
Will you be stacking more than one card on top of each other? If so you should go with a blower style card such as the Asus Turbo 2080Ti.

In my experience Nvidia GPUs boost much higher than is usually stated in the specs. All of my 1080Tis boost up to 1810Mhz although Asus states 1500Mhz as boost clock. This is probably because Nvidias Boost technology controls the boost independently.

It all depends on power and tmeperature of course, but if you had a very well cooled gpu in a cool environment you might be able to have it render at 1800MHz permanently.

That said, if its only one GPU you are planning on getting, I would go with the cheaper evga one, yes, as the boost clock really doesn’t make that big of a difference.

Thanks for the response Alex! Yes I will only be going with one GPU as my tower isnt that big. Since the boost clock doesnt really make a difference, do you recommend I go with ASUS instead of EVGA of the cheaper 2080 Ti’s or does brand not matter either?

Hey Dan,
Brand differences can be noticed the most in the cooling solution and its longevity. Here Asus, EVGA are very good. EVGA probably even has an edge over Asus. Their Return Policy is also very good.

Speedwise you won’t see much difference, but as these High-End Cards like to die sometimes, going EVGA probably is a good Idea, especially in the US.

Great, thank you!

Hi Alex, 2 more quick questions – would the boost clock make a difference anywhere else in Maya like the viewport for example? And also, if I go with the cheaper EVGA Black Edition card instead of the more expensive EVGA XC or XC2 Ultra Gaming cards with the better heat syncs/sensors am I putting my system at risk for overheating or anything damaging while rendering?

Thanks again!

Hey Dan,
Maya Viewport and also any kind of CUDA or GPU accelerated Render Engine such as Redshift, Octane, Vray-RT. Lots of Programs such as After Effects, Premiere Pro have gpu accelerated effects too.

The difference between the cards is marginal and the better cooling lets the card Turboclock longer without having to throttle. All GPUs will throttle on their own if they get too hot, so no worries about risking overheating. Unless you have absolutely abnormal circumstances and live in the desert. 😉

I have 4 blower style GPUs stacked on top of each other in my render nodes and they are rendering for days on end without any problems.

Having the chips run at 85°C all the time of course won’t increase the lifetime of the GPU 😉 But in my experience they tend to stay alive long enough until after a few new generations you want an upgrade anyway.

Awesome, thanks for the info, it is much appreciated!

Hey Alex, back again! Will I be bottle-necked in any way in Maya and a single 2080Ti because of my processor? Its a 6 core i7-5820k 3.3ghz but overclocked to 3.8 per core. I also have 128gb of ram at 2666mhz, Asus X99A Sli Plus mobo, and a samsung M.2 drive. Thanks!

Hey Dan,
The Maya Viewport 2.0 makes good use of GPU features, so your 2080Ti will probably do some good lifting already. That said, 3,8GHz is surely not the best you can get (think i9 9900K at 5.0GHz) but it also depends on what you are doing in maya. If you are using lots of hires meshes that are just populating your scenes in raw state or if you have lots of modifiers applied to your objects such as mirroring, cloning, displacing, beveling and so on.

The latter will need high GHZ cpu the former will need a stronger gpu.

ali

2x 1060 graphic card is better than 1080ti or not?
so thanks

Hey Ali,
2x 1060 is usually slightly slower than a 1080Ti. You also get less Vram. Also, if you are planning on getting more gpus in the future, having one card will give you more room for upgrades, than already having 2 slow cards.

Manodzit

Dear Alex!

Thank you so much for this article and that i stumbled upon this!
i hope you find a minute to answer this for me, i am trying to come up with a rig, i am yearning to buy….
https://www.morele.net/inventory/info/d61f549e/
would appreciate a lot if you could drop your 2 cents on that list

additionally appreciate a ton if you can shed some advice on the issues below .

1 RAM
the most frustrating is picking up RAM for the threadripper 2950X everyone is screaming out to go with G.skill trident Z as its the lastest and best optimized for threadripper (plus lot of people have issues with RAM especially i am reading a lot people have to downclock their RAM to work with the threadrippper).
the dilema is i would like to ideally get 4×16 64Gigs and later upgrate to another 4 sticks of the same. (but tridentZ doesnt have that option out of the box it only come with either 8×8 64GB OR 8×16 128GB sets) – is it safe to mix match?

2 GPU
after reading your post and another post from the pugetsystems (https://www.pugetsystems.com/labs/articles/GeForce-RTX-2080-Multi-GPU-Scaling-in-OctaneRender-and-Redshift-1258/)
i am thinking of getting either 4 x RTX 2070 (hoping redshift will use the RTX better soon) – only catch is how much of a difference would it make that it doesnt support NVlink. although i am not rendering incredibly huge scenes regularly.

thanks in advance

manodzit

I missed to mention, i would be primarily wanting to use it for houdini simulations and Octane/ Redshift lookdevs and renders.

Hey Manodzit,
That Build looks great, don’t see any issues there.

As for your questions: I’d be careful, especially on Threadripper to mix Ram Kits. It’s one thing to get a Kit and get another Kit a week later, but if you wait too long and production lines shift or get bumped to a new revision in the factory then it could very well be that the mixed Kits will only work when downclocked to base frequency. Very seldomly they still produce crashes and bluescreens.

I’d say if it is bound to be a longer time between the purchase of your 2 RAM kits, sell the old one and go 128GB on a full kit.

Concerning the GPUs, NVLink really isn’t all that important. I almost never have the problem of full VRAM and if I do, most renderers (especially redshift) can do out of memory swapping to the system RAM.

Let me know what you think!

Manojit

I never saw an article about the MSI aero rtx 2070 specifically being ‘good’ for multi GPU setup. have read about the Asus Turbo though (i am pretty sure aero is also a blower card) i am leaning on the Aero rather than Asus turbo!

Also, Is the AsRock x399 taichi a decent board to use the precision boost or BIOS Over clocking for the 2950X?

your two cents please!

Hey Manojit,
The Asus Turbo has a slightly larger fan and slightly higher clock speeds, so it might perform better especially when stacking with more cards. That said though, I’d put the asus turbo on pretty much the same level as the msi aero. I think when it is so close it comes down to support quality and rma speed and I would fathom MSI having a slight edge over asus but that just my experience.

Can’t say anything about the AsRock Board you mentioned though unfortunately as I haven’t had the chance to test that feature on it yet, sorry.

Hi Alex,
I’m putting together a build that I intend to use principally for GPU rendering in Redshift. However, at the moment I also still do a good deal of work in Standard Renderer in Cinema and also do a lot of work in After Affects. I therefore need to cover both CPU and GPU capability with the components I choose.. Im thinking a processor with a med.number of cores that still clocks relatively high?.. I put together this setup as an example and would be great if anyone has an opinion to share. Mainly it would be great to know if Im overspending on components that could be equaled with something cheaper here, and also if Im missing anything obvious to consider when putting together a high end setup like this… (4x 1080 Ti VS 3 x 2080Ti was one quick example I thought of)

Cheers, Matt.

Overclocked CPU Overclocked AMD Threadripper 1950X 16 Core (3.4GHz @ up to 4.0GHz)
Motherboard ASUS® ROG ZENITH EXTREME (DDR4, 6Gb/s, CrossFireX/SLI) – RGB Ready!
Memory (RAM) 64GB Corsair VENGEANCE DDR4 2933MHz ~ (8 x 8GB)
Graphics Card 11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
2nd Graphics Card 11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
3rd Graphics Card 11GB NVIDIA GEFORCE RTX 2080 Ti – HDMI, 3x DP GeForce – RTX VR Ready!
1st Storage Drive 6TB SEAGATE BARRACUDA PRO 3.5″, 7200 RPM 256MB CACHE
1st M.2 SSD Drive 256GB ADATA SU800 M.2 2280 (560 MB/R, 520 MB/W)
Power Supply CORSAIR 1200W HX SERIES™ MODULAR 80 PLUS® PLATINUM, ULTRA QUIET
Processor Cooling Corsair H150i PRO Hydro Series High Performance CPU Cooler

Hey Matt,
Some things to note on your components:
– The 2080Ti has a max power draw of about 330W, so 3* 330W + ~200W for the rest of the System comes fairly close to your 1200W psu. Recommend to get a stronger one, as in the 1600W Corsair.
– The M.2 Drive you chose is a SATA Speed drive. Be sure to get a pcie-speed drive, that should have over 2500MB/s Read/Write, like the samsung 970 evo
– The Mainboard is an E-ATX Board, be sure to get a Case that fits this board.

Normally I would recommend getting the 1080Tis, though these have become so rare, that you can only buy them at unreasonable prices nowadays. You could consider 4x 2070 though, as these compare to the speed of a 1080TI and are a lot cheaper.

vikkram

hi alex,
I am an architect trying to set my configurations of 2 computers 1 for rendering of buildings and another for 3d modeling.. Budget for 2 systems is 1500$ each i made 2 systems configuration, Can you please suggest me suitable motherboards and other parts for both processors?

My preferences for above processors above:

System1 for 3d modelling:
CPU : Intel i7 8086k
CPU cooler : ?
Motherboard (suggest the model) with
PCIE slots = 2
Ram Slots = 4
Memory (RAM) : 1 X 16GB DDR4 4166 or 2133 (suggest model)
storage : PCIE M2 SSD Samsung 970 EVO (500 GB)
GPU : NVidia GTX 1070 (2gb)
Case : Please suggest
Power supply : Please suggest for this

System2 for Rendering:
CPU : AMD threadripper 1900X
CPU cooler : ?
Motherboard (suggest the model) with
PCIE slots = 2
Ram Slots = 4
Memory (RAM) : 1 X 16GB DDR4 4166 or 2133 (suggest model)
storage : PCIE M2 SSD Samsung 970 EVO (500 GB)
GPU : NVidia GTX 1070 (2gb)
Case : Please suggest
Power supply : Please suggest for this

Hey Vikram,
Go take a look at finished builds in this Article here:
https://www.cgdirector.com/best-computer-3d-modeling-rendering/

(towards the end)

You can use those builds as base and tweak them until they fit into your budget.

LGA 1151 CPU Cooler would be something like the bequiet dark rock 4.
be sure to get 2x8gb of ram and not just one stick, as they then run in dual mode. No need to get 4266mhz, unless you wan to overclock. Plain old 2400mhz or so will do just fine and save you some cash.

The Nvidia 1070 is 8GB VRAM , where did you find the 2GB info?

The rest of your questions I hope should be answered when you read the article above 🙂

Lucas

Hey Alex
Thank you so much for this article! I want to use Cinema 4D with Redshift from now on with the following setup (Work and Renderstation):

Worksstation:

CPU: Intel i9-9900K (LGA 1151, 3.60GHz, Unlocked)
GPU: NVIDIAGTX 1060 6GB or NVIDIAGTX 1070 Ti 6GB
RAM: 32 GB Kingston Value RAM, 2666MHz
MAINBOARD: ASUS Prime Z370-A
HD: Samsung 970 PRO/EVO 1TB

Renderstation:

GPU: 4x NVIDIA 1070 GTX 8GB VRAM PCI-E x16
CPU: Intel Core i9-7900X (LGA 2066, 3.30GHz, Unlocked)
MAINBOARD: Asus WS X299 Sage

(RAM, HD, Case, Power Supply, Cooling is to be defined)

My questions:
— What do you think about this setup?
— With the amount of the PCIE Lanes of this setup (CPU 44 PCI-E lanes) and the plx-chips on the mainboards: Is it possible to use 4 (up to 7) GPUS, am I right?

Thank you!
Lucas

Hi Lucas,
Yes these are excellent builds. With the Asus Sage Mainboard you really can use up to the amount of gpus that fit onto the mainboard, though you should check in what speeds they will run when fully utilized. Of course these slots are single spaced, so note that you can’t use 7x dual slot gpus without riser cables.

It is usually recommended to get twice the amount of RAM to VRAM. So your 7x 8gb x2 ultimately you should have around 128gb of RAM.

Note though, that GPU rendering is also very straining on the cpu and the more gpus the more the cpu is utilized. Even with in one of my rigs with 4x 1080ti the cpu is often utilized at around 70%-80& in Redshift/c4d on a Threadripper 1900x.

It depends heavily on the type of scenes you render, that is for sure. If you plan on using a rendermanager and split your tasks up onto 1 gpu / per task, then this strains the cpu even more and will most definitely not work, just a heads up. Even splitting up 2gpus / task on a 4 gpu system gives some frame errors sometimes I have found.

Oh and having 7x GPUs you should think about where all of the heat will go. 🙂

Let me know what you think,

sheraz asghar

Hi
I am a miner and 3D Generalist. Since mining is not profitable as it used to be I want to switch my mining rig into a rendering rig(s). I have 19 GPUs 2x 64 vegas, 5x 56vegas 12x, 580 RX nitro’s. All AMD cards. I am hoping if you can tell me exactly what motherboard(s) and processers(s) to buy that will support the maximum amount of cards on each motherboard. I use MAYA/3ds max with VRAY. Currently, for mining I have 2x Biostar 250 pro motherboard(s) this supports up to 12 GPUs with riser cables. Is that something I can re-use. What would you recommend?

Thank you

Hey Sheraz,
As far as I know your Mainboard only supports a bunch of x1 pcie slots. When GPU Rendering though, you should have the gpus run at x8 at a minimum. Also the CPU should support as many pcie-lanes as possible. Your current CPU (LGA1151) will only support 16x pcie lanes to the gpus, so supporting 2x x8 at max.

If you want to get the maximum out of the cards you have i’d get a ASUS WS X299 Sage that supports 7x x8 pcie gpus. You’ll have to get some new 2066 socket cpus though and maybe more ram, depending on what you have right now. 32gb per node should be good.

jerrod

How do you feel about running two rtx 2070s instead of one 2080

Hey Jerrod,
The 1070 and 2070 often surpass the 1080/2080 in the VRAY GPU Benchmark, and have a much much better price/performance ratio. I am not quite sure why the 2070 and 1070 perform so well compared to the 1080 / 2080 but the benchmark results speak for themselves.

It could of course very well be, that the benchmark scene is just not that great, and having the cards run in real-world scenes will result in different performances, but usually this benchmark gets pretty close to the real thing.

In Octane or Redshift the cards scale better, but vray gpu seems to like the xx70 series very much!

Simon

Is it possible to run four rtx 2080 ti in a single computer ?
Thanks

Simo

Hey Simon,
Yes this is absolutely possible. Take a look at the GPU Rendering Build in this Article here towards the end:

https://www.cgdirector.com/best-computer-3d-modeling-rendering/

Just swap the 4x 1080Ti with 4x 2080Ti

Umut

Hi Alex, I have built pc with i7 8700k gtx1070 and 16gb ram. Right now I’m using standart render engine on C4D which is cpu based. Would it make a change in render time if I used a gpu based render engine such as Octane with my current setup? I know gpu renders are better with multiple graphic cards but having a single 1070, is it going to be faster than cpu rendering? How much change would it make to add a second 1070? Is there a way to compare cpu vs gpu render time according to my setup?

Thanks
Umut

Hey Umut,
Because GPU and CPU Render Engines are quite different from each other and normally one can’t render the scene input of the other, it is difficult to compare these two.

In GPU Render engines a huge advantage that you have is the nature of the Render Engine. It usually gives you immediate render updates, so you don’t have to wait for a few minutes to see how your change has affected the scene, but you get an immediate (lower quality) preview of the later final render.

This lets you work more efficiently and lets you iterate much more often, which usually increases the quality of the final outcome a good bit.

If you had to compare GPU to CPU I would say the 1070 is about on the same level as an i7 8700K performance-wise, maybe even a bit faster.

Adding a second 1070 will give you nearly 2x the performance.

Hope this helps,

Andreas

Hi Alex,
Thank you for sharing your knowledge. Thanks to you I understand a lot more of computers know.
I built a workstation a while ago with the help of your website.
Octane runs very fast and I’m happy but the screen freezes often. And sometimes the screen gets black for a second (Screen losing very short the signal but in normal use, not while rendering, maybe weak power supply?). Now I thought, maybe I did something wrong with the hardware:

4 x ASUS TURBO GTX1080 Ti 11GB GDDR5X

Intel Core i7 6850K (LGA 2011-v3, 3.60GHz, Unlocked)

4x Corsair Vengeance LPX (2x 2 Set, 16GB, DDR4-3200, DIMM 288)

ASUS X99-E WS/3.1, CEB, LGA2011-v3 (LGA 2011-v3, Intel X99, SSI CEB) (x16/x16 , x16/x16/x16 , x16/x16/x16/x16) (Do this work in x16/x8/x8/x8 or not? Maybe this is a mistake?)

Dark Power Pro 11 1200 W (Maybe to weak?)

Did i something wrong?
And is it possible to replace the CPU with one it’s a bit better for simulations in my case?
Thanks for your help!

Best regards from Switzerland
Andreas

Hey Andreas,
When I put your specs into a PSU Wattage calculator, it shows me you will need about 1196 Watts. So it can very well be that the 1200W PSU is not enough. Easy to test this though, just take a few GPUs out and work on the Computer for a few days and see if the problem persists.

The Screen going black though doesnt sound like PSU issues to me. Maybe a cable or plug not all the way in. Test with different monitor, less GPUs, other GPU as the display driver.

Check with GPU-Z if your Cards are too hot. Newest GPU Drivers, and maybe try with a clean install. (On a different Drive if you have one lying around)

6850K has 40 pcielanes and the Mainboard has some kind of PLX chips, so a Quad GPU 16x support is there. (It says so in the manual)

What type of workloads to you have in Simulations? Are you using realflow?

Antonios

Hi Alex, thank you so much for all this info.

I am a Vray user and recently have decided to make an upgrade to my old pc but I’m thinking this time to start using gpus for rendering

old pc config:
-i7 5820k 3.3GHz
-MSI X99S MPOWER (MS-7885)
-Crucial Ram 32gb 4×8 ddr4-2400
-Quadro 2000 1gb
-Tesla c2075 6gb
-PSU Corsair 850W

I’m thinking to add 2xGTX 1080 Ti with the existing PSU or 3x (3x maximum for this mboard)
with a bigger PSU and save some money of the total budget buying software, extra storage and a new 4k monitor

I’m still confused about the set up of the gpus. If I stacked 2 gpus will I take double the amount of Vram and if so should I buy an sli 2way for this?
Works also for 3 and 4?

At the end of the day you think this setup worth it for the time or should I make a new one with AMD Threadripper?

Hey Antonios,
Yes you can stack 2 or more GPUs, each in its own pcie slot. The render speed will double with two, near triple with three and so on. The VRAM does not stack though unless you use NVLINK that actually doesn’t seem to be working on most of the apps.

You will have to consider, that gpus have pcie-lanes that are responsible for transferring data from the gpu to the cpu. And the different CPUs have a certain amount of pcie-lanes available. If you need too many pcie-lanes for your gpus then the cards will become slower.

So the i7 8700k for example has 16 pcie-lanes for gpus. If you have two gpus, each gets 8 lanes.

If you want more gpus you will need a cpu that has more pcie-lanes available such as a Threadripper. It has 64 PCIE-Lanes. you could have 4 GPUs running and each would have 16 pcie-lanes to use. (Though such Mainboards almost don’t exist)

To make it easier (since it can be quite complicated) go check out the 3d Rendering Builds at the end of this Article here. They are ready to go and the parts work nicely together:

Joe

Hello, I work at a company that will be building a beastly rendering box, but I’m having a hard time finding a mining sever rack case to house all seven 1080 ti’s and the WS X299 Sage MB. The main problem is the motherboard’s CEB Form Factor (12 inch x 10.5 inch). Most mining cases only support a max of MB size of ATX.

Hey Joe,
Yes I understand. Though I can’t think of one from the top of my head. What type of Riser Cables will you be using and at what pcie-lane speed will these be running at?

I Haven’t decided on which riser cables just yet. The motherboard spec sheet states for seven gpus the configuration will be 16x for the first and 8x for the six remaining slots.

gaston

Hi Alex, you´ve been a great source of inspiration while building our new Octane rendering machine. My question is about how much rendering power I am actually loosing right now that my two 1080TI are running at 8X due to our 8700K setup?. I mean, does it worth it to go for a 7900X?. And if your answer is yes: will I have to also change mobo (asus z370 prime A 1151 socket).
Well thanks in advance for you reply!

Hey Gaston,
No you are loosing almost nothing. Having cards run at 8x instead of 16x makes almost no difference. As you can see from the Graph above, the difference is in the 1% area. Having the cards run at 4x though won’t be good.

Cheers

D3Pixel

This seems to be in dispute. Puget systems ran several x8x8 vs x16x16 tests and found certain apps would lose up to 30% performance on x8x8, https://www.pugetsystems.com/labs/articles/Titan-X-Performance-PCI-E-3-0-x8-vs-x16-851/#Conclusion

Highly optimized multi-gpu software appears to be more affected.

Bart

Amazing source of knowledge, thanks! I’m putting together a PC (moving from Mac) built for C4D, Octane & Redshift, AE and a little bit Houdini. The problem is obviously CPU. I decided to go the highest clock possible and low cores (not ideal for Houdini) but at the same time wanna keep PCI lanes’ number high. I was thinking that Intel Core i9-7900X 3.30 GHz LGA 2066 will be a good compromise between fast viewport (3.3 GHz / 4.3 GHz and a decent amount of PCI lanes (44).. What do you think?

Here’s complete build:
Corsair CC-9011030-WW Carbide Series Air 540 ATX
ASUS WS X299 SAGE Intel X299 DDR4 S-ATA 600
Intel Core i9-7900X 3.30 GHz LGA 2066
NZXT Kraken X62 (280mm)
Samsung 970 EVO 1 TB V-NAND M.2 PCI
Corsair CMK32GX4M2A2400C16 Vengeance LPX 64 GB (4 x 16 GB) DDR4 2400 MHz XMP 2.0
EVGA GeForce GTX 1080 Ti FTW3 HYBRID GAMING, 11GB GDDR5X, HYBRID (start with one and scale up to 4 over time — budget thing)
BenQ PD3200U 32 inch

Hey Bart,
Thats a solid Build. Nothing I can find wrong with it! What PSU are you planning on getting? I recently bought a Corsair AX1600i for my Threadripper + 4 1080Tis Build (Rendernode only). Runs great! Can recommend!

Max

Hi, Alex!
At first, I want to tell you many thanks for source! I found a lot of useful information.
All time i worked on laptops and now i want to buy realy good workstation for After Effects, Cinema 4D, Octan, Redshift.
I try to find compromise between good speed in viewport and not bad rendering in octan with maybe 3 or 4 GTX 1070ti or 1080. I understood that its almost impossible, and i should choose between viewport or rander, but maybe you can advice some options? My budget about 4000 usd.
I need your advice about CPU, GPU, MoBo and cooling fort that .
Many thanks for advance!
P.S. Sorry for my English) it not my native language)

Hey Max,
If you are doing lots of AE, C4D, and GPU Render Work with only 1-2 GPUs I’d recommend a i7 8700K or a Ryzen 2700X. They have good active working performance, good Turbo-Boost, and can also crunch some numbers if they have to.

After Effects especially likes High-Core-Clock CPUS since Adobe got rid of the Multi-Frame Rendering Support. 🙁

Check out the AE Build on this site and the 3D Modeling and Rendering Builds here for more Info (and finished Builds too!):

https://www.cgdirector.com/best-computer-for-after-effects/

Bouri

Hi, Alex
I’ve been using octane on a laptop with 16gb ram and a 1050 ti and i’ve decided to upgrade too a proper set up since rendering is taking forever.
Now i’ve asked a friend whom knows a bit more about pc building and he gave me a list of parts to get. However he’s not very knowledged in good render pc’s so i’d like to ask you if there’s anything is could/should change to this list.

AMD Ryzen 7 1700x
Bitfenix ENSO case black
Bitfenix Spectre Pro RGB 120 fan
Corsair Builder CX 850M Power
G.skill DDR4 Fortis 2x16gb 2400Mhz Ram
MSI x370 GAMING PRO CARBON Motherboard
Samsung 970 Evo 250GB SSD
Seagate HDD 3.5 4tb HDD
VGA Gigabyte Geforce GTX 1080 Windforce OC 2x graphics card

Thanks in advance!

Hey Bouri,
Yes those are good specs. If you have an extra 100€ you can also take a look at the Ryzen 2700x which is the newer generation of the 1700X and worth it in many cases. You might need a Bios Update though for that Mainboard with a 2nd Gen Ryzen.

Aayush

Hi Alex,
I am a upcoming interior designer so I am wanna built a PC for 3DS Max, Google sketchup with V Ray rendering . I decided to choose Intel i7 8700k processor , Nvidia 1080ti GPU ,16 gb of ddr4 ram and 256 gb SSD. So is this spec is good ? My budget is $2,200 to built whole PC . And also which mother board and cooling fan I should choose in perspective to upgrade in future
_ Aayush

Hey Aayush,
Those are good specs, yes! For Mainboard, it depends a bit on the features you want. I would go with a z270 chipset, such as the Asus prime z370-A or a MSI one. CPU cooling you can take a look at Noctua and beQuiet, as these are great and quiet coolers. There won’t be much of an upgrade possibility with the Intel CPUs though. If you absolutely want to be able to upgrade your CPU without changing your motherboard, you will have to go AMD.

Mayu

I want to build faster pc for after effects , premiere pro, 3ds max rendering (vray) , cinema 4d and element 3d. I am just beginner so please let me know the most fast PC (CPU, graphic card , motherboard ) for $2500-$3200 budget. Please list the all equipment that you recommend so that I can buy while I am in VA near Microsoft store. Thank you!!

Hey Mayu,
Take a look at the Cinema 4D PC and After Effects articles on this site. You’ll find your component suggestions there!

Geoff Coope

Hi Alex

Thanks for this post. I am currently teetering on getting a single slave or just upgrading my current machine (4700K + 1080Ti + 1070Ti, 32GB Ram) for use with GPU rendering in Max with either Vray RT or Redshift.

A second machine would free up my main machine when rendering in the day but has the added cost. So not decided which way to go on that front yet.

Anyway, my question. I need a motherboard / CPU that can scale with my work demands so looking for a CPU + Motherboard that can take 4 GPU’s but initially will only use 2. Any recommendations on current MBrd/CPU combo hardware available today? Or is the Sage still a good choice?

Hey Geoff,
Actually I am currently looking into this myself. Using a fast Computer to work on (high-core-clock) but need a high PCIE-Lanes CPU to render on, getting a second PC is somewhat inevitable.
I think the best choice currently is the Threadripper 1900X that has 64 PCIE-Lanes and together with (for example) the MSI X399 Gaming Pro Carbon can take on 4 GPUs (2 in 16x Mode and 2 in 8x Mode). The TR4 Socket will also support upcoming Threadripper generations, so a big plus in terms of upgradeability. The Sage is interesting when using Intel CPUs, since the Intel CPUs have less PCIE Lanes. But with Threadripper you don’t have this problem anymore. And running GPUs on 8x does not make a difference to 16x.

Geoff coope

That’s good to know. The new tread ripper 32 core sounds great too but not sure it needs a new mb. Thanks for reply! Will look into the motherboard

Jarratt Moody

Enjoying all of your wonderful articles and extremely handy price/performance charts.

Looking for best of both cpu/gpu worlds with high processor speed of the 8700k along with a couple 1080tis running at 8x pcie. I do mostly motion graphics using Cinema 4d and After Effects and have started using Octane as my primary 3d renderer. Does that sounds like the perfect setup? Any other pointers if so/if not would be hugely appreciated. I guess I9 7900x could be better more cores, still 4.5 ghz turbo, and allow for more gpus.

Thank you!!

howie

Really great article, it’s helped me a lot to start to understand whats what. I’m switching from mac to PC so I need all the help I can get!
Some of the replies here are a little bit old now so I was wondering if there have been any developments in terms of the best CPU to run 4 GPUs? Looking into it it seems like the i9-7900X is a good option and the i7-6850K is maybe even better whilst being cheaper. Is there anything else to consider between these 2 CPUs or another one that you think is currently better suited to the task?

Hi Howie,
The Computex is currently going on and Intel will release some new HEDT CPUs by the end of the year. These though will be quite expensive, so I would say the i9 and Threadrippers are still the best choice for high PCIE-Lane CPUs

cheers
Alex

Thanks for the reply Alex. So if this machine is purely for GPU rendering and C4D, would you say the i9 is better than the i7 6850k?

Teno

Hi Alex,
I’m so grateful for all your help,
Unfortunately i couldn’t find any good mobo which takes 4 gpu’s as x16 in the country i’m living in at moment,
also couldn’t find EVGA 1080ti hybrid,
decided to go with x2 1080ti gpu’s on asus prime x299 delux, with i7-7820x
there are so many 1080ti versions [not hybrid] , do you recommend any alternative than evga hybrid ?
appreciate your time and effort

also why to go for x299, and not z370 ?

Marc

Hello Alex,
Thank you very much for sharing your knowledge. Very usefull.

I would like to ask you one thing, because I’m going crazy and I’m desperate with my computer problems.
I’m recently using a new computer (I bought from a company specialized in 3d work), with x4 1080ti – Windows 10 – Intel Xeon E5-1650… . I’m using Cinema 4d and Octane and some weeks ago I’ve been experimenting a lot of blue screens, screen freezes, render crashes,etc.. I can’t render a scene by night because always crashes.

I’ve noticed in Octane forums that a lot of people have the same problems, but nobody have a strong solution. I’m trying a lot of drivers versions, and the last method I read was to change the TDR values of the registry, changing 8s by 60s… like this manual: https://support.allegorithmic.com/documentation/display/SPDOC/GPU+drivers+crash+with+long+computations

After changing the TDR values my render don’t show “render crash” but it’s worse, because it stays frozen with no feedback…
So my problems persists, and I’m really desperated…

1. I don’t know if it’s a hardware problem, and Octane poblem or Windows problem…what do you think?
2. Do you recommend me to use Windows 7?? I think it’s my last option…

Thank you very much,

Ziee

Hey Alex,

Thank you for the nice article.
As an architect I will use my computer for complex modeling plus rendering and I guess that needs a processor with max of cores (the more the better).
The problem is the more cores the less core-clock CPU.
What do you think and suggest dear Alex?

Best regards,
Zied

Hey Ziee,
You are absolutely right. The more Cores in a CPU the lower it usually clocks. For modeling I would recommend a high clocking (lower core-count) CPU such as the i7 8700K, but for fast rendering you will need as many cores as possible, such as a Threadripper 1950X. This truly is THE question that we all have to think about. What is more valuable for us. The time we actively work or the time it takes to render. Usually it is the time you work on stuff. Then again, its not like you work twice as slow with a Threadripper 1900X vs a i7 8700K. Its marginally snappier, you might notice differences in a range of 5 – 10%. Tough decision, I know! Get two! 🙂

Cheers!

Adam

Hi Alex,
I am looking for a workstation configuration for redshift and octane in combination with C4d. I am planning to go with 4 gtx1080ti’s or free with any other good option. Could you please guide me with the configuration of cpu and motherboard and Gpu’s also.
Thanks in advance
Adam.

Hey Adam,
Look into Blower Style GPUs such as the Asus 1080 TI Turbo. These blow out heat at the back and not on to other cards. If you can swing, get watercooling or hybrid watercooling solutions. CPU either a Threadripper or a 2066 socket type intel CPU with maximum PCIE-Lanes would be best. There is one Asus Sage WS X299 Board that has PLX chips for additional PCIE-Lanes on it. You could use four x16 cards on this one. Lots of other boards have enough pcie Sockets and Lanes for four 8x GPUs. As stated in the article, check the specification info on the manufacturers Website and it’s best to look into the manual too, to be sure!

Silas

I recently built a 1950X machine for multiple purposes (VRay RT, AE, Premiere Pro, Audio):
1950X
1080Ti & 1080
64 GGB DDR4

The problem is that I found out now that the 1950X has some problems with the navigation in After Effects (and it has problems with an audio tool I often use). I really want to switch to a 8700K. The only thing that held me back was the thought of not being able to render at the same speed in 3D. But after reading your post, the 8700K would be the better solution for GPU renders anyways (since I’m mainly using VRay RT), am I getting this right?

What’s your recommendation? Should I stick with the 1950X? I don’t think Adobe is going to enable Multicore on AE again soon… oh and I plan to use 3 1080Ti sometime in the future.

Thank You!

Hey Silas,
This is in fact one of the big questions in the 3D Hardware world. Get a high-clocked / low core count CPU to work fast (Good for working in c4d and ae), but CPU render slow, or get a high core count / lower clock speed CPU that usually has more pcie-lanes for more GPUs, that you might work slower on, but you can render fast on the CPU can build in more GPUs, though the many GPUs actually need a higher clock speed.

In a perfect world you would have a 16 Core CPU with 4,5 GHZ on all cores, that supports 64 PCIE Lanes. Since this unfortunately is no option you have to either buy two specialized pcs (one for working and one for rendering) or think about what you will be doing the most of, what time is most valuable to you (working or rendering?).

That said, yes you would be able to run your 2 GPUs on a 8700K on both 8x mode. You do have to check your Mainboard specifications though, because many Mainboards will steal PCIE Lanes for e.g. M.2 Slots or SATA Slots.

I myself have a high-clock cpu (i7 7700K) for working and am using render farms for when the projects get too big to render locally, but am thinking about getting a Threadripper for local CPU Rendering.

In terms of AE Rendering after Adobe pretty much broke the Multi-Frame Rendering: A Tip is to use a render manager such as Thinkbox Deadline, that is free for up to 2 PCs. You can set it up in a way, that your pc renders multiple concurrent tasks and so utilizes all of your cores on the Threadripper. (This only works though when rendering Image sequences, not for Film/Video-Rendering)

As for your future Plans 3x 1080TI: The 8700K only has 16 PCIE-Lanes, so no luck there. You will need a i9 or Threadripper to fully utilize the GPUs.

jeremy

Hi Alex, I was curious if you had seen any testing in regards to the new Titan V with redshift?

Cheers,

Hey Jeremy,
Unfortunately not yet. I am looking forward to these too! Though considering the price and the orientation towards Machine learning I expect the performace / price ratio to be much worse than the 1080ti or a following 1180ti (or what it will be named).

Eddy

I’m upgrading my workspace by building up a new render station for Redshift and Octane. My issue is that im torn between wither going the Ryzen route or the new 79xx route. The reason being, I want to run either 3 1o8oti’s or 4 1080s at the best speed possible.

Looking at the require PCIe lanes available, i need your advice or what could potentially work best and still be future proof. I know i will get more PCIe lanes on ryzen i just havent jumped on it before.

Could you recommend some ideal compatible parts for a build for redshift, specifically on the CPU and MOBO side. Im building inside a Corsair Obsidian 900D.

Druid

Alex,
Great info! Question: Have you noted the impact of overclocking (GPU and/or CPU) with respect to the results in your charts? When comparing my configuration results in VRAY Bench to the ones listed on the website, I score in the middle of similarly-configured systems. If I run everything at stock speeds, I’m out in left field. I’d be interested in your take on this, if any.

Druid,
Vray Bench unfortunately does not record the Clock speed of the Hardware. I would suspect that, since there are so many different AIB Companies that all tune their cards a bit differently, and also custom overclocking, this most probably will be the reason for the differences in render time.

Also, I am not sure if the VRAY GPU Bench is the best bench to rely on, since the 1080 performs worse than the 1070 here on average.

What card are you testing and on what system?

Alex,
My system is not typical because I’ve done so much enhancement over the few years I’ve had it. And while it’s several generations older now, it still makes a good workstation for the small projects I do in 3DS Max.

It’s an intel i7-4960X based system on an Asus RIVE Black Edition MoBo. 64Gigs of 2800 DDR3 Memory. Processor overclock at 4.6Ghz (under water) as well as other things overclocked on the MoBo. The graphics cards are: 1-NVIDIA GeForce GTX 1080 Ti, and 2-NVIDIA GeForce GTX 1080’s running x16 / x8 / x8 on the bus. My best score on VRAY Bench is 2:28 CPU and 0:26 GPU. I want to add another 1080 Ti but I need to find a PCIe x8 riser/splitter (1 to 2 or more). MSI’s 1080 Ti card is very fat and makes it so you can’t put another card next to it on my specific MoBo. I’m not going to water-cool these things as I don’t run them that hard. I know it might be difficult, but I’d like to see a chart showing the speed differences in VRAY GPU Bench with 1 card running in a x16, x8, x4 and x1 slot compared. I really like the “Bang for the buck” approach your article features. Again, this is very good info you don’t see everywhere.

Jay

Hi Alex, first off – thank you so much for all this useful knowledge on this site, learning a lot about how I should go about building my first PC for Octane for Cinema 4D and some Aftereffects motion design.

I am torn between the i7 7700K and the 8700K, I’m planning to buy one GTX 1080Ti and add another one in the near future. If I was to go up to 2, and maybe even 3 or 4 1080Ti’s in the future, which MoBo should I consider?

Hey Jay,
There is no LGA 1151 (v2) Mainboard out there (i7 7700k or 8700k) that would support 4 1080 Tis at full speed (x16) since the CPUs are limited to 16 PCIe Lanes for GPUs. If you want something a bit more future proof in terms of extending it with multiple GPUs you will have to go in a 2066 Socket (and Mobo) Direction with the new Intel i9 type CPUs (that have 44 PCIe Lanes) or in a AMD Threadripper Direction (64 PCIE Lanes)

Tiago

First of all, great post, congrats!
My concern is about GPU’s, i want to build a rendering workstation for 3ds max + Vray 3.6 and in terms of graphic cards i’m planning to use a gtx 1070 for viewport and a gtx 1080 ti for render and add 1 or 2 more gtx 1080 ti in the future.
My doubt is, it is possible to have different cards in the same build? i’m asking this, because all the people that i see building this kind of machines with multiple gpu’s, uses the same cards.

Thank you in advance.

Hey Tiago,
Yes, this is possible. You could just use your 1080Tis for rendering by deselcting the 1070 in the render options. As far as I know octane, redshift and vray have this option.

Or you could have the 1070 render too, but will be limited by the lesser amount of VRAM on this card, also your viewport will get sluggish, if you decide to continue working while rendering.

The 1080Tis will run in headless mode (no monitor connected) meaning they will also have more VRAM free to use for the GPU Render engine.

Kurt jefferson macacando

Hi sir.. im planning to have my first ever cpu.. im planning to have a i7 7700k processor and a gtx 1070 8gb gpu.. and a 16gb ram.. can i have a fast rendering time for vray 3.0 sketchup 2017? Or a ryzen 5 1600 processor due to its 6 cores and 12 thread.. im a bit confuse.
Due to tight budget.. i really need your help for this..

Thank you sir!

Hey Kurt,
Check out this Post for CPU Rendering Speed: https://www.cgdirector.com/best-cpu-for-rendering/

Ryzen 5 1600 has great performance / dollar. The 7700K has very high single Core Speed, though its not as fast for multicore-rendering. If you are on a tight budget, taking the Ryzen route is a good choice. The Problem with the 7700K is, that you cannot upgrade this Socket to a better processor, since the next generation (the 8700K) already needs a new Mainboard and slightly different Socket.

1070 is a great budget choice of GPU.

Hippetuinen

7700K is the way to go with Sketchup, because it just uses a single core.

Polystorm

Hi Alex. Your blog posts have really helped me understand pc builds for working in Cinema 4D. I’m trying to do a build with the i7 7700k and 4x1080ti cards on PCPartpicker. I can’t seem to find information on a motherboard that would support this configuration. Any advice? Thank you in advance!

Hey Polystorm,
If your plan is to GPU render with these 4 1080tis, then you will want to look into getting a different cpu. The i7 7700k supports only 16 PCIE Lanes, meaning your 4x 1080TIs will run in 4x mode. That said though, it seems Intel does not offer support for 4x 4 Cards as you can see on this page in the section PCIE Configuration:

https://ark.intel.com/de/products/97129/Intel-Core-i7-7700K-Processor-8M-Cache-up-to-4_50-GHz

I recommend going the Skylake-X route that offer 28 PCIE Lanes on 7800X and even 44 PCIE Lanes on 7900X Series CPUs.

Thanks for the quick response! Interesting. I selected the 7700k on your recommendation for faster viewport speed. If I got the 7900x would overclocking solve that? Also, would the motherboards you recommended allow for 4 gpus?

It is interesting indeed! Unfortunately there is no CPU that satisfies all needs. You will have to make a tradeoff in some apsects. You will have to think about what is most important for you.

The tradeoff in a 8700K (or 7700K) Cpu vs a 7900X in terms of viewport performance is marginally. Considering you are thinking about getting 4 1080TIs that alone cost around 2500+$ I would look into making them run the best they can, even if you loose those 5% in Viewport Speed in heavy scenes, that you might only have rarely.

I don’t think I recommend a Skylake-X mainboard anywhere but it will have to be something like the MSI X299 GAMING PRO CARBON, which has 4x PCIE x16 support.

And you could of course overclock your CPU, but you’ll have to look into decent cooling and good RAM. Then of course you could also overclock a 8700K and get even more single core speed there. 😀

Okay thanks again. Time to remortgage!

Deb Pearl

My son has been looking for a new graphics card and I wanted to look up some information about them. That is great news that they have matured quite a fit over the last couple of years. That means my son could probably find exactly what he wants for his computer easier. Thank you for all the information! I will be sure to show my son.

Glad to help!

Frank Gomez

Thank you for your posts, they are very helpful. Two questions:

1. Regarding the i7-7700K. If I use 1TB Samsung 960 EVO M.2 NVMe PCIe in my build. Would this affect the PCI lanes for the GPU? Or would the GPU still run x16?

2. If I want to run 4 GTX 1080ti at full speeds (x16), would i have to move to the Xenon processors? What mobo should I use?

Thanks Again.

Frank

Hi Frank,
1. There are extra Lanes for the PCH (Platform Controller Hub) that manages all the additional Hardware such as M.2, Sata, USB and so on. This means you can use the 16 Full PCIE Lanes for your GPU(s) and still have extra lanes depending on the mainboard chipset to have M.2 at full speed running (usually x4)

2. You wont be able to run 4 x16 GPUs on a LGA 1151 Board with a i7 7700k CPU, that would require 64 PCIE Lanes. You can run them at x4 Speed, but this would reduce the speed somewhat and yes you should then switch to a 2011-3 x99 Board with a CPU that supports up to 40 PCIE Lanes.

ANDRES GARCIA CASTILLO

Hi, Aelx

Good Post!

A question if I want to have 4 gtx1080 in a corei7 6800k that would have lost for rendering with octane? Is insignificant? or it would be worth buying 6850k to be able to have the 4 gtx.

And what kind of cooling you recommend for these 4x gtx1080, thank you very much for your time!

Cheers,
Andres

Hey Andres,
Technically you would have 3 1080 at 8x pcie speed and one at 4x pcie speed. Meaning that last card will have some speed drawbacks. That said though, adding cards does not scale the rendering performance linearly, so the last card usually will not give you an added 100% speed but less.

With more GPUs there is also a higher strain on the CPU for managing these, so I would recommend getting the 6850K that has a higher clock-speed.

As for cooling, depending on your budget, you should look at either blower style cards (that stay cooler when stacked), hybrid cards (youll need lots of room in your case for mounting the radiators) or watercooling if you are up to it.

What is your take on noise? 4x Founder Edition Blower style cards will be quite noisy, hybrids a bit better, and the water cooling will be the quietest.

Let me know what you think.

swaroop

Hi, Aelx
Can i use 4 GPUs (my preference 1070) with 7700k? for octane.
If can, then which Motherboard to choose.
Thanks

Hey swaroop,
You can, but those 4 cards will only run at 4x pcie speeds, meaning quite a decrease in speed. Although there is almost now drawback in 16x vs 8x pcie, there unfortunately is when running the cards at 4x pcie speeds.

For 4 gpus I recommend getting a i7 6850k that has 40 lanes and you will be able to rund 3x 8 pcie and one in pcie x16.

Also consider getting the 1080 since they have dropped in price quite a bit recently. You could get 3x 1080 in 8x pcie and get a 6800k that has 28 pcie lanes.

At current prices 4x 1070 = ~1600$ and 3x 1080 =~1500$. Both combinations will have roughly the same rendering speed.

Best Hardware for GPU Rendering in Octane – Redshift – Vray (Updated)

Best Hardware for GPU Rendering

Processor

Best Memory (RAM) for GPU Rendering

Best Graphics Card for Rendering

GPU Cooling

Power Supply

Mainboard & PCIe-Lanes

Best GPU Performance / Dollar

GPU Benchmark Comparison: Octane

GPU Benchmark Comparison: Redshift

GPU Benchmark Comparison: VRAY-RT

Speed up your Multi-GPU Rendertimes

Buying GPUs

Redshift vs. Octane

Custom PC-Builder

178

Comments

Popular Posts

Post Categories

Best Hardware for GPU Rendering

Processor

Best Memory (RAM) for GPU Rendering

Best Graphics Card for Rendering

GPU Cooling

Power Supply

Mainboard & PCIe-Lanes

Best GPU Performance / Dollar

GPU Benchmark Comparison: Octane

GPU Benchmark Comparison: Redshift

GPU Benchmark Comparison: VRAY-RT

Speed up your Multi-GPU Rendertimes

Buying GPUs

Redshift vs. Octane

Custom PC-Builder

Popular on CGDirector Right now!

178 Comments

Popular Posts

You May Like

Post Categories

178

Comments