Reuse older hardware for distributed GPU rendering - does a P5000 speed up the rendering?

[3d-laboratory]

I try to use my "old" Quadro P5000 in my Render node for distributed rendering with GPU. In my workstation I have the RTX 4090.
I thought I can get some additional speed but it seems, that the administrative work is a brake for the whole system. The RTX 4090 renders the scene in 3:32, the P5000 in the rendernode needs 22:36. But both together needs 5:27, longer than the 4090 itself. Is this a normal behaviour? 5:27 is for bucket rendering, with progressive rendering I had 3:30 again, so the P5000 didnt really help. Any suggestions to speed it up (beside buying another 4090)?

[Nejc Kilar]

Honestly to me it really doesn't sound like it is worth it. The P5000, with all due respect, is a really old GPU.

Based on the two entries on the V-Ray benchmark page your 5950x, if you do Hybrid rendering, should be roughly a bit more than 2x the speed of the P5000. So maybe, if you have to, that route is worth exploring?

If it helps, I've recently retired my 2x 2696v3 Xeon render node for much the same reasons. Sure, it did help speeding up the rendering but compared to the more modern 96 core Epyc nodes I have it really wasn't efficient at all. And that's another thing to consider - In my case the Xeon pulls 300W whereas the Epyc 400W. The Epyc is pretty much ~5x faster though. Hurts the wallet and the planet I suppose

As for bucket rendering being slower, I would think the P5000 was stuck doing the last couple of buckets whereas the 4090 is just chilling and waiting for the P5000 to complete the rendering so you can move to the frame.

[3d-laboratory]

To be honest: I didn´t expect much but I wanted to get both GPUs to render with distributet rendering, more as a technical gimmick. It is of course sad, I bought it used about 5 years ago for my workstation for about $1000. Now I wanted to sell it but best bit was $200. This is the course of things. The CPU in the rendernode helps more. It has about 80% of the render speed of the workstations CPU. But if I render with DR on both CPUs your EPYC is surely 4 times faster, congrats. Your last sentence is sometimes also true for CPU DR rendering.

[Nejc Kilar]

To be honest: I didn´t expect much but I wanted to get both GPUs to render with distributet rendering, more as a technical gimmick. It is of course sad, I bought it used about 5 years ago for my workstation for about $1000. Now I wanted to sell it but best bit was $200. This is the course of things. The CPU in the rendernode helps more. It has about 80% of the render speed of the workstations CPU. But if I render with DR on both CPUs your EPYC is surely 4 times faster, congrats. Your last sentence is sometimes also true for CPU DR rendering.

Yeah, tech moving forward can both be a great and no so great thing at the same time, agreed I very much felt it when I retired my dual Xeon.

The Epyc I have is pretty awesome indeedio but at the same time we can acknowledge there's Zen 5 Epycs with 128 cores now so there's even faster solutions. Not to mention 2 socket systems as well where you get 2 CPUs in one box.

As for DR, yeah it does still happen from time to time for me as well but rather rarely I suppose.