Benchmarks  ⋮  Hardware

Single vs Dual vs Quad Channel Memory on Ryzen Threadripper (Benchmarks)

CG Director Author Alex Glawion  by Alex Glawion   ⋮   ⋮   19 comments
CGDirector is Reader-supported. When you buy through our links, we may earn an affiliate commission.
Single vs Dual vs Quad Channel Memory on Ryzen Threadripper (Benchmarks)

Does Single vs Dual vs Quad Channel Memory impact performance on Ryzen Threadripper CPUs? Ever since AMD has introduced the glorious new Ryzen Family of CPUs, there has been lots of talk about Memory Compatibility with said Processors.

Intel CPUs have been able to use pretty much any kind of RAM but with Ryzen and Threadripper CPUs, XMP, Memory Clocks, Latency, Infinity Fabric, Single Channel vs. Dual vs. Quad Channel and other buzzwords have turned out to be quite confusing, especially when choosing a new System for your 3D Rendering, Video Editing or other kinds of Graphics Design needs.

In this Article Series that will analyze Threadripper CPUs to see how they can be optimized as much as possible, let’s start with how memory channels affect Ryzen Threadripper.

This is what Wikipedia has to say about Memory Channels:

In the fields of digital electronics and computer hardware, multi-channel memory architecture is a technology that increases the data transfer rate between the DRAM memory and the memory controller by adding more channels of communication between them. Theoretically this multiplies the data rate by exactly the number of channels present.

Having Quad Channel Memory, as is the case when you have at least 4 Memory Modules in a Threadripper System, should then quadruple your data rate between your RAM and the memory controller. Sounds pretty good!

Let’s see how Single vs Dual vs Quad Channel Memory impacts some of our most popular benchmarks:

System Specs

  • CPU: AMD Threadripper 1900X @ Stock Speed
  • Memory: Corsair Vengeance LPX DIMM Kit 64GB, DDR4-2666, CL16-18-18-35
  • Mainboard: MSI Gaming Pro Carbon AC X399
  • GPUs: 4x 1080Ti Asus Turbo
  • SSD: 860EVO 500GB
  • OS: Win10

Single vs Dual vs Quad Channel Memory – Benchmark Results

Cinebench R15

Single vs Dual vs Quad Channel Memory - Benchmark Cinebench R15
In Cinebench R15 there is a slight increase in performance from Single Channel to Dual Channel to Quad Channel Memory. It is only about 1%, but it is consistently reproducible.

These Benchmarks have been run 5 times with the average of those runs displayed here.

Cinebench R20

Single vs Dual vs Quad Channel Memory - Benchmark Cinebench R20

Cinebench R20 does not seem to benefit from Quad vs Single Channel RAM. The problem with this benchmark is that the results were all over the place, making it difficult to pinpoint a strong average midpoint. The Benchmark results could fluctuate +/- 100 points within the same runs.

VRAY CPU Benchmark

Single vs Dual vs Quad Channel Memory - Benchmark Cinebench VRAY

As the VRAY CPU Benchmark only gives us increments of 1 second, it is difficult to see if there really is any difference between Single, dual and quad channel memory. These Benchmark runs though seemed to be very consistent across many runs, never deviating from around 1:17s / 1:18s all that much.

Redshift & Octane

Single vs Dual vs Quad Channel Memory - Benchmark Cinebench Redshift

Single vs Dual vs Quad Channel Memory - Benchmark Cinebench Octane

Although Redshift and Octane are GPU Rendering Benchmarks, they still make good use of CPU resources and might be impacted by memory channel performance.

After running the benchmarks several times though, it seems that the differences in benchmarking times lie within the margin of error.

Conclusion

The only Benchmark that is consistently reproducible and benefits from having more memory channels, is the Cinebench R15 Benchmark.

Having Quad vs Single you gain about 1,2% in performance. All of the Benchmarks don’t seem to be very Memory demanding, due to the fact that the Benchmarking Scenes are very simple and there is no reason to access large amounts of Memory while Rendering.

It might very well be that you experience slightly higher performance boosts on Quad Channel Memory and the Threadripper Platform, depending on the tasks and characteristics of your Scenes and Projects.

 

What is your experience in Single vs Dual vs Quad Channel Memory? What Memory are you thinking of buying? Comment below!

Join the New CGDirector Forum! Expert Advice & PC-Build Planning with a warm and friendly Community! :)

Alex Glawion - post author

Hi, I’m Alex, a Freelance 3D Generalist, Motion Designer and Compositor.

I’ve built a multitude of Computers, Workstations and Renderfarms and love to optimize them as much as possible.

Feel free to comment and ask for suggestions on your PC-Build or 3D-related Problem, I’ll do my best to help out!

19
Comments

Carlo

I agree with one of the comments below, To make it fair Test you should test it with capped max size, Since you use those builds for Rendering i would advies to make a valid test 1x32gb single channel, 2x16gb dual and 4x8gb run os same Frequency. Already while booting windows you will note the difference.

Aston

Hey, can you check this and help me out to finish pc building
CPU: Intel I7 9700K Coffee Lake
GPU: Nvidia Gigabyte GTX 1660 Ti 6GB OC
RAM: Corsair 32GB 2x16GB DDR4/3200 Vengeance LPX Black, CMK32GX4M2B3200C16
SSD: Samsung SATA III MZ-76E500B 860 EVO 500GB
Motherboard: Asrock Z390 Pro4
I am not sure about motherboard is that good enough, I am gonna use this pc for sculpting, texturing and game development. Also if you can recommend me some power suplay(under 100$ will be nice)

Aston

PS.
Or maybe is better to go with Ryzen 7 3700x and some other motherboard?

The other guy

When I look at these benchmarks, I feel like there are too many variables to make the comparison legit. Your test is comparing 1, 2, & 4 channel RAM; however, you are also increasing the total amount of RAM with each test. This may partially explain why the results are so close. If these applications were to get max benefit at 16GB, then increasing the amount to 32GB and 64GB isn’t going to see much improvement.

Instead, you should cap your test at 16GB. The single channel test should be w/ one 16GB stick. While the dual and quad channel tests use 2 x 8GB and 4 x 4GB respectively. This will allow your benchmarks to focus on the use of multiple channels as intended. Obviously, the RAM will need to be the same speed and timings.

Alex Glawion
Alex GlawionCGDIRECTOR

Hey and thanks for the feedback! I agree the size increase is suboptimal. I’ll include some ram usage screenshots, as those benchmarks need much less than 16gb of ram.

Wouldn’t you say, that because the benchmarks need less than 16gb of ram to perform optimally, the only thing relevant here is the ability of the higher channel bandwidth?

These benchmarks are not very ram heavy and therefore don’t show large differences in performance. I’ll do some benchmarks with 3d scenes that really do a lot of ram reading and writing, which should give us a better picture.

Thanks for your comment!
Alex

Robin

Someone told me that with Ryzen Dual Channel is giving the most performance and that their CPUs greatly benefit from more ram.
So if that were to be true, going from dual to quad should decrease performance, but doubling the RAM Size should benefit performance.
So I hoped that this test would give some insight in this. But those two forces/factors might have counteracted each other…
But I don’t think going batsht with team capacity brings any performance increase. Especially for the money you have to pay more.

dude

this is flawed.
For months I couldn’t understand why my sun FPS are “limited” to 120-190fps at overwatch when the processor is i5-9400f and graphic is gtx-1070.
yesterday I found I mistakenly connected A0/A1 memory and not A0/B0 etc.
Immediately the FPS went through the roof. to 300FPS. Minecraft got a similar boost – about two times FPS.
what’s more – all the weird hangs and bouncing FPS rates disappeared.

Robin

Dude, that’s what Dual Channel stands for. You can’t just randomly slap ram into your Mainboard.
I mean it’s even colored… xD

Julio Camarena

Wow, I just ordered an AM4 pc with the 3900x and just realized about the dual channel limitation on it vs the quad channel config on TR4 platform. I think I could just cancel my order and change to a 1950x with quad channel (which turns about for the same price and performance for rendering) but reading this makes me think I’m ok with dual channel, right?. I primarily work with 3ds max and corona/vray.

Alex Glawion
Alex GlawionCGDIRECTOR

Hey Julio,
Yes absolutely, Dual vs Quad in Rendering and 3D-Software makes almost no difference. Much better to go with the higher-clocking cpu (ipc) as is the case with the 3900X.

Cheers,
Alex

Julio Camarena

Yeah I think so too Alex. It could be a limitation going for 128 gb of ram because 32 gb modules are slow atm (2.6666 mhz) but I’m currently on 48 gb and doing not so bad so 64 is going to give me some time so that faster 32gb modules come out… if they ever do

Nuno

I suppose the improvement in single-core performance going from dual channel to quad-channel is also negligible in Xeon platform?

Alex Glawion
Alex GlawionCGDIRECTOR

Hey Nuno,
Yes that is the case at least in the Benchmarks and tasks listed here. There might be some datacenter tasks that make much better use of increased channel bandwidth but 3d rendering doesn’t seem to be one of them.

Cheers,
Alex

Nuno

Thank you Alex! 😉

Lukasz Szeflinski

What about the simulation in 3dsmax? Like particles or cloth? Still dual vs quad doesn’t matter?

Alex Glawion
Alex GlawionCGDIRECTOR

Hey Lukasz,
I don’t have any concrete numbers for those specific tasks and apps, but given the fact that particle sim or cloth sim isn’t too bandwidth demanding, Id’ say you see just small improvements in the low single digit area.

Cheers,
Alex