NVIDIA GeForce RTX 50 “Blackwell” Series to Usher In a New World of AI Computer Graphics This Month, Led by RTX 5090 ($1,999) with up to 2x the...

ON the topic of DLSS, I was resistant to DLSS upscaling for a long time, what I have enjoyed the most is DLAA (I'm an image quality fiend), and apparently DLSS 4 AI Transformer model should have better DLAA as well, so I'm looking forward to that. I would only use DLSS Upscaling if I'm not achieving a playable framerate with all the bells and whistles in the game turned on, it's an 'only if i really need it' function for me.

As for Frame Generation, I will never use it. I don't like the idea of it, I don't like the implementation of it, and I could go into a lot of detail about that. In summary, it bypasses going through the game engine/pipeline, and hallucinates what it thinks the image should look like, I'm not ok with that. I am also not on board with the fact that it bypasses the responsiveness part of the pipeline, and skips right ahead to the smoothness aspect, by bypasses responsiveness, and adds latency, it just doesn't feel the same to me. DLSS Upscaling at least actually improves responsiveness, but since Frame Gen doesn't add to the responsiveness, it is completely removed from the feel of the game, it's floaty. I'm sure DLSS 4 Frame Gen will look better, but it's not just the looks for me, it's also how the game feels as well. There's a lot of technical reasons I am just not on board with Frame Gen as a concept.

Why even introduce the potential fof corruption, and errors, when I can just avoid it all together and not use Frame Gen.
 
Last edited:
What really chaps my arse is when these AIB's sell a card with a waterblock on it and charge an extra $600-800 for it. When I can buy the same card air cooled and a separate block for another $300 and save $300-500.

Gigabyte and Asrock, specifically, did this with the 7900XTX and 4090. I remember that Asrock 7900XTX being $1800 and the Gigabyte 4090 was $3000 or something retarded.
 
Yep, not a fan of Frame Gen. I've been tinkering with it on some newer games and it's a mixed experience, at best. Sure the FPS counters may jump but you can definitely tell it's not real. Add in the increased lag and it becomes something I just don't want to use unless I'm testing various max settings but I'm tending to avoid now.

I've had mostly positive experiences with DLSS and am fine using it if I can't get 60ish, or more at native 4K. There have even been some odd instances where DLSS seems to improve over native AA solutions (I recommend checking out the slider comparisons for KCD 2 from Game GPU linked in today's post, SMAA looks like crap compared to DLSS for it).

I didn't really catch on to DLAA until recently and now it's one of the first things I look for when firing up a game for the first time and I hope more developers support it.

What really chaps my arse is when these AIB's sell a card with a waterblock on it and charge an extra $600-800 for it. When I can buy the same card air cooled and a separate block for another $300 and save $300-500.
Too true and here's hoping MSI will be the ticket again. The Liquid Surprim X 4090 was ~$200 ish more than the FE which I and many others found it to be more than fair and performed well.
 
What really chaps my arse is when these AIB's sell a card with a waterblock on it and charge an extra $600-800 for it. When I can buy the same card air cooled and a separate block for another $300 and save $300-500.

Gigabyte and Asrock, specifically, did this with the 7900XTX and 4090. I remember that Asrock 7900XTX being $1800 and the Gigabyte 4090 was $3000 or something retarded.
I was talking to Gigabyte about this - basically said they got one initial shipment of the 4090 aio model and that was about it for the last couple years. So I'd expect the 3k pricing to be more scalper driven than anything. Their 5090 aio model will likely similar volume. No insight on the ASRock side...
 
I'd be more interested in the 5080, to be honest. US$2000 for the 5090, even if justified by performance, is just too much to swallow. The 5080 should still be the second fastest GPU available for several years upon release, and at half the price...
 
The 5080 should still be the second fastest GPU available for several years upon release, and at half the price...
Yep, but just keep an eye out that there could be another version out later on for the 5080, be it SUPER/Ti. The cutdown mobile version is already getting a VRAM upgrade that makes more sense than the announced desktop version for 4K gaming, which 4080/5080 are more than capable of when using DLSS and sometimes native.

 
Yep, but just keep an eye out that there could be another version out later on for the 5080, be it SUPER/Ti. The cutdown mobile version is already getting a VRAM upgrade that makes more sense than the announced desktop version for 4K gaming, which 4080/5080 are more than capable of when using DLSS and sometimes native.

They might have dropped the memory controller bus width to 192bit and used higher density DRAM to hit that capacity though; not sure that's indicative of a future desktop part.

However, apparently 3Gbit GDDR7 is coming (like with DDR5 24GB / 48GB modules) which may allow 24GB on a 256bit bus without having a GTX970 situation.
 
Become a Patron!
Back
Top