Announcement

**vladimir.nedev** · 15-01-2021, 06:53 AM

Is this the original scene for the competition or have you made some modifications to it ?

Greetings,
Vladimir Nedev

**vasile_bulgac** · 15-01-2021, 07:17 AM

Hello Vladimir,

Yes, it is the original scene for the competition, untouched by me. I have also sent you yesterday an e-mail with this issue.

**vladimir.nedev** · 15-01-2021, 10:20 AM

Hi, thanks for the info, we will do some more tests here although so far we haven't observed issues with this driver.

I tested the 460.89 driver (non-studio, I guess it's the same since the version number matches) on the 2080 Ti a couple of weeks ago.
It was on average 5-6% faster on my test suite, with some tests 8-10% faster, compared to 457.09.
So it seems this driver has indeed been optimized, whether for Vantage specifically or in general, I don't know.

However, one of the more specific tests was severely bugged (all the rest were basically unchanged).
The bug involved lights when the light tree is disabled (not the default).
I haven't seen any texture issues.

I haven't received an e-mail from you by the way. Are you sure you sent it to me ?

Greetings,
Vladimir Nedev

**vasile_bulgac** · 15-01-2021, 11:34 AM

Hi again,

About the e-mail, yes I'm 100% positive that I've sent you an e-mail regarding this matter. I've sent the e-mail from zomb_killer@hotmail.com to your address (vladimir.nedev@chaosgroup.com).

Now about this issue: you said you tested the non-studio version of the driver with a 2080 ti. I'm starting to think, maybe I'm wrong but can it be the Ampere architecture to be buggy with this driver? From what I've seen, since I replaced my Titan RTX with the 3090, indeed I've seen a strong perf. boost in games but in creative applications, believe me I have noticed slightly drops in performance.

I really hope we can figure out what is goin' on, because I have a strong feeling it can be related to drivers or the GPU architecture.

BTW, should I try a complete reinstall of the SD drivers? DDU first and then clean reinstall? Should I try that?

Thanks

**vladimir.nedev** · 15-01-2021, 11:39 AM

I just ran our performance tests on the 3080 FE with driver 460.89 and there is indeed some issue, most likely in the shading graph code, like you mentioned.
This issue is not present on 2080 Ti with the same driver, even though the speed-up provided by the driver is larger on the 2080 Ti.

For now, you can go back to a previous driver to have the issue resolved (I've been testing with 457.09).

I wasn't able to test the newer 461.09 driver to check if it fixes the issue.
The remote machine got stuck and there is no one to restart it on Friday

We will get in touch with nVidia to figure out if the issue is on our side or on their side.
We will also ask whether we can run our own tests before these drivers are released.

Greetings,
Vladimir Nedev

**vladimir.nedev** · 15-01-2021, 11:45 AM

About the e-mail, yes I'm 100% positive that I've sent you an e-mail regarding this matter. I've sent the e-mail from zomb_killer@hotmail.com to your address (vladimir.nedev@chaosgroup.com).

Now that you included your e-mail address, I was able to find the e-mail in my spam folder.

Now about this issue: you said you tested the non-studio version of the driver with a 2080 ti. I'm starting to think, maybe I'm wrong but can it be the Ampere architecture to be buggy with this driver? From what I've seen, since I replaced my Titan RTX with the 3090, indeed I've seen a strong perf. boost in games but in creative applications, believe me I have noticed slightly drops in performance.

You are right, the issue is with the Ampere architecture, like I mentioned above.
However, Vantage is much faster on the 3090 compared to the Titan RTX.
You can see some tests I did here : https://forums.chaosgroup.com/forum/...3080-on-lavina

BTW, should I try a complete reinstall of the SD drivers? DDU first and then clean reinstall? Should I try that?

I don't know if that's really necessary. A normal install of an older driver should be fine, I guess.

Greetings,
Vladimir Nedev

**vasile_bulgac** · 15-01-2021, 03:07 PM

Vantage is much faster on the 3090 compared to the Titan RTX

I know that, that was the reason why I replaced my Titan RTX with a 3090 FE

In the end I think I will get a NVIDIA^® RTX^™ A6000, especially for Vantage (One of the best real time ray traced app I have ever played with).

Anyway, in the week-end I will run some tests too (put back my Titan RTX, roll back to older drivers, etc.) and I will post my results here.

Again, thanks a lot Vladimir

Best regards,
Vasile B

**vasile_bulgac** · 16-01-2021, 11:07 AM

Hello again,

Some of my preliminary results from a series of tests I have performed today:

Scenario 1: Replaced my 3090 FE with my old Titan RTX. Ran a basic render on Vantage with Studio Driver Version 460.89 -> no issue(s).
Scenario 2: Replaced my old Titan RTX again with 3090 FE. Ran a basic render on Vantage with Studio Driver Version 460.89 -> issue present.
Scenario 3: Uninstalled the drivers using DDU, and installed Studio Driver Version 457.30. Ran a basic render with 3090 FE on Vantage -> no issue(s).

From what I have observed so far, it seems that Ampere architecture may be bugged with the latest Studio Drivers 460.89 or it may be a potential conflict between the drivers and the OS (sorry to say but Win 10 is like cancer).

**vladimir.nedev** · 17-01-2021, 08:02 AM

Scenario 1: Replaced my 3090 FE with my old Titan RTX. Ran a basic render on Vantage with Studio Driver Version 460.89 -> no issue(s).
Scenario 2: Replaced my old Titan RTX again with 3090 FE. Ran a basic render on Vantage with Studio Driver Version 460.89 -> issue present.
Scenario 3: Uninstalled the drivers using DDU, and installed Studio Driver Version 457.30. Ran a basic render with 3090 FE on Vantage -> no issue(s).

From what I have observed so far, it seems that Ampere architecture may be bugged with the latest Studio Drivers 460.89

Yes, we have come to this conclusion as well. Thanks for testing.

Greetings,
Vladimir Nedev

**vasile_bulgac** · 17-01-2021, 10:23 AM

I'm glad we sorted this out. Let's hope Nvidia will fix this annoying issue ASAP.

**vladimir.nedev** · 18-01-2021, 10:23 AM

I just tested driver 461.09 and it's bugged in the same way.

Greetings,
Vladimir Nedev

**vasile_bulgac** · 18-01-2021, 02:37 PM

I just tested driver 461.09 and it's bugged in the same way.

Damn...

Now the situation is getting serious... What if these new branch of drivers, since 460.89 are deliberately creating bugs in creative applications, like Chaos Vantage and 3ds Max with V-Ray on RTX 30 series?

Now I am really curious if RTX^™ A6000 behave the same...because if it is...well...

But, given the circumstances, I think I will be forced to reinstall my old Titan RTX, at least until this "bug" will be sorted out by some more possibly stable Nvidia SD drivers.

Personally I feel very sorry because Vantage doesn't deserve to suffer like this, because it is so awesome and I love it very much

**vladimir.nedev** · 18-01-2021, 03:20 PM

Now the situation is getting serious... What if these new branch of drivers, since 460.89 are deliberately creating bugs in creative applications, like Chaos Vantage and 3ds Max with V-Ray on RTX 30 series?

No, there is no way this is deliberate.

But, given the circumstances, I think I will be forced to reinstall my old Titan RTX, at least until this "bug" will be sorted out by some more possibly stable Nvidia SD drivers.

Why ? It's better to just use the older drivers that work.

You should generally stick to drivers that work and change them only if absolutely necessary.

I will try to explain why I think changing drivers can lead to such instabilities.
A CPU program like 3ds Max or V-Ray CPU gets (practically) fully compiled to machine instructions on our (Chaos Software) build machines.
When we change the compiler (MS Visual C, clang, Intel compiler) or just its version, we do extensive testing and only then switch our builds to the new compiler.

With GPU programs, the code is compiled to an intermediate representation on our build machine and then the final (with optimizations) compilation to machine instructions is done on the user's machine by the GPU driver.
This allows nVidia to use different instruction sets on new architectures unlike the x86 instruction set which is pretty much fixed (not counting some new instruction set extensions which are rarely used).
This also allows nVidia to optimize the compiler for better performance without the need to rebuild all software/games and reinstall them.

Unfortunately compilers are very complicated pieces of software so they can be buggy.
I suspect nVidia perform very extensive testing to ensure there aren't issues, but sometimes bugs slip undetected.

Also note that the new drivers may have simply uncovered a bug in our code, I haven't done extensive debugging yet.

Greetings,
Vladimir Nedev

**vasile_bulgac** · 19-01-2021, 02:48 AM

You ware right Vladimir, it may not be deliberate but from what I've saw since the launch of this new branch 460, the Nvidia forums are filled with hundreds, maybe thousands of users complaining of different kind of bugs/issues (90% of them are on RTX 30 series).

Another thing that may interfere with drivers development is these "feature updates" from Win 10. With every feature update more and more bugs appear, but like you've said I should generally stick to drivers that work and change them only if absolutely necessary.

Anyway, I have a strong feeling that in a week or two this issue will be resolved.

Announcement

Chaos Vantage & NVIDIA Studio Driver Version 460.89

Chaos Vantage & NVIDIA Studio Driver Version 460.89

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment