I have a really hard time liking Vray GPU.
The performance is terrible on my machine and wonder why that is.
For comparison I made a screen recording, comparing the performance to Octane, now that is fast an fun to use! I whish Vray GPU could have this level of performance.
Running the latest Vray update (v7 Update 2 hotfix 2)
Nvidia driver 580.97
Windows 11
RTX 5090 FE 32 GB
128 GB system Memory
The video doesn’t exist anymore, I watched it a few times last week. Could you please re-upload this video?
Please check this thread I wrote about performance. You will find all tips and tricks to make V-Ray GPU as fast as possible. Interactive rendering performance is one of the big advantages of V-Ray GPU, while Octane is super fast, V-Ray GPU catches up in most cases. Here is a quick comparison, https://www.youtube.com/watch?v=5Izfx9EpmOo
We are waiting on Nvidia to find a solution. This only affects Time-To-First-Pixel on Blackwell GPUs, there are 8 seconds wasted every time you start the IPR/Production rendering
Hi Muhammed,
thanks for the reply. The problem I encounter is that by every change you will see that Vray GPU is ‘Building GPU Dynamic Nodes’. This is not affected the first time, but every time and take 4-8 seconds. This you can clearly see in my screen recording.
At 4.08 you see me working in the same scene using Octane, and you can see how fast that works. It is so nice, I whish we have this speed and response with Vray GPU
Hi Muhammed, I think I have to make a correction here…
I was working on another Vray scene and the response was what I expected, really great!
So in regards the scene I was showing in the screen recording, I merged the whole scene into a new file and get similar response now as I showed with Octane :o
So clearly this was scene related. Perhaps some legacy stuff since the project was originally created in 2025.?
Oh I did down grade my nvidia driver to 580.97 Don’t know if that did the trich though :?:
Thanks for the information and the video, it is very helpful!
From the video the issue is very clear! It seems Time-to-first-pixel is the issue when you go through the timeline. Are you using Tyflow in this scene? or displacement?
I’m glad you have a conclusion now. It looks like you hit a bug in the original version of the scene, is there a chance we could get this scene to investigate?
I downloaded the video, feel free to delete it now if you wish.
We are in the process of testing a new recommended driver. Expect it to update later next week.
Sooo… When buying a 5090, I wasn’t aware that even after 6 months release, there are still issues with the 50XX series? I’m getting only 11K Paths in benchmark, while most of the results are around 15K. I don’t have the recommended studio drivers though, might this be an issue with Benchmark as well? I’m not using it (yet) for Vantage or V-Ray GPU, but was just testing my new GPU. I’m on Windows 11, and running latest 581.29 driver. I can see people having 16K paths on W11 with these drivers. What am I doing wrong?
There are currently no issues with the RTX 5090, the kernel compilation issue was solved by Nvidia earlier this month.
Very odd what you describe here. I suggest clean install of drivers, the latest studio driver will be the new recommended driver this week. Please use it.
You need to uninstall, with clean install option enabled. Then restart the machine then test. The Benchmark tool is fine on my end.
If you still have issues with performance could we please schedule a meeting to take a look at you machine
Myeah it’s strange. I’ve updated the drivers (clean install). Here you can see my scores with 2 different drivers, doesn’t make much difference. I even installed MSI afterburner to overclock it a bit but that didn’t do much either (2 latest scores are with afterburner).
This is a new machine, latest bios, drivers, etc. MSI GeForce RTX 5090 32G SUPRIM SOC Threadripper 9980X, 256GB ECC DDR5 RAM running at 5600 Mhz on a Gigabyte TRX50 AI TOP motherboard.
Your scores are normal and actually higher than mine, I’m getting 10500 in CUDA. What if you swap your engine to RTX? What scores do you get?
The benchmark site will not filter CUDA and RTX results, they are kept in the same place.
And… we are back with performance issues in 3Ds Max and Vray GPU…
I am preparing an upcoming project which I would like to do in 3Ds Max. So I collected the main elements I needed and did some testing. How discouraged I was (again) by using Vray GPU in 3ds Max.
So I transferred the scene over to Cinema and wow that is the performance I want. @Muhammed_Hamed Please take a look at the following screen recording of both applications.
Can you tell me that the performance you see in 3Ds Max is what it should be? It looks and feels crippled and no fun to work with.
@menno_vandijk Thanks for the video! It is very helpful
You should be able to achieve the same interactivity you have in C4D, in 3Ds Max
In the performance tab, please use these settings,
IPR Rays Per Pixel 2
IPR Ray Bundle Size 256
Undersampling 2
In the GI tab, please use these settings,
Secondary GI engine = BF, C4D uses BF/BF GI by default. LC is usually only useful for Interactive rendering for interior scenes.
On another note, How many Render elements are you using? are you using Light Mix?
Is there a chance you could upload this scene and send it to me, muhammed.hamed@chaos.com
I will test it on my 5090 and report back to you. From my quick testing now, 7.2 has no issues with GPU interactive rendering
I removed all of my render elements and switched to BF - BF with CPU, and the interactivity is more or less the same. When the orbit or pan moves are too fast V-Ray just waits until I stop moving, something I don’t see happening in the C4D videos. Are the settings in C4D for CPU also standard set to BF - BF?
