Benchmark Vray 3.3 vs Vray 3.4

[Nearl]

We finally got around to upgrading to 3.4 and I noticed a slowdown on one of my scenes. So, I had one of my animators run a benchmark test.

After rendering the benchmark scene with both Vray 3.3 and Vray 3.4, we got these results:

Vray3.3 took 3 min 50 sec
Vray3.4 took 4 min 6 seconds.

3.3 rendered it 15 seconds faster, so 3.3 is indeed faster than 3.4

Has anyone else experience a slowdown?

[vlado]

Has anyone else experience a slowdown?

Nope, normally there is no slow-down; I've double-checked many times. Of course, it's always possible that there is a bug somewhere, so it would be best if you can get us that scene to support@chaosgroup.com so that we can check it out.

Best regards,
Vlado

[Nearl]

Ok, I'll have the animator get me the file so I can post it. Hopefully it's just a bug. Thanks Vlado!

[Nearl]

Here's the benchmark scene. I also, sent it to support@chaosgroup.com

Benchmark.zip

[Nearl]

Second benchmark...

The below image was rendered in 3.4 and then in 3.3. Nothing was done to alter the scene between versions. The only thing I can think of that might cause this drastic of a slowdown, is that these models have UDIMs with 8k maps (tiledEXR) and V-RayDispMod. Other models in the same scene (without UDIM and Disp) render fine. Could UDIM and/or VRayDisp have a bug in 3.4?

V-Ray 3.30.05 - 0h,5m,24.8s
V-Ray 3.40.02 - 3h,0m,16.5s

[vlado]

The only thing that I can think of is that in V-Ray 3.40 we moved the texture cache value from the VRayHDRI texture to the renderer settings; if you modified the value in the VRayHDRI texture originally, make sure the same value is in the renderer settings.

If the issue is somewhere else, then I would need that scene too - looks like something that I should definitely take a look at. 3h seems like a horribly long time for such a simple render.

Best regards,
Vlado

[vlado]

Here's the benchmark scene. I also, sent it to support@chaosgroup.com

The scene is missing two textures. Do you know where I can find them?

Best regards,
Vlado

[Nearl]

I emailed the missing textures for the benchmark scene.

Can you point me to the direction of the texture cache value in the renderer settings? I might be overlooking it.

[Nearl]

FYI, I turned off the displacement and it rendered in 0h, 4m, 52.9s

[vlado]

The texture options are in the System rollout, at the bottom:
https://docs.chaosgroup.com/display/...exture+Options

Best regards,
Vlado

[Nearl]

Bad news...I matched my texture cache value from 3.3 and it still took forever in 3.4.

Good news...I got it to render in 3.4 in 0h, 5m, 34.6s.

Bad news...Elliptical Filtering is now broken in 3.4 for displacement. Switching it back to default filtering fixed the issue. You might look into that.

Good news...I regained my sanity and won't have to down grade our farm.

Party on!

[vlado]

So for the first scene, Benchmark.zip, the difference seems to be because of a fix that was done to improve sampling when "clamp output" and "subpixel mapping" are both enabled - V-Ray 3.30 could produce horribly noisy results in some cases. This means that V-Ray 3.40 traces more rays, as can be seen in the VRaySampleRate render element, as well as the slightly reduced noise in some places.

If you disable "clamp output" and "subpixel mapping", the render times between the two versions are very similar, with V-Ray 3.40 being marginally faster.

Best regards,
Vlado

[vlado]

Elliptical Filtering is now broken in 3.4 for displacement. Switching it back to default filtering fixed the issue. You might look into that.

Yes, definitely we will. Thanks for reporting it and apologies for the inconvenience.

Best regards,
Vlado

[Nearl]

Happy to help! I've been using V-Ray for 10 years and love the product and the support.

....now can you help me with this?

http://forums.chaosgroup.com/showthr...airVrShadeable