What is the best CPU for Maya if I am GPU rendering?

What is the best CPU for Maya if I am GPU rendering?

Anonymous
Not applicable
5,510 Views
9 Replies
Message 1 of 10

What is the best CPU for Maya if I am GPU rendering?

Anonymous
Not applicable

I am building a workstation for Maya. I am going to use Maya for modelling and animation, and Redshift for rendering. I will not be using Maya for simulations.

 

From what I have read online, GPU rendering with V-Ray or Redshift is faster than CPU rendering, so I am going with a 2080 ti as my GPU and avoid CPU rendering entirely. Here are benchmarks showing rendering speeds for GPUs:

https://www.gamersnexus.net/guides/3399-best-workstation-gpus-2018-for-adobe-premiere-autocad-vray-a...

 

But which CPU is best? Having a CPU with a lot of cores and lower clock speeds (Threadripper, Skylake) seems like a waste now. Since I am not CPU rendering, what other Maya functions requires consideration between high core counts and core clock speed? The only example I can only think of is view port performance. Is there anything else to consider?

 

The fastest single core CPU right now is the i9-9900K. But I cannot find any benchmarks comparing the i9-9900K to any other processors in Maya specviewperf benchmarks. The excellent Gamers Nexus article linked above uses an i9-7980EX as its test CPU with specviewperf benchmarks, but I have no idea how the i9-7980EX compares to the i9-9900K in terms of specviewperf because it was a GPU only comparison.

 

So this leaves me with some questions:

  1. Which CPU is best for Maya with GPU-only rendering? i9-9900K or i9-7980EX? 

  2. Is there some other activity or function in Maya besides view port experience that might benefit from higher core counts over clock speed? Perhaps simulations, but I am not doing that.

  3. Which processor has the better view port experience? i9-9900K or i9-7980EX? 

The CPU choice matters because their platforms have different amounts of PCIe lanes. If I go Skylake, I can add another GPU for rendering to the workstation without sacrificing PCI speeds. For example, an i9-9900K would limit me to one GPU with x16 PCI; two GPUs would put me at x8 PCI speeds for each GPU. So I would like to know what I am truly gaining in viewport performance (or anything else) with an i9-9900K when I give up having a second GPU for rendering that I could have both at x16 speeds if I go with a Skylake CPU.

 

TL:DR: What is the best CPU for Maya if I am GPU rendering?

 

Sorry for the long post and thank you for your help.

0 Likes
5,511 Views
9 Replies
Replies (9)
Message 2 of 10

Anonymous
Not applicable

You kind of answered your own question. If you plan on doing GPU rendering and not use the CPU for rendering, then your best bet is to get a processor with higher clock speeds.

 

Plus you are comparing the i9-9900k to the i9-7980xe with a price of ~$400 to ~$1300 (used) that's a +300% difference!

 

Why would you pay 3x more for a processor that runs slower at single clock speeds when you don't plan to use all the cores?

 

I will say that I put together a threadripper 1950x build beginning of 2018 myself. It runs 16 cores @ 3.9GHz stable overclock. It makes working with Arnold renderer in Maya very smooth, and originally I got it because I was doing physical rendering in C4D, so at the time it made sense for me and I often see all cores being utilized for those tasks. Makes a big difference!

 

I've also been using Redshift and have a gtx 1070 for that. It runs well, and I recently just got a used 1080ti on ebay for $520 which I'm excited to add to my rig to speed up that process.

 

So in the end, even though I have a high core count cpu myself, if you specifically don't plan on using so many cores, why pay the premium? With the money you save you can easily get a second GPU.

0 Likes
Message 3 of 10

Anonymous
Not applicable

Also I should add that you might not have to worry about the difference between x16 vs x8 pcie lanes for your GPU's, I'm pretty sure I've seen tests online where they haven't found that to be an issue in rendering performance. I think the pcie lanes are for the initial transfer of data to the GPU at the beginning of rendering, and generally even at x8 lanes it's so fast you're not going to notice any difference at x16. But I would do the research because I'm not 100% positive.

0 Likes
Message 4 of 10

Anonymous
Not applicable

Hi jhsu2R56R,

That makes sense. I will go with the i9-9900K. It's interesting that we are in a situation where consumer desktop CPUs and GPUs can perform better than high end desk top CPUs and professional cards, at least for Maya-like software.

 

Good luck in 3D!

0 Likes
Message 5 of 10

Anonymous
Not applicable

No problem. I definitely hear that, I grew up in a time when the only way you could get into CG was if you had access to +$10K machines. It's amazing what we have access to now that technology is so much more affordable.

 

Good luck with your rig 🙂

 

0 Likes
Message 6 of 10

vedran.klemen
Participant
Participant

I think sim times are depending on cpu. So more cores on cpu, the faster the sim time. Iff this is true, you will be better with 1950x or 2990wx (cheaper vs faster). Single core speed on this cpus are good enough for gpus. And you will have powerfull cpu for After Effects.

0 Likes
Message 7 of 10

Anonymous
Not applicable

Hi Vedran,

Right on, that is a good point. If I was doing simulations or After Effects I would look hard at cores over clock speed. 

 

I think what is somewhat frustrating is the lack of benchmarks in general for professional software such as Maya. Gaming benchmarks are everywhere, but pro benchmarks are harder to come by, and when they are available they tend to just cover just a few things and lack the details we are looking for.

0 Likes
Message 8 of 10

vedran.klemen
Participant
Participant
Well there isn't much to it (ok, fume fx is gpu viewport and rendering
driven). Firepros in Maya are much faster in viewport than regular gpus or
even Quadros (my W7100 with 4 cores cpu has 148 points in Cinebench and
2080 ti has 120 points. And in Maya viewport 2080 probably have the same
speed as any other card. But you dont need Firerpro iff you just started to
learn. Good cpu and gaming card for rendering should be enough.
0 Likes
Message 9 of 10

Anonymous
Not applicable

Cinebench is not useful for benchmarkin gpu's:

https://linustechtips.com/main/topic/910876-how-is-this-for-a-gpu-cinebench-score/

 

Also the main gpu renderer's on the market (Redshift & Octane) do not support AMD (OpenCL) gpu's, they require Nvidia (CUDA) gpu's. If the OP wants to do gpu rendering, unfortunately AMD cards are not a good choice.

 

Does Redshift support AMD GPUs?

https://www.redshift3d.com/support/faq

 

"We didn't give up on OpenCL, OpencCL gave up on us."

https://render.otoy.com/forum/viewtopic.php?f=9&t=61904

 

Currently the only gpu render supporting OpenCL is ProRender in Cinema 4D, and it's known to be very slow compared to Octane & Redshift.

0 Likes
Message 10 of 10

vedran.klemen
Participant
Participant

Well I do not recommend Firepro for rendering except for Vray which accepts Open cl. As you will soon discover Maya doesn't only have rendering tools but others also like animation and modeling, which I recommend to develop that skills also, and not just render the cup of tea and two speheres in the box, which sadly most do.

0 Likes