New Nvidia drivers

[Comixbooks]

https://www.reddit.com/r/nvidia/s/2Rb9bhi0JQ

700 comments in 24hrs every comment is negative about the latest drivers 573.83. Good reason to hold off buying a 5000 series card for now. The previous 572.70 seems pretty stable.

 

[Wade88]

Weird stuff like this is why I don't buy brank spanking new cpus, chipsets, or gpus anymore.

 

[DogsofJune]

Weird stuff like this is why I don't buy brank spanking new cpus, chipsets, or gpus anymore.

It affects 4000 series as well.

I have black screen on a 4080 and quirky issues on a 4070. Had to revert back. The 4060 just isn't getting updated since the blue screens debacle. So I don't know why people think it's just the new GPU's that have problems.
I haven't tried them on a 3000 series card, but I suspect.

It's the drivers.

 

[Niner]

Good reason to hold off buying a 5000 series card for now.

It doesn’t just affect the 50 series cards as others have mentioned. I had some minor black screen issues with my 4090 on boot, but those went away as the newer drivers came out. No issues with two 50 series cards so far..

 

[III_Slyflyer_III]

Which driver is that you are running? 572.75? I've been having a lot of issues with all the drivers past the 566 series with black screens on my 4090.

I am running 572.70 and have 0 issues with my 4090 now. No black screen when waking up my monitor and games run great. The only issue I noticed is in some games, Frame Generation seemed "Choppy", however, I noticed that was because it was not automatically enabling vsync for that game (which is required for Frame Gen). So that's an easy fix on a per game basis if the issue presents itself.

 

[Wade88]

I haven't had any black screen issues with 4080s in a x570 system and a b650 w/o any niggles. I am familiar with these issues with generations of nvidia cards past but some of it was due to crappy monitors and cables too.

 

[Wade88]

Enable the swap file completely windows stock, not altered and it should work better. Even with alot of ram this game doesn't like the swap file changed or disabled.

I would get a refund for that game on steam if it bitched about that.

 

[Absalom]

I'm surprised no one has posted this yet. Nvidia just updated the Nvidia App. You can now set DLSS Super Resolution render resolution to an arbitrary %:

You'll probably have to force the game specifically in the app. Note the min and max though.

About to find out what 33% looks like.

Edit: Ok, tried it out in Stalker 2. Some things to note with 33%:

• Looks pretty good when standing still and there's no motion.
• Becomes a blurry mess when objects or the screen is in motion. Power lines have ugly aliasing or gaps.
• Performance might be better? I can't tell honestly. Stalker 2 is probably not the best example of using this.

Don't trust screenshots. They'll make it look like it's better than it actually is.

I'm glad they added this. Because the more user options, the better. Some might put it at 90% just to give them that extra bit of fps vs DLAA.

 

[Sycraft]

One thing to note about custom resolutions is that it is possible they could look worse even if they are higher than one of the preset levels because the data the neural networks get trained on is at the preset resolutions.

Not saying don't try it, I am going to for sure but it is completely possible that Quality will end up looking better than setting it to 75%, even though 75% is a higher input rez.

Also completely right about testing in motion with things like DLSS because both bad AND good really get revealed there. That's also where it shines for dealing with things like shimmering vs something like SMAA.

 

[GoldenTiger]

75%

Just to note, 75% is actually a dlss "ultra quality" preset that most games don't implement. I don't know about arbitrary resolution percentages being better or worse as you postulate, but they very well could be.

 

[Absalom]

One thing to note about custom resolutions is that it is possible they could look worse even if they are higher than one of the preset levels because the data the neural networks get trained on is at the preset resolutions.

Not saying don't try it, I am going to for sure but it is completely possible that Quality will end up looking better than setting it to 75%, even though 75% is a higher input rez.

Also completely right about testing in motion with things like DLSS because both bad AND good really get revealed there. That's also where it shines for dealing with things like shimmering vs something like SMAA.

It's probably game dependent too. In Stalker 2 I noted I was not gaining any fps by setting the render resolution lower (33% vs Quality) indicating that something else is hamstringing the performance. But I could tell a visual difference, which Quality looked better obviously.

 

[Sycraft]

It's probably game dependent too. In Stalker 2 I noted I was not gaining any fps by setting the render resolution lower (33% vs Quality) indicating that something else is hamstringing the performance. But I could tell a visual difference, which Quality looked better obviously.

Stalker 2 I think gets fairly CPU hung up and particularly with your older CPU it is likely that. But ya, the performance gains are always something you need to check.

 

[Absalom]

Stalker 2 I think gets fairly CPU hung up and particularly with your older CPU it is likely that. But ya, the performance gains are always something you need to check.

I am brutally reminded of that, especially when it comes to minimum framerates. But, I know enough that it's not the only factor. And averages should still increase, which in this case, they didn't.

In HL2-RTX demo, my framerate literally doubled when going from Performance to Ultra Performance. While I'm not going to reinstall the demo to experiment with this, I'd say framerates in that will go way up with 33%. It was pretty obvious HL2-RTX was not CPU bound in any way.

 

[Sycraft]

I am brutally reminded of that, especially when it comes to minimum framerates. But, I know enough that it's not the only factor. And averages should still increase, which in this case, they didn't.

In HL2-RTX demo, my framerate literally doubled when going from Performance to Ultra Performance. While I'm not going to reinstall the demo to experiment with this, I'd say framerates in that will go way up with 33%. It was pretty obvious HL2-RTX was not CPU bound in any way.

There may have been another issue but Stalker 2 seems to uniquely have issues with hitting a single CPU thread hard and getting limited. HL2 RTX is almost the complete opposite, it is extremely GPU limited, given that raytracing is such a shader-heavy process and more pixels = more slows for RT.