by Alex ⋮ 
Redshift Benchmark results are not easy to find, that is why I compiled a list for everyone to take a look at. I myself use Redshift for Rendering Projects on a daily basis and recently purchased a new Rendernode with lots of GPUs.
When you build a new PC or Rendernode for Rendering, there is no way around digging deep into benchmarks.
Cinebench Scores are great for comparing CPU Rendering performance, but if you use Redshift heavily, these benchmark results will be invaluable:
Redshift Benchmark Results
| GPU(s) | OS | CPU Threads | CPU GHz | Time (Minutes) |
|---|---|---|---|---|
| GPU | OS | CPU Threads | CPU GHz | Time |
| 8x GTX 1080Ti 11GB | Windows | 32 | 2.1 | 01.57 |
| RTX 2070 8GB | Windows | 32 | 2.1 | 12.50 |
| 4x GTX 1080Ti 11GB | Windows | 32 | 2.1 | 03.25 |
| 8x RTX 2070 8GB | Windows | 32 | 2.1 | 01.56 |
| 8x GTX 1080 8GB | Windows | 40 | 2.2 | 02.26 |
| 4x GTX 980 Ti 6GB | Windows | 56 | 2.3 | 04.48 |
| Quadro K5200 8GB | Windows | 24 | 2.6 | 38.50 |
| 2x GTX 1060 6GB | Windows | 32 | 2.6 | 11.07 |
| RTX 2080 8GB | Windows | 36 | 2.6 | 10.59 |
| RTX 2080 Ti 11GB | Windows | 36 | 2.6 | 08.38 |
| GTX 1080 8GB | MacOS | 8 | 2.9 | 16.00 |
| 4x GTX 1080Ti 11GB | Windows | 36 | 2.35 | 03.21 |
| GTX 960 2GB | Windows | 4 | 2.68 | 44.56 |
| 8x Tesla V100-SXM2 16GB | Linux | 64 | 2.74 | 01.03 |
| GTX 780 3GB | Windows | 8 | 2.81 | 31.05 |
| Titan V 12GB | Linux | 12 | 2.90 | 07.06 |
| 2x GTX 1080 Ti 11GB | Windows | 16 | 2.99 | 06.15 |
| 2x RTX 2070 8GB | Windows | 16 | 3.0 | 06.29 |
| Quadro RTX 6000 | Windows | 20 | 3.1 | 08.47 |
| 2x GTX 1070 Ti + 1x GTX 1070 | Windows | 12 | 3.2 | 06.12 |
| 2x TITAN Xp 12GB | Linux | 16 | 3.2 | 06.16 |
| 2x GTX 1080 8GB | Windows | 12 | 3.2 | 07.26 |
| GTX TITANX 12GB | Windows | 32 | 3.3 | 17.33 |
| 1x GTX 1070 + 1x GTX 1070 Ti | MacOS | 12 | 3.3 | 08.17 |
| GTX 780Ti 3GB | Windows | 12 | 3.4 | 25.58 |
| GTX 1080 Ti 11 GB | Windows | 12 | 3.4 | 11.44 |
| Quadro M6000 12GB | Linux | 12 | 3.5 | 14.23 |
| GTX 980 4GB | Windows | 12 | 3.5 | 23.21 |
| GTX 1070 8GB | Windows | 12 | 3.5 | 17.11 |
| Quadro M6000 24GB | Windows | 12 | 3.5 | 15.42 |
| GTX TITANX 12GB | Windows | 12 | 3.5 | 15.40 |
| Quadro M6000 24GB TCC | Windows | 12 | 3.5 | 14.33 |
| GTX 970 4GB | MacOS | 12 | 3.5 | 27.22 |
| GTX TITANX 12GB | MacOS | 12 | 3.5 | 14.25 |
| 2x Quadro P6000 24GB TCC | Windows | 12 | 3.5 | 05.55 |
| Quadro P6000 24GB | Windows | 12 | 3.5 | 12.30 |
| GTX TITANX Pascal 12GB | Windows | 12 | 3.5 | 11.59 |
| Quadro P6000 24GB TCC | Windows | 12 | 3.5 | 11.31 |
| 4x GTX 1070 8GB | Windows | 12 | 3.5 | 04.15 |
| 2x Quadro GP100 16GB TCC | Windows | 12 | 3.5 | 04.25 |
| GTX TITANX Pascal 12GB TCC | Windows | 12 | 3.5 | 10.54 |
| Quadro GP100 16GB | Windows | 12 | 3.5 | 09.57 |
| 2x Quadro GP100 16GB | Windows | 12 | 3.5 | 05.04 |
| 2x Quadro P6000 24GB | Windows | 12 | 3.5 | 06.28 |
| Quadro GP100 16GB TCC | Windows | 12 | 3.5 | 08.43 |
| RTX 2060 6GB | Windows | 8 | 3.6 | 15.04 |
| RTX 2060 Super 8GB | Windows | 16 | 3.6 | 12.17 |
| GTX 1060 6GB | Windows | 8 | 3.6 | 23.11 |
| GTX 1660Ti 6GB | Windows | 8 | 3.6 | 24.53 |
| RTX 2070 8GB | Windows | 6 | 3.6 | 11.35 |
| 4x GTX 1080Ti 11GB | Windows | 12 | 3.6 | 03.07 |
| RTX 2070 Super 8GB | Windows | 16 | 3.6 | 11.17 |
| 2x RTX 2080Ti 11GB | Windows | 16 | 3.8 | 04.28 |
| GTX 1070 8GB | Linux | 12 | 3.8 | 15.05 |
| 4x RTX 2080Ti 11GB | Windows | 16 | 3.8 | 02.28 |
| GTX TITAN X 12 GB | Linux | 12 | 3.78 | 14.07 |
| 2x GTX 1070 8GB | Linux | 12 | 3.79 | 07.39 |
| 2x GTX 1070 8GB | Windows | 8 | 4.0 | 08.06 |
| 2x GTX 1080 Ti 11GB | MacOS | 8 | 4.00 | 06.30 |
| 4x GTX TITANX 12GB | Linux | 12 | 4.1 | 03.36 |
| 2x GTX 980Ti 6GB | Windows | 20 | 4.3 | 07.35 |
| 4x GTX TITAN X 12 GB | Windows | 16 | 4.3 | 04.28 |
Some interesting findings are, that Different OS’s have an impact on rendering performance.
Linux, for example, seems to be doing somewhat better than Windows. These benchmarks are also great in analyzing what CPU is best for GPU Rendering in Redshift and how low the Clock Speed can go before major performance impact is inevitable.
Multi-GPU Scaling seems to work quite well in Redshift, though with those 1-2 minute Results the benchmark probably is just too short to make good use of all that power.
This list is also available on the Redshift Forums, where some of these Scores originate from, so head on over if you have an account and get a good talk with all of the great Users.
If you are thinking about getting a new GPU Rendering Computer altogether, be sure to check out this article on what hardware to get for the best GPU Rendering Workstation.
Missing any GPU Benchmark Results or having trouble with one of your builds? Let me know in the comments.
Hello, Alex, I am a student who occupies a new system to render a short film of at least 5 minutes, there are currently rumors of the new Nvidia architecture, which makes me think if it is better to wait or invest in a new PC, I would appreciate it in advance a tip as to the investment that I plan to make and what kind of equipment I could acquire, my only concern in general would be the temperature of the cards when having to work continuously, since only a single equipment would be available to render. I have at least $ 4000 available.
I plan to render in Maya, I have an i7 6700K CPU, GTX 1070, 32 RAM 3200 and in Windows. Thank you
Hi José,
Thanks for dropping a line!
True, there are rumours about the upcoming 7nm NVIDIA Ampere architecture but these won’t get launched until late March, if the rumours are to be believed. However, the first Ampere products alleged to get released are thought to be for the datacenter and will be AI-focused. Once that happens, we might see the introduction of RTX 30-series cards. Sad to say, there are no specifics on that just yet.
If you want to wait for the next generation of NVIDIA GPUs, that’s one option for you to take. If you want to build a workstation now, we can come with a build for you with a great upgrade path. You can then just opt to upgrade the GPU to an RTX 30-series card once they become available.
Here’s a powerful build the won’t even use up your full $4,000 budget:
Parts List: https://www.cgdirector.com/pc-builder/?=Db1Gf0i0jib
CPU: AMD Ryzen 9 3950X 3.5GHz 16-Core Processor ($749.00)
CPU Cooler: be quiet! Dark Rock Pro 4 AM4 ($89.90)
Motherboard: Gigabyte X570 Aorus Elite ATX AM4 ($179.99)
GPU: NVIDIA RTX 2080 TI 11GB – MSI Gaming X ($1249.00)
Memory: 64GB (4 x 16GB) Corsair Vengeance LPX DDR4-3200 CL16 ($302.54)
Storage PCIe-SSD: Samsung 970 EVO PLUS 1TB M.2 Solid State Drive ($199.99)
Power Supply: Corsair RMx Series RM650x 650W ATX 2.4 Power Supply ($109.99)
Case: Corsair Carbide Series 275Q ATX Mid Tower Case ($89.98)
This build will set you back $2970.39 which is considerably lower than your $4,000 budget but it doesn’t necessarily mean that the performance will also be lower. In fact, it’s the opposite because this build gets you nothing but the best components.
It has a very powerful 3950X CPU coupled with 64GB of RAM to ensure that your workflow is always fast and smooth. Not only that, the high core count of the 16-core 3950X will definitely deliver a great Maya CPU render performance. No worries on GPU temperature though as the MSI RTX 2080 Ti Gaming X Trio comes with three (3) fans to help with heat dissipation. All in all, this is a beast of a build that will handle any task you throw at it with ease.
Cheers,
Alex
Hi Alex. I have a GTX 1660 ti but it doesn’t work with my octane 3.07 and I got this error “there is no CUDA device which is selected” and unfortunately I have no budget to buy a higher octane version like Octane 4 that may solve the problem of CUDA. So do you think that might I have to overcome this problem with RedShift engine? Does Redshift support GTX 1660 ti? Thank you in advance for any advice!
Hey John,
Octane supports the 1660ti and the 1660ti also has cuda cores. You will need Octane 3.08 at a minimum, better octane 4 to make it run though. Also be sure you have the newest nvidia drivers installed.
Redshift too supports the gtx 1660ti.
Cheers,
Alex
Thank you so much Alex!
Hi there,
an extremely interesting topic. I am currently planning a new workstation build and generally I am pretty certain what tech I need. But what I cannot guess at this point is the following. If I should get a single 1080Ti vs 2070 or higher model.
Basically, how much and if the future integration and optimization of OptiX into Redshift, Blender (Cycles, EEVEE), Houdini GPU accelerated calculations, etc. will boost significantly the performance of the RTX lineup. Can you comment on that?
Regards,
Ivan
Hey Ivan,
You can test the Redshift beta with integrated RTX optimizations already. There are cases where it is 40% faster, some cases where it is barely faster. All depending on how your scene is set up.
If your scene is heavy on:
– Shading (complex shaders, lots of layering, blending etc)
– Volumes (lots of VDBs)
– Particles or hairs (lots of primitives)
– Deformation blurring (lots of deformed objects in the scene)
– Point-based techniques like irradiance point or photons or point based SSS
… then you will probably not see big speed gains. It doesn’t mean it will render slower but you might see little to no benefit.
1080Ti vs 2070 of course also is a question of vram, though in general I find redshift to perform well even if vram is full and it has to outsource to system ram.
Also, it depends on how much you are paying for the gpus, as the sweet spot currently is the 2070 super, and that gpu is slightly faster than a 1080ti and quite reasonably priced usually.
Personally I’d go with the RTX generation unless you are getting a great deal on the 1080Ti. I wouldnt go higher than 2070 super, as the performance/price gets a lot worse higher up. Better to get a second or more 2070 supers to add to the build.
Cheers,
Alex
Thanks a lot! Yes the question was related to the fact that I have an offer for 1080Ti at 50% the price of 2070 Super, therefore I needed an estiamtion of the RTX capabilities to get faster with the future OptiX adoption.
I found a more detailed andwer here:
https://techgage.com/article/testing-blender-2-81-alpha-with-nvidia-rtx/
Regards,
Ivan
Hey Ican,
Given that RTX Rendering in Blender is still in the early stages and already showing a good speedup through the line, I could fathom it becoming even more potent in the future, while with a 1080ti you would be stuck on the “older” path. 50% of a 2070 super, though is quite a nice price, I must say. Benchmark scenes of course aren’t always the best indicator of real world performance, so if you know what type of scenes you usually render, it might be interesting to do some testing on those. Feel free to send me a scene of yours if this is something you would want me to test.
Cheers,
Alex
Hi, Alex. I’ve just started out with 3d animation, but mostly I work on modeling characters and environments. I wanted to know, what options do I have if I wanted to buy a laptop to work on my projects. My budget would be between 1000 to 1500 $
Hi Akesh,
Thanks for asking!
For a budget of around $1,500, the best laptop you can get is the MSI GF63 Thin Gamer Notebook which goes for around $1,479.99. Below are the specs of the MSI GF63:
CPU Intel Core i7-9750H 2.60GHz 6-Core Processor
Graphics Card NVIDIA GeForce GTX 1650 4GB
Memory(RAM) 32GB DDR4-2666
SSD 2TB NVMe Solid State Drive
HDD 2TB 5400rpm Hard Disk Drive
Weight 1.8 kg (~3.96pounds)
Display 15.6″, 1920×1080
For it’s price of $1,479.99, the MSI GF63 Thin Gamer Notebook gives you the most bang for your buck, so to speak, when it comes to its specs. You basically get an i7-9750H CPU working together with a whopping 32GB of RAM so you can expect a faster and smoother workflow. Not only that, this laptop gives you 2 storage options – one 2TB NVMe SSD and one 2TB HDD. Given all those specs, the MSI GF63 Thin Gamer Notebook is a bargain for its price and I highly suggest that you go for it!
And in case you want to see what other options you may have, please see the site’s PC Builder Tool at https://www.cgdirector.com/pc-builder/. Input your budget and use case and the tool will give you the best recommendations.
Cheers,
Alex
Hi! We have just bought redshift in our studio, and I am testing in on Autodesk Softimage. I have RTX 2060 Super 8GIG and a GTX 970 4GIG.
My problem is that my test renders only give a 30 percent increase in renders speed, from 970 to 2060.
Can it be possible? Why? According to the benchmark results above, the render speed difference should be more than double.
Any ideas?
Thanks a lot!
Robert
Hey Robert,
Thanks for asking!
To answer your question, it’s possible to only get a 30% increase in render speeds when comparing an RTX 2060 Super to an older GTX 970 GPU. GPU render speeds depend on a lot of factors and it doesn’t necessarily mean that you’d get the same results as that of the Redshift Benchmark Results.
For example, the OS alone already plays a big role in GPU render speeds. The GTX 970 was on a MacOS system while the RTX 2060 Super was on a Windows machine. Your CPU and its clock speed also play a part so you also have to take that into consideration. The scene you’re trying to render is also something you should look at. At the end of the day, you can’t expect the same performance from benchmark results because of the differences between the machines used in the benchmarks and your system.
Two more things you should check is how long does the frame you are rendering prepare? Because only the bucket rendering phase really gets a speedup with stronger cards. Also – how many pcie-lanes are available for your 2060? At least x8? How does it perform with it being the only GPU in your system?
Cheers,
Alex
Y porque no hay información sobre las rx de amd?
Hey Marco,
Redshift runs on CUDA capable GPUs only, which is only supported by Nvidia.
Cheers,
Alex
Hi Alex, Here i am sharing my existing PC config details. if i want to replace with new GPU RTX 2080 ti with existing PC. What all parts would be compatible to use further and what all should i consider to replace to get best performance for my day to day Houdini with redshift rendering use.
Bellow is my Existing PC config. in that, i like to replace GPU card RTX 2080 ti:
– CPU – i7 – 5820k 6 Core
– RAM – G.SKILL – 2400Mhz – DDR4 Ram (16GB x 4 = 64GB) ( Quad Channel )
– .Mother Board – As-Rock x99e-itx – with 4 Ram Slots o9n board
– SSD Hard Drive – 250 GB ( for OS and Softwares )
– 2tb sata Seagate ( for Storage )
– Nvidia m /4000 ( this would like to replace with RTX 2080 ti) or is it good to use anyways !!!!
– .800w SMPS corsair
Hey Kalesha,
Thanks for asking!
Upgrading your GPU to an RTX 2080 Ti will definitely bring an increase in performance in terms of GPU rendering with Redshift but if you want to get the best performance and have the means for it, my suggestion is for you to start from scratch and come up with a new build.
If I may ask, what’s your budget like?
Also, the site’s PC Builder Tool at https://www.cgdirector.com/pc-builder/ will give you build recommendations based on your budget and use case scenario so you may want to check that out.
Cheers,
Alex
Hi Alex, Thank you for your reply. If i have to build from scratch, My budget would be around $4000. which includes 2 x 2080 ti card.
Hi Kalesha,
If you have a budget of around $4,000, I suggest that you go with a build like what I put together below:
———————
Your Custom GPU Rendering Build: (https://www.cgdirector.com/pc-builder/?=rm2gfd00edd)
Parts List:
CPU: Intel i9 9900k 3.6GHz 8-Core Processor ($474.99)
CPU Cooler: Noctua NH-D15 1151 ($89.95)
Motherboard: ASUS Prime Z390-A ATX 1151 ($179.99)
GPU: NVIDIA RTX 2080TI 11GB – Asus Turbo ($1149.99)
GPU #2: NVIDIA RTX 2080TI 11GB – Asus Turbo ($1149.99)
Memory: 64GB (4 x 16GB) Corsair Vengeance LPX DDR4-3200 CL16 ($319.99)
Storage SSD: Samsung 860 EVO 1TB 2.5″ Solid State Drive ($139.99)
Power Supply: be quiet! Straight Power 11 1000W ATX 2.4 Power Supply ($199.90)
Case: Corsair Obsidian 450D ATX Mid Tower Case ($204.99)
Total: $3909.78
Generated by CG Director PC Builder on 30/9/2019 @ 21:40:6
———————
The total of the build comes up to around $3909.78 and which leaves you around $90 and some change which you can still use for other parts of your build like extra storage.
Cheers,
Alex
Alex Thank you for giving me good PC Built . here few points like to ask you. By Tweaking my budgets a little bit. what if i make few changes ? if i do that. how much difference will it be in my Houdini and Red shift workflow performance.
1) CPU : AMD Ryzen 9 3900X ( instead of – Intel i9 9900k 3.6GHz )
2) Motherboard: MSI Prestige X570 CREATION _ E-ATX ( instead of ASUS Prime Z390-A ATX 1151 )
3) Memory: ( i wonder if i can get 32Gb x 4 Kit 3200 mhz) 128GB ram ( instead of – 64GB )
4) SSD : Samsung 970 Evo Plus NVMe PCIe M.2 1GB ( instead of – Samsung 860 EVO 1TB )
Please help me ,if its good this way or is there any better advise .
Thank you.
Hi Kalesha,
Changing your CPU to the Ryzen 9 3900X will not bring significant increase in your Redshift and Houdini workflow performance. In fact, the i9-9900K might even be around 2% snappier than the 3900X but the Intel platform is not that future proof with its limited upgrade path.
The same goes with choosing the MSI Prestige X570 CREATION _ E-ATX over the ASUS Prime Z390-A ATX 1151 for that matter. Both motherboards support up to 128GB of RAM but the challenge is getting 32GB x 4 RAM kits as these are hard to come by, not to mention only a couple come with speeds of 3200MHz or higher. If you go for the MSI Prestige X570 CREATION _ E-ATX motherboard, the fastest 32GB RAM stick in its QVL list is the VENGEANCE® LPX 64GB (2 x 32GB) DDR4 DRAM 3000MHz C16 with model CMK64GX4M2D3000C16 ver3.40.
As for the SSD, I suggest going for the Samsung 970 Evo Plus NVMe PCIe M.2 1GB as an NVMe SSD is a little faster than the Samsung 860 EVO 1TB. Be advised though that the increase in performance is not much of a game-changer, so to speak.
Cheers,
Alex
Hi Alex, Thank you very much for being so helpful..
One last thing i want to ask.
1) As i am planing to place GPU: NVIDIA RTX 2080TI 11GB – Asus Turbo x 2 of them. in one machine. So power supply of 1000W, will it be goo enough ??
2) I have observed there are different brands of RTX 2080 ti GPU’s. But you have suggested RTX 2080ti 11Gb Asus Turbo. ( which i can blindly trust you, it must be good one, and indeed i am going ahead with your suggestion ). But for my curiosity, if you can say in short and sweet., How to make best choice from various GPU’s 2080 ti brand as per our needs!!!
Thank you once again for guiding me trough helpful knowledge. 🙂
Hey Kalesha,
1000W is enough for 2x 2080Ti! 🙂
The reason I recommend Asus Turbo Series GPUs is because they are Blower-Style GPUs that have special Coolers that perform best when stacking many GPUs on top of each other. As 3D Artists often need lots of Render performance, this means many buy more than just one GPU. And if you use multiple GPUs you need Blower-Style GPUs so the Air is transported outside of your case and does not stay inside of the case heating it up. Blower style GPUs handle this well. Asus is just a personal preference as I have used many of those and have had no issues so far. MSI Aero though is another brand with blower-style GPUs that you can look in to.
Cheers,
Alex
Hello alex, I wonder where the quadro rtx4000 ll be in this benchmark?
I m building a tower and I need your help, should I.give the extra 350€ to go with the quandro or I should I pick the rtx2070?? Also I dont care about gaming, I care only for pure performance on c4d, after effect etc
Hey Fotis,
The 2070 is slightly faster than the Quadro RTX 4000 at least in octanebench. As both octanebench and redshift scale very similarly, I’d say you would probably get the better deal when going with the 2070. Especially because it is a lot cheaper but performs on par or even better than the Quadro.
Cheers,
Alex
Hey Alex, with 4 x 2080Ti Hybrid GPU’s (standard clock) with can pull 2m, 21 but this is using x16,x8,x16,x8 PCIe configuration.
We’ll be submitting results with 4 water-cooled, all x16 slots with the Xeon W-32XX series of CPU’s shortly.
This is going to hopefully get us under 2m on the benchmark for Redshift! For all interested, we have found that 4 way GPU is the optimal for Redshift, we build 7 and 11 way rigs and they just don’t scale up correctly unlike in Octane.
Hey Richard, that’s awesome! Let me know once you have the results 🙂
Cheers,
Alex
Hey Alex. This benchmark has been super helpful in both work and planning out a personal CPU.
My build will eventually have two 2080Ti’s and a 1950x threadripper, I intend to get into heavier rendering and I am wondering if the Noctua NH U14s tr4/sp3 fan is going to be enough to keep it cool? If not what would my options be? Id prefer not to use water cooling unless its the only way.
Hey Jake,
Yes that Cooler will be sufficient. I have it running on a 2950X and its doing great! An alternatives would be a beQuiet Dark Rock Pro 4 TR4 which is a bit bulkier though.
Cheers,
Alex
Wow that reply was way faster than I expected! I appreciate it, feeling a lot better about the build now.
I’m also curious as to what motherboard you prefer to run threadrippers on? I intended to use the ASRock x399 because it had less bad reviews. seems like the TR4 boards can come with some issues.
Hey Jake,
Many of the reported issues are still from a time when the boards where very new, most of them have been fixed with bios updates.
I usually build Threadripper GPU Builds so 4x GPUs. The best board for this is the Gigabyte Designare Ex, as it has 5 pcie-slots for 4 double-height GPUs, but ASROCK has some good boards too! Are you looking into the Taichi?
Cheers,
Alex
Oh I see. yes i was looking at the Taichi. though other than those bios issues I really don’t know the difference between the TR4 boards.
Do you think the Gigabyte Designare Ex is better than the ASRock Taichi even if I only plan to use 2 GPUs?
The thing is there really aren’t many differences between the boards, Mostly just amount of usb, wireless lan, # of pcie-slots, bios flashback option, fan headers, LAN capability. That’s pretty much it.The taichi is a bit cheaper than the designare ex, so I’d go with that one for your needs!
Cheers,
Alex
Ill do that then. thanks a bunch for the advice, feeling much more excited for my new build now!