Nvidia shows signs in [2023]

  • Thread starter Deleted member 2197
  • Start date
Status
Not open for further replies.
What's the relationship between Reflex and DLSS SR and/or tensors? Since Reflex is available on 900 series GPUs, and a basic understanding of the tech when it was released, I believe the answer is nothing. However, that is also now "DLSS3".

Not confusing at all for anything, of course we're going to know what we're talking about when someone mentions DLSS3.

I'll just reiterate my point that we should no longer use the term DLSS since it's meaningless now, reduced to yet another marketing gimmick.
 
I do wonder if all these results and Nvidia talking to Intel will mean that they'll move to them especially if TSMC still maintains its relationship with Apple being No 1 with the latest node monopoly a Year or two.
 
What's the relationship between Reflex and DLSS SR and/or tensors? Since Reflex is available on 900 series GPUs, and a basic understanding of the tech when it was released, I believe the answer is nothing. However, that is also now "DLSS3".

Reflex is part of FrameGeneration.
 
They should have marketed the suite as "DLX", as a subset of RTX.
 
Reflex is part of FrameGeneration.
But can be used separately which is no problem. When FG was first introduced one Nvidia engineer stated it was possible to get FG to run on older cards but would not get the full performance benefit as seen on the RTX 4000 series that have OFA. Maybe they have a surprise in store ...
 

Insane gross margins, Nvidia is making a shit ton of money on AI DC sales.
Gaming is also up however, +21% Y/Y.

I'll just reiterate my point that we should no longer use the term DLSS since it's meaningless now, reduced to yet another marketing gimmick.
DLSS is an SDK containing a bunch of tech. Whether you "should" or not use the term depends solely on your understanding of what it is. If you want to talk about Frame Generation feature specifically then just do that instead of talking about "DLSS 3" - this would solve all issues and save us from the last couple of pages of the discussion.
 
When FG was first introduced one Nvidia engineer stated it was possible to get FG to run on older cards but would not get the full performance benefit as seen on the RTX 4000 series that have OFA.
From memory the older RTX cards had OFA but only the new iteration in 4000 series was good enough for FG.
 
So I guess the talk of Nvidia’s gpu pricing being way too high was incorrect. People are buying it enough that they made more money than before the 40 series launch. Guess I’ll expect 50 series to launch at luxury pricing and then maybe slowly drop over a couple of years.
 
I do wonder if all these results and Nvidia talking to Intel will mean that they'll move to them especially if TSMC still maintains its relationship with Apple being No 1 with the latest node monopoly a Year or two

Nvidia will at least triple the production of its top H100 AI processor in 2024 to 1.5mn to 2mn H100s, a massive jump from the 500,000 expected this year, the Financial Times reports, noting the AI processors are already sold out into 2024.
 
Last edited:
So I guess the talk of Nvidia’s gpu pricing being way too high was incorrect. People are buying it enough that they made more money than before the 40 series launch.
30 series had a higher ASP than 40 series so far even before we account for the mining price inflation on the 30 series.
It's not that surprising to see gaming side selling well enough IMO. But there really is no reason to expect 50 series to launch in a significantly higher price range - unless they'll want to make some Titan class product in it again.
Current GPU pricing isn't "too high", it's just unimpressive if you're looking at perf/price gains over the previous GPU generation.
 
30 series had a higher ASP than 40 series so far even before we account for the mining price inflation on the 30 series.
It's not that surprising to see gaming side selling well enough IMO. But there really is no reason to expect 50 series to launch in a significantly higher price range - unless they'll want to make some Titan class product in it again.
Current GPU pricing isn't "too high", it's just unimpressive if you're looking at perf/price gains over the previous GPU generation.

It's unimpressive looking at price/perf gains but people are buying it so ... ¯\_(ツ)_/¯
 
Last edited:
A somewhat good illustration about DLSS versions.

u4tmp23diujb1-jpg.9444
 

Attachments

  • u4tmp23diujb1.jpg
    u4tmp23diujb1.jpg
    63.6 KB · Views: 121
A somewhat good illustration about DLSS versions.

u4tmp23diujb1-jpg.9444
At this point Nvidia should just allow frame generation on 20 & 30 series GPUs (since all supported games allow you to toggle it independently from DLSS) and simply say "your results may vary" and be done with it.
 
At this point Nvidia should just allow frame generation on 20 & 30 series GPUs (since all supported games allow you to toggle it independently from DLSS) and simply say "your results may vary" and be done with it.
Well I'm sure part of the reason is to encourage people to buy 40 series GPUs instead...
 
Well I'm sure part of the reason is to encourage people to buy 40 series GPUs instead...
Sure, but then you'd think Ray Reconstruction would have been limited to 40 series GPUs as well, and they could have simply continued with "DLSS3 is 40 series only".
 
It's unimpressive looking at price/perf gains but people are buying it so ... ¯\_(ツ)_/¯
It's unimpressive but is still an improvement on previous generation and at the very worst is at about the same perf/price. If people bought 30 series then why wouldn't they buy 40?

At this point Nvidia should just allow frame generation on 20 & 30 series GPUs (since all supported games allow you to toggle it independently from DLSS) and simply say "your results may vary" and be done with it.
So Nvidia should just make DLSS worse for no other reason but because some people can't be bothered to read?
 
So Nvidia should just make DLSS worse for no other reason but because some people can't be bothered to read?
"I used to think ray tracing was awesome. Then Nvidia enabled it on the GTX 10 series and now I think ray tracing sucks." - No one ever. 😉

Seriously though, I can understand why they'd not enable Frame Gen on 20 & 30 series if it in fact would be unusably slow on them.
1. From a marketing point of view, articles and Youtube videos out there trashing Frame Gen for doing nothing but adding input latency on the 20 & 30 series might dilute the intended signal that it's only meant to be used on the 40 series.
2. It makes things clearer for the normies. I can envision some poor bastards out there playing a game with Frame Gen on and it would add nothing but input latency and visual artifacts in their game not understanding what setting was responsible for that.

Enabling it a year or two after launch (enough time for marketing to have made the message clear) from some hard-to-find option in NVCP that also dramatically warns the user of abysmal performance on their GPU might be one way to open it up. I doubt it'll happen though.
 
Status
Not open for further replies.
Back
Top