NVIDIA discussion [2025]

[trinibwoy]

So what's the deal with this "The Information" outfit? They're just leaking false info to manipulate Nvidia's stock price?

 

[DavidGraham]

So what's the deal with this "The Information" outfit? They're just leaking false info to manipulate Nvidia's stock price?

Precisely, they do this every quarter, usually it drops the stock a little, but this time it appears to have very little effect.

 

[DavidGraham]

NVIDIA has a big supercomputer running 24/7, 365 days a year improving DLSS. And it's been doing that for six years.

Actually, we have a big supercomputer at Nvidia, with many 1000s of our latest and greatest GPUs, that is running 24/7, 365 days a year improving DLSS. And it's been doing that for six years

What we're doing during this process is we're analysing failures. When the DLSS model fails it looks like ghosting or flickering or blurriness. And, you know, we find failures in many of the games we're looking at and we try to figure out what's going on

why does the model make the wrong choice about how to draw the image there?

We then find ways to augment our training data set. Our training data sets are always growing. We're compiling examples of what good graphics looks like and what difficult problems DLSS needs to solve.

"We put those in our training set, and then we retrain the model, and then we test across hundreds of games in order to figure out how to make DLSS better. So, that's the process."

Turns out there's 'a big supercomputer at Nvidia… running 24/7, 365 days a year improving DLSS. And it's been doing that for six years'

It's all part of the process.

www.pcgamer.com

 

[pjbliverpool]

NVIDIA has a big supercomputer running 24/7, 365 days a year improving DLSS. And it's been doing that for six years.

Turns out there's 'a big supercomputer at Nvidia… running 24/7, 365 days a year improving DLSS. And it's been doing that for six years'

It's all part of the process.

www.pcgamer.com

And that's exactly why every other vendor will have a very hard time catching up with DLSS. I've been saying this for years.

 

[trinibwoy]

And that's exactly why every other vendor will have a very hard time catching up with DLSS. I've been saying this for years.

Isn't XeSS pretty decent already?

 

[Albuquerque]

In my experience, I prefer XeSS results to FSR -- at least in Cyberpunk. I'm not sure how relevant it is for other games...

 

[DavidGraham]

Isn't XeSS pretty decent already?

It is decent, but DLSS still has superior upscaling, and it has Ray Reconstruction and Frame Generation.

What's more, the transition to the transformer model will boost DLSS quality in all categories even further.

 

[trinibwoy]

It is decent, but DLSS still has superior upscaling, and it has Ray Reconstruction and Frame Generation.

What's more, the transition to the transformer model will boost DLSS quality in all categories even further.

Yeah I know but presumably Intel hasn’t been training their model for 6 years so maybe it doesn’t take much to get decent results. 80/20 rule.

 

[Kaotik]

It is decent, but DLSS still has superior upscaling, and it has Ray Reconstruction and Frame Generation.

What's more, the transition to the transformer model will boost DLSS quality in all categories even further.

XeSS has Frame Generation too.

 

[arandomguy]

NVIDIA has a big supercomputer running 24/7, 365 days a year improving DLSS. And it's been doing that for six years.

This is one of reasons why when I see people saying something like this should be open sourced or made vendor neutral I ask them about specifics.

I know the consumer side might not be want to hear this but sans directly monetizing software improvements like this it's going to need to be captured via the margins on hardware.

 

[Potato Head]

Yeah I know but presumably Intel hasn’t been training their model for 6 years so maybe it doesn’t take much to get decent results. 80/20 rule.

Intel hasn’t had XESS implemented in over 600 games to gather training data from.

 

[Broopster]

They have moved forward their plans in a great way, Blackwell Ultra (B300) is coming up 6 months early, and Rubin is the same. It's the same situation as H100 and H200. NVIDIA keeps pumping up new hardware and customers buy what's available according to their budget, order volume and development plans.

As for reports of delay:

https://twitter.com/x/status/1879346126492737860

The answer is pretty obvious - until whoever keeps leaking them covers their shorts! (Only half joking).