News NVIDIA GeForce RTX 5080 Is 2x Faster Than 4080 At $999, RTX 5070 Ti 2x Faster Than 4070 Ti At $769, RTX 5070 Faster Than 4090 For $549

[Tracer_Bullet]

I was wondering when someone was going to come up with that. 1000 and 3000 series cards imo are probably a bit of an outlier.
Could Nvidia have done better for the 5000 series, maybe. Its possible they've hit a performance plateau.

They are prioritizing for AI. So probably the design is less than optimal for gaming and gaming doesnt get the latest chip node.
Plus VRAM nonsense for segmentation. Its upto competition to try and take market share, NVIDIA only seems to respond then.

People have to stop making NVIDIA the default choice. Else what happens is if AMD comes up with something good, then Nvidia reduces price and people buy Nvidia.
So for AMD, you dont win. You only get less margin. But anyway, they said that they will target market share this time and Nvidia cards seem to be bad, maybe this gen there is a chance,

I sincerely doubt that. At best, it will match 4070 Ti super performance. AMD loves shooting themselves in the foot, so better to keep low expectations

Best case is it matches 4070ti super in RT and 4080 in raster. But yeah, AMD often disappoints and price is also unknown.
FSR4 becoming good will make them much more competitive.

 

[mayank11280]

 

[iPwnz]

Is this a real tweet?

 

[mayank11280]

Is this a real tweet?

Yes!

https://twitter.com/x/status/1885363355734413321

 

[k660]

I sincerely doubt that. At best, it will match 4070 Ti super performance. AMD loves shooting themselves in the foot, so better to keep low expectations

ekdum satya vachan, everywhere i go i see this similar comments

I don't have much experience with it especially anything other than text generation models, but the 3090 about which you asked should be pretty alright for most of those use cases. You can also look up pretty much everything you want on YouTube, different models running of different hardware and other queries you might have or just create a thread asking for help.

i will keep that your exp in text generation models in my mind.
also, sighs....
the 3090 gpus arent enough unfortunately. relatively only good thing about it is its vram amount other than that as the workflows get complex, as fixes and detailers climb on it slows down a lot.
for now 5090 (best for a consumer grade gpu) is the only gpu who has put a significant dent like its 40-60% faster in img generation compared to 4090 but the amount of power it is uses holy sh1t it makes it not worth it, (maybe new process node might be a help there) .
on youtube most of them test it on 512x512 or dont have enough gpus with them to test on different models (and yt search engine has gone down the drain), that is the ai community isnt there yet, like the ones we have e.g. ltt,gn,hd-unbx and similar who does rigourous and in-depth tests. so i was and am looking for as many people with img generation exp as much possible to gather info.
thank you for your input btw

 

[Skoomahed]

ekdum satya vachan, everywhere i go i see this similar comments

i will keep that your exp in text generation models in my mind.
also, sighs....
the 3090 gpus arent enough unfortunately. relatively only good thing about it is its vram amount other than that as the workflows get complex, as fixes and detailers climb on it slows down a lot.
for now 5090 (best for a consumer grade gpu) is the only gpu who has put a significant dent like its 40-60% faster in img generation compared to 4090 but the amount of power it is uses holy sh1t it makes it not worth it, (maybe new process node might be a help there) .
on youtube most of them test it on 512x512 or dont have enough gpus with them to test on different models (and yt search engine has gone down the drain), that is the ai community isnt there yet, like the ones we have e.g. ltt,gn,hd-unbx and similar who does rigourous and in-depth tests. so i was and am looking for as many people with img generation exp as much possible to gather info.
thank you for your input btw

I have tried playing with stable diffusion, generating a 512x768 or vice versa image takes almost 30 seconds on my laptop 1660 ti. lil guy is chugging along somehow.

Found this video by the way

 

[Kaching999]

I'm hoping for better in the next gen. However If the RTX 6000 series underwhelms the same way that would be problematic.

I don't think that the 6000 series will have similar gains, I think it will be much better something in line with the 3000 and 4000 series gains. But I think that the series is going to take much more longer to arrive than the average cycle.

There were rumours of the AMD 9000 series GPU's offering RTX 4080 Super like performance at $ 549. If that pans out, fantastic.
I think many gamers will ditch Nvidia and jump onto the AMD bandwagon if that happens. We'll just have to wait till March.

I hope of that happening too, it is rumoured to have 4080 raster and 4070 Ti super ray tracing performance and is going to cost around $600-700.

I wouldn't be surprised if Nvidia comes out with 5070 Super, 5070 TI Super, 5080 Super if AMD offers competitive pricing for their products.
On the whole however GPU prices have been on an increasing trend for quite a while. Again if they come down fantastic. Will they? Your guess is as good as mine.
Basically the used markets see more activity and become more important for gamers.

Well the Super and Ti models are obviously going to come and they are going to arrive atleast 1-1.5 years from now. The 5080 will have many more models just like the 4070 series did last generation. I think that the prices are going to come down just like they did last time with the release of the Super and Ti series most probably because the 50 series sales are going to be much worse similar to what happened with the 4080 with it's initial price.

I was wondering when someone was going to come up with that. 1000 and 3000 series cards imo are probably a bit of an outlier.
Could Nvidia have done better for the 5000 series, maybe. Its possible they've hit a performance plateau.

The 1000 and 3000 series were big jumps due to Nvidia getting better deals for larger GPU dies, the 3080 had the same dies as the 3090. But the cost of 4nm dies from TSMC has increased mainly due to the demand an in order to keep their investor and board partners happy Nvidia has to maintain the same profit percentage margins.

They have not hit a performance plateau, the reason the 5090 series has such low gains is because it uses the same 4N node for it's dies which the 4000 series used there have been high leaps in performance each generation due them using smaller nodes and thus extracting power out of hit more efficiently but that hasn't happened this time and that is also why we have seen such high power requirements for any meaningful performance gain.

I think there are a lot of unsold 4080 and 4080 super still sitting on shelves mainly because that was also bad value and thus one should buy those at a much cheaper price than spending 1.5l on basically the same card.