NVIDIA GeForce RTX 5080 Founders Edition

[neatfeatguy]

The 4080 came out swinging and walked over the 3090Ti. The 5080 came out and wheezed away from the 4080, barely, and can't even catch the 4090.

I was at least expecting to see 4090 performance out of this card, not less than.

 

[Contra]

Just noticed in gpu-z, PhysX removed from 50-series?
@W1zzard

 

[Garrus]

RTX 5090 and 5080 are both insanely lame. Enough said.

 

[Chrispy_]

It isn't number of defects that goes up with square of the increased size, but the amount of dies you can get out of a single 300mm wafer goes down by square of increase in die size. Number of defects per mm2 stay the same whether you have a 1mm2 die or a 800mm2 die.

Of course you have a higher chance of a defect hitting a bigger die than a smaller one but for a mature process the actual number is quite low.

Disabling units are only benefitting them because there's a fixed amount of people that will buy a GPU at a particular price point, so it's beneficial to disable them and call it a lower end GPU even if there's no technical reason(like defective units) to do so, because you can't have a product line full of xx90 GPUs. By calling it xx80 you get to sell those instead. You can't rely solely on defective dies to sell lower end cards, because they are significantly higher volume.

Yeah that's a better way to think of it. The yield calculators always do a much better job of explaining it in pictures and stats

Example yields for 4090 (left) vs 5090 (right):

For a 25% increase in die area, fab yields drop by 11-12% and the number of perfect dies drops by 34% per wafer as a combination of both the number of dies per wafer and the lower fab yield.

I don't think the 5090's GB202 is a fully-enabled perfect die anyway, and the 4090's AD102 definitely isn't.

The 4080 came out swinging and walked over the 3090Ti. The 5080 came out and wheezed away from the 4080, barely, and can't even catch the 4090.

I was at least expecting to see 4090 performance out of this card, not less than.

But MFGEEEEeeeee!!!!
If you fake frames twice as hard, the fps numbers are twice as biggerer!
Remember kids, a 5070 is a 4090.

 

[mkppo]

Yeah that's a better way to think of it. The yield calculators always do a much better job of explaining it in pictures and stats

Example yields for 4090 (left) vs 5090 (right):
View attachment 382581View attachment 382583
For a 25% increase in die area, fab yields drop by 11-12% and the number of perfect dies drops by 34% per wafer as a combination of both the number of dies per wafer and the lower fab yield.

I don't think the 5090's GB202 is a fully-enabled perfect die anyway, and the 4090's AD102 definitely isn't.


But MFGEEEEeeeee!!!!
If you fake frames twice as hard, the fps numbers are twice as biggerer!
Remember kids, a 5070 is a 4090.

GB202 is cut down pretty hard with 170 out of 192 SM's active. With a die that massive, i'm not surprised. 5090Ti is there if nvidia wants it I suppose.

 

[lexluthermiester]

Paul did a nice summary on this GPU, definitely worth watching:

Paul is wrong on at least one point; There ARE bad GPUs.

However he makes a ton of great points, including a few that I've been taking about. This is the first time I can think of that a new gen of cards didn't surpass the previous gen flagship and yet the price didn't scale appropriately. Paul's points about NVidia's business model are seemingly spot on and it's a piss-poor, pathetically predatory thing.

Even four years ago, TSMC's yields for N5 were good.

Do you know what those numbers mean? .10% defects per square centimeter is equal to a 37% yield, give or take.. That is a terrible yield!

 

[N/A]

You have to account for the fact that GB202 is a more stretched ADA 24 by 32 mm 3:4 ratio not 5:6 used in the calculation. It probably makes a difference.

GB203 is the same as the predecessor AD103. With just the 512 units that were missing in ADA added. How generous of them.

5070 is actually going to be good but there is a catch 120 base FPs and a 480Hz display for the MFG 4x to be useful and then it's a 4090.

I don't think the 5090's GB202 is a fully-enabled perfect die anyway, and the 4090's AD102 definitely isn't.


But MFGEEEEeeeee!!!!
If you fake frames twice as hard, the fps numbers are twice as biggerer!
Remember kids, a 5070 is a 4090.

 

[hsew]

For the 5080, yes I agree there is no point to testing beyond 4k. The 5090 however...

There is value in testing the 5080, heck, even the 4080 above 4K. They’ll provide a reference for the 4090/5090. I’d also include the 7800 and 7900 series with those benchmarks as well.

 

[lexluthermiester]

TSMC 3nm yields are around 90% right now so yeah...

I'm not sure believe that. Not calling you a lair as you likely read that elsewhere, only that 90% on such a new lith process is very unlikely.

There is value in testing the 5080, heck, even the 4080 above 4K.

You seem to have missed the context of my point.

Presumably Nvidia are stockpiling defective dies to have partially-working RTX Blackwell workstation/server compute GPUs that will replace the RTX 5000 and 5880 Ada generation, either that or a 5080Ti down the line.

This. Very believable. Hell, I'd bet money on it. The performance/specs gap between the 5080 and 5090 almost demand it.

 

[hsew]

You seem to have missed the context of my point.

The entirety of what you wrote, verbatim:

“For the 5080, yes I agree there is no point to testing beyond 4k. The 5090 however...”

Was there some other context in that response?

Is the 5080 a beyond 4K card? No. I never claimed it was by asking for those benchmarks. I asked because I’d like to see how it compares to the 5090 at those resolutions. It’s basically half of a 5090.

 

[lexluthermiester]

The entirety of what you wrote, verbatim:

“For the 5080, yes I agree there is no point to testing beyond 4k. The 5090 however...”

Was there some other context in that response?

Is the 5080 a beyond 4K card? No. I never claimed it was by asking for those benchmarks. I asked because I’d like to see how it compares to the 5090 at those resolutions. It’s basically half of a 5090.

And did you read the comments preceding what I said?... Just sayin..

 

[DavidC1]

Someone said the multi fake frames are only useful if your frame rates are already high enough to be playable. In the situations where you really need it, it doesn't do much. In fact when the difference between real and fake is too high it's going to create a disjointed feeling at some point.

 

[remekra]

Someone said the multi fake frames are only useful if your frame rates are already high enough to be playable. In the situations where you really need it, it doesn't do much. In fact when the difference between real and fake is too high it's going to create a disjointed feeling at some point.

So the same as with single frame generation.
Turn on RT or PT, get to 40-70 fps with DLSS upscaling then interpolate the frames to get to 120-240. That seems to be the way Nvidia marketing goes.

Until my 5080 arrives can't say how good it feels. However I used FSR 3.1 FG on my XTX on couple of titles and experience varies from title to title.
In Wukong I was having around 90-100 with frame gen and it felt good, stalker 2 110-120 also felt better than running without it at 60.
Robocop on the other hand was far better off with FG disabled, same with Cyberpunk.

If someone is very sensitive about latency then I guess FG is not an option.

 

[lexluthermiester]

Someone said the multi fake frames are only useful if your frame rates are already high enough to be playable.

Sort of. Framegen has potential for making games that are only just barely on line of playable more smooth, but that assumes the GPU can render the interposer frames at a consistent and useful rate.