i7 5960x or dual Xeon E5-2687w

[shermanjas]

I am going to buy a new rendering system. I would like to go with four Titan X's or four 980ti's. Does anyone know if the i7 5960x chip can push enough data fast enough to the four large cards. Would I get better render times with four of these cards and dual Xeon E5- 2687w's pushing them data?

Please help!

Sincerely,
Jason

[smicha]

Unless you render on cpu engines xeon is a waste of money. 2687W x2 will cost you 4k eur nett, but 5930k for 500eur is enough for a great 4x titan X machine. If you want to save 1.2k eur you may go with 980ti but you loose 6gb of vram and little performance.

Recently I built this machine:
http://render.otoy.com/forum/viewtopic.php?f=11&t=47678
if you have any more questions just feel free to ask.

[garytyler]

I'm assuming you are rendering animation...

Depending on your render settings, Octane may or may not need to prep scene data to the GPU between each frame. If it does, the faster CPU will help some, but no more than a few seconds per frame. Once the GPU is processing, the CPU is no longer under strain. The CPU only preps the scene for the GPU. You can run as many GPUs as you want on a small CPU. It will just take longer to prep.

[shermanjas]

Thanks for the input. Yes, I am doing animation work and need to speed up my output.

I was looking at machines from mediaworkstations.net and got this as for an answer from Chris.

"Yes, there's a limit to how many bytes an i7 5960X can push, and that's what GPUs need to perform. I just had a conference call with my new media lead Evan about a benchmark comparison looking at CPU and GPU performance and limits with a new 28 core i-X2 with 128GB RAM and 4 GTX TitanX purchased by a Japanese architecture firm, vs. an i7-X with 2 GTX TitanX and 64GB RAM the week of June 8. I will also use the oppty to test the feedback from other professionals about processing limits of the i7-5960X. The performance roughly doubles going from 1 to 2 cards in most scenarios, but does it triple if you add a third card? Does it increase at all? Most ISVs are updating code left and right now to take advantage of GPU acceleration, but the simple fact remains, that I have seen benchmarks in Premiere Pro and Maya where the third card adds 5-10% GPU performance increase, not the 50% you might expect with absolute scaling. It's not a RAM issue, though that can come into play - it's that the CPU cannot push any more data. But dual Xeons? They can and do. "

does any of this make sense for Octane? If I am going with a four GPU setup wouldn't I want the GPU's to be able to all perform at there highest level for all application like After Effects and Premiere Pro Does this mean I go with a dual Xeon setup?

I think I may be getting more confused!

Thanks for all the help

Sincerely,
Jason

[glimpse]

Hi Jason.

Regarding to scaling using multiple GPU with Octane Render the effect is linear, so if You have 3cards they will produce You result in thrid of a time You would get using single card, while 8 GPUs will get the calculations done in 12.5% time (compared to single GPU).

Only thing You need to keep in mind is how You plug thsoe GPUs. If You're working with heavy scenes it's better to have more lanes (do not bottleneck systems connecting GPUs into too slow PCIe slots).

As for CPU, as Guys mentioned before You need CPU only to prepare & deliver all the data to GPU - the computation itself is working inside GPU (where processing unit toalks with vRAM).

In the end..remember that once You put multiple cards inside Your case they heat up each other under full load & thermal throttling might come into play - with this You'll start loosing render speed.

In order to avoid this You need propper cooling that would not even keep Your cards at lower temperatures (keeping stability & extending longenity) but would allow You to overclock further gettting extra performance.

for strictly OctaneRender centric "workstation"/render node quad core CPU is enough, 6-8 core is better, but..if You're looking to use Your rig for other jobs..plan acordingly based on Your needs for that software.

Xeons will always carry less value, unless You need specific features they could provide or density of power You could not get otherwise (stability is another thing Guys choose to spend more for those).

[shermanjas]

Thanks for the info. It all makes sense but I am still getting conflicting information Boxx did some testing for me and this is what I heard back.

"Hi Jason,
My apologies for taking long to respond. First, we appreciate the feedback because yes, your information was essentially right on the money. We had one of our engineers perform a quick test and while the i7 did show some gains with 4 cards over 2 cards, it wasn’t taking the full advantage of all 4 cards. The Dual 10 core (20 core machine) which got peak performance from all 4 cards.

So – you will want to stay with our 7901 model which is dual Xeon, or the 8901 model (only if you think you might add a 5th card).

What are you basing your information on. Have you seen any benchmarks comparing more than two GPU's on both the Dual Xeon and the i7 CPU's?

Sincerely,
Jason