V-Ray Benchmark 1.0 release

[vlado]

ok so the person at number 1 with their 88 cores is making my computer feel inadequate..stop it

It's one of the machines that we got for testing multithreading in V-Ray

Best regards,
Vlado

[vlado]

Thanks Vlado, it could well be this workstation I'm using at the moment. I've had all kinds of issues with it but if I can help by providing logs or something then let me know.

If you go to the temp folder for the current user (type %temp% in Windows explorer), there is a sub-folder "V-Ray Benchmark" that contains the text log outputs from the benchmark run (vraybench.err and vraybench.out). Do they say anything meaningful?

Best regards,
Vlado

[varosi]

1. Are you behind a proxy server?
2. Check your firewall settings
3. On Submitting, V-Ray Benchmark should open you a default browser so you could log-in and then submit and compare.

Unfortunately I wasn't able to submit my result, it just hangs on "submitting" but what I found strange was that my GPU time was 01:38 with this workstations 780 Ti, which is a bit confusing when a 1080 is listed at 01:50.

[varosi]

We're thinking of it, so if there is enough interest we will build one.

Very nice, thanks!
I hate to throw requests like this, but any chance we'll see something headless for Linux?

[savage309]

is there a way to use the same render image for both cpu and gpu? Good for deciding if cpu or gpu is the route to go for when purchasing machines etc

We don't plan to do that (at least not in the near future).
The reason for that is that it is very hard to do a proper fair test. While V-Ray can raytrace render scenes in both CPU and GPU, and in general, GPUs tend to run faster, the differences in render times can vary significantly between different types of scenes. For this reason, we decided to keep each test to it’s own scene. Users should therefore compare CPU to CPU and GPU to GPU.

Best,
Blago.

[Fredrik Averpil]

We're thinking of it, so if there is enough interest we will build one.

We're currently using a home-made scene to test performance in GCE and on our local farm. This is nice because then we throw a lot of technologies at it at the same time which also probes e.g. a nightly/stable build for issues. I'm sure your current benchmark does not try to implement all types of lights, proxies, fur, motion blur and is suitable for this type of thing. But in case you would go down that route, I think it would be super useful (perhaps even internally for you).

Using the same scene, although for a benchmarking purpose, we also measure the time it takes to render.

So you could say we try to nail two flies in one clap (or however that translates into English...) by using one scene to both proof a new version and measure a benchmark. And so this becomes very useful to also be able to run on Linux.

[vlado]

We have more than a thousand unit test scenes for V-Ray for Maya and for 3ds Max separately that the nightly builds and the releases go through. There are additional several hundred scenes for V-Ray Standalone and V-Ray GPU. They test various features and in addition to tracking potential differences in the rendered images, we also track changes to the render times and RAM usage.

Best regards,
Vlado

[Fredrik Averpil]

Wow, that's awesome. I actually had no idea you did that. As usual, hats off. Wonderful!

[LudvikKoutny]

I really like the benchmark. It has good design and nice compact feeling to it. Although, I think it would be bit more helpful if both CPU and GPU rendered the same scene. Both scenes look simple enough that they should achieve almost perfect parity between CPU and GPU, and people could then compare speed of their CPU against their GPU

[vlado]

I really like the benchmark. It has good design and nice compact feeling to it. Although, I think it would be bit more helpful if both CPU and GPU rendered the same scene. Both scenes look simple enough that they should achieve almost perfect parity between CPU and GPU, and people could then compare speed of their CPU against their GPU

We considered this, but for the moment the speed of CPU vs GPU is very scene-dependent and we didn't want to mislead people. The benchmark might render X times faster with the GPU, but that doesn't mean that all scenes will render like that.

Best regards,
Vlado

[LudvikKoutny]

Yep, I guess that's a fair point. From my own experience, GPU loses edge over CPU as the general scene complexity increases.

[vfx_film_televsion]

+1 for a linux build. Since all our workstations run on centos this would be sweet

[Jiri.Matys]

Hi,

I record two videos:

i7:
https://drive.google.com/file/d/0Byx...ew?usp=sharing

2x xeon:
https://drive.google.com/file/d/0Byx...ew?usp=sharing

Look at 2x xeons: is it really ok? Time is too different..

And as my friend said, scenes looks like 10 years old. I dont think it should render one hour. But it should look like from 21 century Scene for CPU has only 4 objects or so, dont think it represents powerful and modern renderer as Vray should be, but its an opinion. We can compare with this scenes too of course, just saying how it looks compared to what vray can render. But real problem is seconds counting, I think we can put all results in a trash.

[cmjohncheng]

Hi,

I record two videos:

i7:
https://drive.google.com/file/d/0Byx...ew?usp=sharing

2x xeon:
https://drive.google.com/file/d/0Byx...ew?usp=sharing

Look at 2x xeons: is it really ok? Time is too different..

And as my friend said, scenes looks like 10 years old. I dont think it should render one hour. But it should look like from 21 century Scene for CPU has only 4 objects or so, dont think it represents powerful and modern renderer as Vray should be, but its an opinion. We can compare with this scenes too of course, just saying how it looks compared to what vray can render. But real problem is seconds counting, I think we can put all results in a trash.

tried to download, but it show"Sorry, the file you have requested does not exist."

[Jiri.Matys]

hmm, weird. Try that one for xeons:

https://mab.to/gzhO9j2FH