Nvidia Geforce RTX 5090 reviews

[trinibwoy]

The different reviewer conclusions are interesting. GN thinks nobody should buy the 5080. Because clearly everyone already owns a 4080. TPU on the other hand recommends the 5080 to the consternation of some readers. The logic there is that if you have $1000 to spend that’s the card he would recommend.

The second approach is obviously more in line with how people actually make purchase decisions but the emotional takes are fun to watch too even though they’re completely useless.

 

[techuse]

It isn’t unreasonable for GN to have that opinion. If you were looking to spend 1000 on a video card and the 4080 didn’t compel you why would this?

 

[trinibwoy]

It isn’t unreasonable for GN to have that opinion. If you were looking to spend 1000 on a video card and the 4080 didn’t compel you why would this?

I can buy a $1000 card today even though I didn’t want to last month. Time and circumstances aren’t static.

 

[Nappe1]

The different reviewer conclusions are interesting. GN thinks nobody should buy the 5080. Because clearly everyone already owns a 4080. TPU on the other hand recommends the 5080 to the consternation of some readers. The logic there is that if you have $1000 to spend that’s the card he would recommend.

The second approach is obviously more in line with how people actually make purchase decisions but the emotional takes are fun to watch too even though they’re completely useless.

In U.S.
For $1000, I think 5080 would not be extremely bad deal for some one like me with 3070Ti or less. More like okayish... slightly steep, but not too bad.

Unfortunately $1000 (or 962 euros, 34 cents, according to xe.com ) is not nearly enough here to get one. MSRP in europe was 1229 euros and after those cards were sold about in one minute, now AIB card's average price is here (finland, 25.5 % vat included) slightly under 1,600 euros. (which is $1,661.81. Even if you rip off the vat, you still end up to price 1332.00 USD.)

Too bad for us.

 

[Florin]

The 4080 and 4090 sure seem to have undergone some major rehabilitation as time has gone on. Those cards also received quite a few mocking and scathing reviews at the time but we've now arrived at a point where some of the same people are making them sound like really solid choices.

Okay, the 4080 price cut certainly helped a bit but it's still fundamentally the same device. And the 4090 never even got that. If anything the opposite happened.

Does this mean anything? To me, two things - perspective changes everything, and shitting on everything is just a core part of some reviewers' brands.

 

[trinibwoy]

shitting on everything is just a core part of some reviewers' brands.

Ha sad but true.

 

[arandomguy]

The different reviewer conclusions are interesting. GN thinks nobody should buy the 5080. Because clearly everyone already owns a 4080. TPU on the other hand recommends the 5080 to the consternation of some readers. The logic there is that if you have $1000 to spend that’s the card he would recommend.

The second approach is obviously more in line with how people actually make purchase decisions but the emotional takes are fun to watch too even though they’re completely useless.

I feel relative to Nvidia's own stack in theory, as we don't actually know how the rest of the stack will play out for sure yet especially in terms of street prices, I don't feel the RTX 5080 is really a good buy either. Just compared against the 5070ti in theory it might only be 20% faster (could end up less) but you will pay 33% more MSRP to MSRP. It also has the same VRAM, so that would not be a difference either. The only real selling point I feel is it has a FE model while the 5070ti won't. We won't know for certain right now but at least speculatively I feel the 5070, 5070ti and 5090 are really the ones to consider depending on criteria.

As an aside I just realized something interesting about the 5080 FE. On paper officially it's boost clock (which is what the official TFLOP numbers are derived from) is ~100mhz higher than the 4080 but if you look at TPUs review as they examine effective clockspeeds in practice it's ~100mhz lower than their 4080 sample. The one exception from their data is Cyberpunk RT (they don't have other RT numbers) in which they end up roughly the same and interestingly enough that benchmark has their performance delta on the larger side at 18%. We don't have much samples but the 5090 compared to the 4090 with their numbers matches the official boost spec numbers in that effective clocks lower on the 5090 vs the 4090.

https://images.nvidia.com/aem-dam/S...ell/nvidia-rtx-blackwell-gpu-architecture.pdf (see appendix b)

5080 FE clocks -

NVIDIA GeForce RTX 5080 Founders Edition Review

The NVIDIA GeForce RTX 5080 is out now. Priced at $1000 it includes all the new GeForce RTX 50 features and offers good gaming performance. On the other hand, the gen-over-gen performance increase is smaller than expected, but NVIDIA didn't raise their pricing either.

www.techpowerup.com

4080 FE clocks -

NVIDIA GeForce RTX 4080 Founders Edition Review - Energy-Efficient Beast

GeForce RTX 4080 launches today and offers fantastic gaming performance at 4K, with highest details. In our RTX 4080 Founders Edition review, we confirm that NVIDIA's new $1200 card beats the last-generation flagship RTX 3090 Ti conclusively, both in performance and value.

www.techpowerup.com

 

[trinibwoy]

I feel relative to Nvidia's own stack in theory, as we don't actually know how the rest of the stack will play out for sure yet especially in terms of street prices, I don't feel the RTX 5080 is really a good buy either. Just compared against the 5070ti in theory it might only be 20% faster (could end up less) but you will pay 33% more MSRP to MSRP. It also has the same VRAM, so that would not be a difference either. The only real selling point I feel is it has a FE model while the 5070ti won't. We won't know for certain right now but at least speculatively I feel the 5070, 5070ti and 5090 are really the ones to consider depending on criteria.

Good point. The 5070 Ti looks to be the better buy. Hard to form any conclusions until the 5070 Ti is actually reviewed though. So maybe an advisory to wait until then before deciding whether to buy a 5080.

 

[trinibwoy]

Is 4xDSR + DLSS Perf better than using DLAA at native res? I feel like you are dabbling in the dark arts and it's making me nervous

Looks like I may be on to something No idea why this is the case but here you go.

“Well, believe it or not, DLSS Performance Mode at 8K can provide a better image quality than Native 4K with DLAA.”

Is 8K Gaming Possible With NVIDIA RTX 5090 & DLSS 4?

So, is 8K gaming now possible with NVIDIA's latest GPU, the NVIDIA GeForce RTX 5090, and DLSS 4? Let's find out.

www.dsogaming.com

 

[homerdog]

Looks like I may be on to something No idea why this is the case but here you go.

“Well, believe it or not, DLSS Performance Mode at 8K can provide a better image quality than Native 4K with DLAA.”

Is 8K Gaming Possible With NVIDIA RTX 5090 & DLSS 4?

So, is 8K gaming now possible with NVIDIA's latest GPU, the NVIDIA GeForce RTX 5090, and DLSS 4? Let's find out.

www.dsogaming.com

I believe it. In my experience, even with the same base number of pixels rendered, raising the DLSS output resolution increases image quality substantially. E.g. 1440p DLSS Performance looks better than 1080p DLSS Quality even though they are both 720p internally.

 

[trinibwoy]

You would think DLAA has more information to work with than DLSS. Maybe the upscaling process has some intelligence that’s not applied when running at native DLAA.

 

[Scott_Arm]

WIth upscaling like DLSS post-processing is done after the upscale, or at least that's how Nvidia recommends that its done. Post-process effects do not upscale well. Better to apply depth of field, motion blur, lens effects after the upscale. That could lead to a much better image on higher resolution screens even if the higher resolution screen is upscaling from the same base resolution, but it will also run worse.

Alan Wake 2 is a good example of this. There is a toggle to switch pre or post application of post-processing effects. The game looks much better and much sharper if you apply the post-processing after upscale, but there is a significant performance hit because that game uses a lot of very heavy post-processing effects.

 

[homerdog]

You would think DLAA has more information to work with than DLSS. Maybe the upscaling process has some intelligence that’s not applied when running at native DLAA.

Does DLSS jitter sample positions? IDK if it's always an ordered grid or if it selects sample positions in some other way that might relate to the output resolution.

 

[DegustatoR]

You would think DLAA has more information to work with than DLSS. Maybe the upscaling process has some intelligence that’s not applied when running at native DLAA.

Well you are using DSR to output 8K on a 4K screen. So DLSS P 8K is essentially DLAA (1/4th of 8K is 4K) + DSR 4x so it's not that surprising that the resulting quality is higher.

 

[trinibwoy]

Well you are using DSR to output 8K on a 4K screen. So DLSS P 8K is essentially DLAA (1/4th of 8K is 4K) + DSR 4x so it's not that surprising that the resulting quality is higher.

He’s saying it looks better than 4K DLAA.

 

[DegustatoR]

He’s saying it looks better than 4K DLAA.

Yes.
There are two ways to view that: either as a 4K DLAA vs 8K DLSS-P downscaled to 4K display via DSR 4x or as a 4K DLAA vs 8K DLSS-P on an 8K screen.
In both cases it's pretty obvious why DLSS-P would look better than 4K DLAA.

 

[trinibwoy]

Yes.
There are two ways to view that: either as a 4K DLAA vs 8K DLSS-P downscaled to 4K display via DSR 4x or as a 4K DLAA vs 8K DLSS-P on an 8K screen.
In both cases it's pretty obvious why DLSS-P would look better than 4K DLAA.

I assume it’s the former. Why is it obvious that 8K DLSS-P would look better than 4K DLAA on a 4K screen?

 

[DegustatoR]

Why is it obvious that 8K DLSS-P would look better than 4K DLAA on a 4K screen?

Because 4K DLAA and 8K DLSS-P are basically the same rendering resolution with similar DL TAA in that resolution but the latter also gain improvements from DSR 4x downsampling which are MIP LODs adjustments and the downsampling AA on top of the DL TAA. You're basically comparing 4K TAA with 8K reconstructed from 4K with DLSS and then downscaled back to 4K. The latter is doing a lot more work.

 

[trinibwoy]

Because 4K DLAA and 8K DLSS-P are basically the same rendering resolution with similar DL TAA in that resolution but the latter also gain improvements from DSR 4x downsampling which are MIP LODs adjustments and the downsampling AA on top of the DL TAA. You're basically comparing 4K TAA with 8K reconstructed from 4K with DLSS and then downscaled back to 4K. The latter is doing a lot more work.

LOD differences make sense.

 

[Florin]

And it arrived already! I'm surprised how well that paper launch worked out.
Ordered on the 30th of Jan right here in Singapore, and on my desk on Monday evening the 3rd.

It's hard to express in words just how big this box is. I placed it next to some choice pieces to give some indication.

I'm about to try me some 4K Stalker 2



- edit - shrunk img size a bit