Will AMD FirePro Graphics cards work ok with VrayRT?

[savage309]

Since it is the strong side of OpenCL to adress CPU and GPU likewise it seems like a good idea to use that as good as possible. I tried the Benchmark 2.0 with GPU and Coherent tracing on. Result: 12 min 15.6 s which puts the w8100 in to the expected range of a Quadro M6000 according to the Benchmark Spreadsheet. The same scene on GPU+CPU, Coherent Tracing off, finishes at 7min 59s.

Since most people will get a Quadro or Firepro for their deep color support (like me) or some certified driver stuff, it is pretty interesting to see that a mid range FirePro card is performing pretty well against a Quadro six times as expensive.


Sounds exciting

Are you sure about those results ? Using coherent tracing reduced the render time for benchmark1 from 224s to 200s at mine Fury X. The same scene runs on M6000 for about 90s here. Keep in mind that 3.3 benchmark results can not be compared with 3.2 (image sampler is totally different and can change a lot of stuff in the render time).

[bibi5000]

Are we talking about the same scene here? I used this in my second test: https://dl.dropboxusercontent.com/u/...mark_scene.zip ("Benchmark 2.0")
Didn't change any settings, just clicked "render".

The results for this scene on the benchmark spread sheet for the M6000 are: 1 x M6000 3.20.02 / 346.35 (Linux) - 12m 26s

[savage309]

Are we talking about the same scene here? I used this in my second test: https://dl.dropboxusercontent.com/u/...mark_scene.zip ("Benchmark 2.0")
Didn't change any settings, just clicked "render".

The results for this scene on the benchmark spread sheet for the M6000 are: 1 x M6000 3.20.02 / 346.35 (Linux) - 12m 26s

Yep, thats correct but you should use the same versions of the software to compare the hardware, since the software may (and if fact does) behave very differently between different (major in this case) versions.
I think that with 3.30.04 M6000 will give (much) smaller time than those 12 minutes. I will see if I can find one to test and update here with the results.

[bibi5000]

use the same versions of the software to compare the hardware

haha

I think that with 3.30.04 M6000 will give (much) smaller time than those 12 minutes

ok, I'm curious what the M6000 can do. Altough I'd rather buy a car for that amount of cash

[PGD]

I've installed 3.30.04 and run the benchmark scene and it is much better. The render did continue as opposed to stopping at 500/512 so I had to stop the manually.

Screen shot attached for the render info but 2m 58s for 543 paths

(why does the yellow render info show a different time than the frame stamp?)


This is with both AMD cards installed checked to render:
AMD FirePro W9100 - 16GB GDDR5 Workstation Graphics Card (Hawaii)
AMD FirePro W5100 - 4GB GDDR5 Workstation Graphics Card (Bonaire)

[PGD]

Further test screen shots.

2 AMD cards and CPU - 1m 43
AMD W9100 - 2m 25
2 AMD cards retested - 2m 14 (exactly as per the post above @2.58 but with different results?)


2 cards 01 already uploaded

[savage309]

I think the difference in the results is only in the noise. In the case of CPU and GPU working together, this is expected since the hardware in both is doing the calculations differently.

[PGD]

I meant a difference in render times given the same scene setup to be honest. As the first try was 2.58 yet the second was 2.14 (3ds was closed and opened again in between that)

I haven't looked closely at noise in these..

It's no biggie, just an observation

[savage309]

ok, I'm curious what the M6000 can do

The scene (benchmark 2) is completed for 2816 path per pixel made in 14976 passes, which took 13m 21s on a single M6000, V-Ray 3.30.04. Please make sure to compare vs 3.30 only.

[bibi5000]

Thank you for the benchmark savage309. Ich will try this again on the w8100 as soon as I find the time.
Just to clarify things: Should I leave Vray RT running after 2048 p/p are reached like you did?

[savage309]

Yes, let it run until it is done.

[Libertyman86]

Thank you for the benchmark savage309. Ich will try this again on the w8100 as soon as I find the time.
Just to clarify things: Should I leave Vray RT running after 2048 p/p are reached like you did?

I have the AMD Firepro W8100. Any updates on benchmarking on the Vray RT?

[bibi5000]

Here is the result: 2816 paths/pixel, 14336 passes in 50m 33.8s.

I expected the w8100 somewhere at 50% or 30% of the M6000, based on that this benchmark, but comparing benchmarks may be a bit pointless.

[savage309]

Here is the result: 2816 paths/pixel, 14336 passes in 50m 33.8s.

I expected the w8100 somewhere at 50% or 30% of the M6000, based on that this benchmark, but comparing benchmarks may be a bit pointless.

Thanks for sharing !
For our benchmark scene it seems to be around 26% of the M6000 - 13m vs 50m. Which is should be around the same percentage of the price as well ?
I would like to note again that RT GPU does not use any of the pro GPU features, so being a pro one doesn't give any benefit for this benchmark.

[Libertyman86]

Thanks for sharing !
For our benchmark scene it seems to be around 26% of the M6000 - 13m vs 50m. Which is should be around the same percentage of the price as well ?
I would like to note again that RT GPU does not use any of the pro GPU features, so being a pro one doesn't give any benefit for this benchmark.

interesting stuff. So, does Vray take advantage of 'pro' features in Nvidia cards? Its my understanding that most of the increase in performance in the Nvidia cards has to do with Vray's optimization towards CUDA tech rather than AMD's GCN/OpenCL tech, yes?

Any chance RT GPU will make some changes in the future to take advantage of more AMD pro GCN/OpenCL 2.0 features?

should I ditch AMD altogether and get a couple titan cards instead? My other other option is just get another W8100 and increase my VRAM to 16GB and hopefully increase my performance that way.