Nvidia Geforce Drivers Release Announcement thread

[Scott_Arm]

That's amazing for 720p. Is the ghosting under control?

At 1440p DLSS Performance I think you could reasonably run path tracing on a 4070. Add on FG which I've found works fine at 45-50fps base and you could have a pretty decent experience. Wouldn't be good for competitive shooters but I don't think those games have PT anyway.

I've only done a really quick look, so not sure about ghosting. There wasn't any noticeable ghosting on those particles around the gemstone in Remnant 2, and those are constantly moving. Didn't see anything that stood out for me in Remnant but wasn't look for it. I do generally adjust my settings to get 120 fps or more, and that really helps minimize issues with DLSS anyway.

 

[homerdog]

I'm only half way through the video and already very confused. Is the new FG model used on Ada at all? The VRAM saving would be most welcome but I'm not sure if Ada stays with the old model or gets the new model but capped at 2X FG.

 

[arandomguy]

I'm only half way through the video and already very confused. Is the new FG model used on Ada at all? The VRAM saving would be most welcome but I'm not sure if Ada stays with the old model or gets the new model but capped at 2X FG.

The frame gen model is updated on Ada as well according to Nvidia.

 

[Scott_Arm]

I'm only half way through the video and already very confused. Is the new FG model used on Ada at all? The VRAM saving would be most welcome but I'm not sure if Ada stays with the old model or gets the new model but capped at 2X FG.

This might be the worst video they've ever done. I just finished watching. He spends most of the time memeing and intentionally making things more confusing than they actually are. It's like listening to a person complain about their job for 40 minutes. He could have made a good video, but he spent most of it griping.

40 series gets the enhanced version of the old FG (2x) that reduced memory usage and increased "performance." It is the version for the optical flow accelerator. MFG is exclusive to the 50 series.

 

[homerdog]

This might be the worst video they've ever done. I just finished watching. He spends most of the time memeing and intentionally making things more confusing than they actually are. It's like listening to a person complain about their job for 40 minutes. He could have made a good video, but he spent most of it griping.

40 series gets the enhanced version of the old FG (2x) that reduced memory usage and increased "performance." It is the version for the optical flow accelerator. MFG is exclusive to the 50 series.

View attachment 13010

I should've watched for another 60 seconds and my question would've been answered

The video was very annoying to watch. Like the script was written by a budget Leslie Nielsen.

 

[Florin]

I should've watched for another 60 seconds and my question would've been answered

The video was very annoying to watch. Like the script was written by a budget Leslie Nielsen.

The very first sentence of the video is 'Nvidia is being a dick'. Not the most auspicious start. But ok at least he's not leaving any uncertainty about where he stands. What makes the video so unpleasant is that he then proceeds to be a dick for pretty much the rest of it. It's just exhausting.

And it's a shame because there's a really good point there about branching off an obsolete version of PresentMon.

I find it ironic how people keep bringing up this complaint that they can't tell how many frames are real and how many are fake. I mean, sure. But also, that pretty much means it is working, right?

 

[Scott_Arm]

@Florin The whole thing about PresentMon is interesting, but there are good reasons why a company might want to branch off of an open tool like PresentMon, especially during R&D, and he explores none of them. He mostly seems mad that it makes his job more difficult because it makes it harder to do apples-to-apples comparisons, which is something a consumer doesn't need to do. Most people are not running multiple PCs and trying to compare results across them. Also, the stuff about the review guide is a bit weird, because he acts like the information isn't necessary because intel already solved it in PresentMon, but the whole point of review guides is to inform reviewers. Steve himself shows that he had to personally educate JayzTwoCents, so clearly the reviewing community had not gotten the memo, which means the information in the review guide is fully necessary. Just because GN knows doesn't mean everyone else knows.

Would it be better if Nvidia participated in PresentMon? Absolutely. Honestly think it would help if PresentMon wasn't Intel branded. I don't think companies should care about that stuff, but they do.