NVIDIA GeForce RTX 5080 Founders Edition

[Visible Noise]

The small die strategy was the golden era for AMD GPUs. Evergreen hit a 49% market share with the HD 5000 series, the highest AMD/ATI ever recorded.

Evergreen was a large die (Terascale) Polaris was small die (GCN).

The HD 7000‘s were the rebrand cards. They could be Terrascale or GCN based upon their model number.

 

[lexluthermiester]

What a terrible idea.... lmao...

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

 

[gffermari]

What?
View attachment 382416
The 5080 is less than a stellar performance and isn't worth $1000, but half of a 5090 it is not. 2/3rds sure.
The 4070ti/3090ti are half a 5090. Let's keep things accurate shall we?

When you put them in heaviest situations out there, the 5090 is nearly double the fps.
In any other situation, 5090 is a victim of a component that causes bottleneck.

 

[lexluthermiester]

When you put them in heaviest situations out there, the 5090 is nearly double the fps.

That is a specific situation that does not reflect the over all experience which is why I showed the graph above. And even with your examples that's still not half. Sooo...

 

[AusWolf]

What?
View attachment 382416
The 5080 is less than a stellar performance and isn't worth $1000, but half of a 5090 it is not. 2/3rds sure.
The 4070ti/3090ti are half a 5090. Let's keep things accurate shall we?

I'm sure the general narrative is "oh look, it's 2/3s of a 5090 for half the price, what a good value this is", but my narrative is "oh look, it's a 4080 Super at the same price, how boring".

 

[gffermari]

That is a specific situation that does not reflect the over all experience which is why I showed the graph above. And even with your examples that's still not half. Sooo...

The overall experience is irrelative if you want to show the actual performance difference.
It's like the nonsense RT average graph where some games have only RT shadows and others Full RT.

 

[lexluthermiester]

I'm sure the general narrative is "oh look, it's 2/3s of a 5090 for half the price, what a good value this is", but my narrative is "oh look, it's a 4080 Super at the same price, how boring".

Either way, the 5080 is not a great performance. It shouldn't have been released as a 5080. However, my above point was that it is not 1/2 of a 5090. It's better than that. Not much better, but better.

The overall experience is irrelative if you want to show the actual performance difference.

That statement is a contradiction and makes no sense.

It's like the nonsense RT average graph where some games have only RT shadows and others Full RT.

Also nonsense.

 

[AusWolf]

Either way, the 5080 is not a great performance. It shouldn't have been released as a 5080. However, my above point was that it is not 1/2 of a 5090. It's better than that. Not much better, but better.

I agree. So far, the 50 series is a massive letdown. Only the 5090 is faster than the predecessor, but at a massively increased price. Sure, Nvidia didn't increase price on the rest of the product stack, but didn't increase performance, either. They didn't increase price simply because they couldn't, the cards aren't worth more. Stagnation at its worst. If AMD can't trump this with a good, or even just acceptable value offer on the 9070 XT, I won't know what to say.

 

[oxrufiioxo]

I agree. So far, the 50 series is a massive letdown. Only the 5090 is faster than the predecessor, but at a massively increased price. Sure, Nvidia didn't increase price on the rest of the product stack, but didn't increase performance, either. They didn't increase price simply because they couldn't, the cards aren't worth more. Stagnation at its worst. If AMD can't trump this with a good, or even just acceptable value offer on the 9070 XT, I won't know what to say.

For realz if they can't look appealing vs a 5070 they are doomed.

 

[lexluthermiester]

I agree. So far, the 50 series is a massive letdown.

Yeah, it's kinda shaping up that way. This is bad for NVidia, but good for Intel and AMD.

 

[Chrispy_]

What?
View attachment 382416
The 5080 is less than a stellar performance and isn't worth $1000, but half of a 5090 it is not. 2/3rds sure.
The 4070ti/3090ti are half a 5090. Let's keep things accurate shall we?

I said half a 5090, not half the performance of a 5090. Just like the 4080 12GB was about half the 4090. For the 50 series, the 5090 is cut down this much to make a 5080:

10752 cores vs 21760​

45.6bn transistors vs 92.2bn​

378mm die size vs 750mm​

256-bit bus vs 512-bit​

16GB VRAM vs 32GB​

HALF (well, technically a little less than half)

In fact, the only thing that's not physically halved is the ROP count.

 

[TheinsanegamerN]

Evergreen was a large die (Terascale) Polaris was small die (GCN).

Incorrect. Evergreen was AMD's small die strategy

AMD's Radeon HD 5870: Bringing About the Next Generation Of GPUs

www.anandtech.com

More details on the RV800: Evergreen is a big coup for AMD

AMD's released its new GPU, the RV800, which brings massive performance …

arstechnica.com

"This dynamic is another step in the continuing vindication of AMD's small-die strategy; build a midrange part on a cutting-edge process, and build two GPUs into one card for higher performance ala the 3870X2, 4850X2, 4870X2, upcoming 5870X2, and NVIDIA's GTX295 and 7950GX2 parts."

AMD's 5870, a 334mm2 part, was outperforming nvidia's 529mm2 GTX 480.

Meanwhile, with polaris, the RX 480 at 232mm2 was roughly 4% slower then the 6GB 200mm2 GTX 1060. GCN, as an arch, was not small die focused (hello RX 290x!) the way the terrascale arch was.

The HD 7000‘s were the rebrand cards. They could be Terrascale or GCN based upon their model number.

Mostly incorrect. Evergreen 2 (hd 6000 series) were, quite literally, the HD 5000s. The 6870 was a 5870, the 6750 was a 5750, ece. You could crossfire them together with 0 effort. Then when the GTX 570 and 580 came out, AMD was caught with their pants down and had to rush out the 6900 series in response, as nvidia fixed fermi's flaws and delivered significant improvement. This resulted in the 6900 series being VILV 4 instead of VILV 5.

None of the HD 7000 series desktop parts were terrascale. The HD 7700, 7800, and 7900 series were all GCN. Only the low end 73-76xx mobile parts were terrascale, the midrange 7700m and up were all GCN. This was normal at the time, all the 520-540 series from nvidia were previous gen as well. The HD 7750/GTX550ti were considered the baseline as entry level cards, anything below that was regarded as a display adapter or hobbyist item, not a serious card.

 

[sn2x]

Two things I like from this card are the efficiency and the overclocking headroom. I might get one since its the most powerful GPU that can work with my PSU.

I wonder if the 10%+ overclocking headroom is normal though, seems like a lot of performance to leave on the table... with a 10% OC this card should be within 5% of the 4090.

 

[oxrufiioxo]

Two things I like from this card are the efficiency and the overclocking headroom. I might get one since its the most powerful GPU that can work with my PSU.

I wonder if the 10%+ overclocking headroom is normal though, seems like a lot of performance to leave on the table... with a 10% OC this card should be within 5% of the 4090.

I seen another review where it was ten also but they would really need to run a full 20-30 games to see what real gains it has.... Synthetics a lot of time scale better than games do.

 

[MxPhenom 216]

He stated the 5080 beats the 4090. It does not.

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

Not even with framegen does it beat the 4090. Jensen very likely didn't see for himself the side-by-side perf comparisons(and seriously that's part of his job). Instead he likely just read from the speech the marketing nitwits made for him. So yes, Jensen ends up looking like a liar and that's not a good look.

Well when the 4090 isnt running with frame gen but if the 5080 is, obviously it'll put out higher frames. (Same claim they made with 5070 matching 4090 performance - fine print on that is with FG enabled on 5070 and 4090 doesnt)

Id have to go back and see, but I don't recall him saying pure rasterization of a 5080 is better than 4090. And many were able to extrapolate from the plots posted after the keynote that the performance uplift on the 5080 was going to be 10-20% more than a 4080 which would conclude it was still not better than 4090.

 

[AusWolf]

Well when the 4090 isnt running with frame gen the 5080 will obviously put up more frames with frame gen. (Same claim they made with 5070 matching 4090 performance - fine print on that is with FG)

Not that such information is of any use, though. I could say my 6750 XT puts out more frames at 720p low with FSR than the 5080 does at native 4K ultra, but what's the point?

 

[Visible Noise]

Incorrect. Evergreen was AMD's small die strategy

AMD's Radeon HD 5870: Bringing About the Next Generation Of GPUs

www.anandtech.com

More details on the RV800: Evergreen is a big coup for AMD

AMD's released its new GPU, the RV800, which brings massive performance …

arstechnica.com

"This dynamic is another step in the continuing vindication of AMD's small-die strategy; build a midrange part on a cutting-edge process, and build two GPUs into one card for higher performance ala the 3870X2, 4850X2, 4870X2, upcoming 5870X2, and NVIDIA's GTX295 and 7950GX2 parts."

AMD's 5870, a 334mm2 part, was outperforming nvidia's 529mm2 GTX 480.

Meanwhile, with polaris, the RX 480 at 232mm2 was roughly 4% slower then the 6GB 200mm2 GTX 1060. GCN, as an arch, was not small die focused (hello RX 290x!) the way the terrascale arch was.

Mostly incorrect. Evergreen 2 (hd 6000 series) were, quite literally, the HD 5000s. The 6870 was a 5870, the 6750 was a 5750, ece. You could crossfire them together with 0 effort. Then when the GTX 570 and 580 came out, AMD was caught with their pants down and had to rush out the 6900 series in response, as nvidia fixed fermi's flaws and delivered significant improvement. This resulted in the 6900 series being VILV 4 instead of VILV 5.

None of the HD 7000 series desktop parts were terrascale. The HD 7700, 7800, and 7900 series were all GCN. Only the low end 73-76xx mobile parts were terrascale, the midrange 7700m and up were all GCN. This was normal at the time, all the 520-540 series from nvidia were previous gen as well. The HD 7750/GTX550ti were considered the baseline as entry level cards, anything below that was regarded as a display adapter or hobbyist item, not a serious card.

Huh. Somebody needs to update Wikipedia then.

 

[MxPhenom 216]

Not that such information is of any use, though. I could say my 6750 XT puts out more frames at 720p low with FSR than the 5080 does at native 4K ultra, but what's the point?

I know, its f***** stupid. I despise this frame generation shit because of the misleading marketting they can spew out. It also is holding back rastorization r&d, and I think pre-maturely. I dont think we are too the point yet where we cannot get more out or a traditional rendering pipeline, but all money is going into AI, so here we are. Cards for said software.

We will be in for some fun when we move on from FINFET to GAAFETs though. Believe it!

 

[EsliteMoby]

I said half a 5090, not half the performance of a 5090. Just like the 4080 12GB was about half the 4090. For the 50 series, the 5090 is cut down this much to make a 5080:

10752 cores vs 21760​

45.6bn transistors vs 92.2bn​

378mm die size vs 750mm​

256-bit bus vs 512-bit​

16GB VRAM vs 32GB​

HALF (well, technically a little less than half)

In fact, the only thing that's not physically halved is the ROP count.

Could that be L2 cache size not doubled hence the bottleneck?

Or someone could test in native 8K maybe the gap between 5080 and 5090 will widen.

 

[Baba]

3GB GDDR7 IC's dont exist yet. 2GB only right now. 3GB will probably be available by the end of the year

Samsung Develops Industry’s First 24Gb GDDR7 DRAM for Next-Generation AI Computing

New GDDR7 offers industry-leading capacity and speed of over 40Gbps, significantly raising the bar for graphics DRAM powering future applications Validation with major GPU customers to begin this year, production slated for early next year

news.samsung.com

Validation for the 24Gb GDDR7 in next-generation AI computing systems from major GPU customers will begin this year, with plans for commercialization early next year. - posted October 17, 2024

 

[dyonoctis]

Incorrect. Evergreen was AMD's small die strategy

AMD's Radeon HD 5870: Bringing About the Next Generation Of GPUs

www.anandtech.com

More details on the RV800: Evergreen is a big coup for AMD

AMD's released its new GPU, the RV800, which brings massive performance …

arstechnica.com

"This dynamic is another step in the continuing vindication of AMD's small-die strategy; build a midrange part on a cutting-edge process, and build two GPUs into one card for higher performance ala the 3870X2, 4850X2, 4870X2, upcoming 5870X2, and NVIDIA's GTX295 and 7950GX2 parts."

AMD's 5870, a 334mm2 part, was outperforming nvidia's 529mm2 GTX 480.

Meanwhile, with polaris, the RX 480 at 232mm2 was roughly 4% slower then the 6GB 200mm2 GTX 1060. GCN, as an arch, was not small die focused (hello RX 290x!) the way the terrascale arch was.

Mostly incorrect. Evergreen 2 (hd 6000 series) were, quite literally, the HD 5000s. The 6870 was a 5870, the 6750 was a 5750, ece. You could crossfire them together with 0 effort. Then when the GTX 570 and 580 came out, AMD was caught with their pants down and had to rush out the 6900 series in response, as nvidia fixed fermi's flaws and delivered significant improvement. This resulted in the 6900 series being VILV 4 instead of VILV 5.

I also remember that a lot of people were confused about the naming change: people stacked the 6870 against the 5870 and were mad to see a slight performance regression overall when that GPU was replacing the 5770.
What’s In a Name? - AMD’s Radeon HD 6870 & 6850: Renewing Competition in the Mid-Range Market

When it came to light that Barts, oversimplistically a successor to the Juniper GPU (which makes up the Radeon HD 5700 series), is going to be branded under the HD 6850/6870 series, it created quite some drama; with some users claiming it to be very gimmicky of AMD to release a series that isn't much of an upgrade option for existing users of HD 5800 series GPUs. That's not the case, because AMD made it adequately public through the press, its reasoning behind using the HD 6800 series as the "gamer's sweet spot" series, and consolidating all higher-end SKUs into the Radeon HD 6900 series, slated for next month. Besides, it's not like AMD is asking Radeon HD 5800 series kind of prices for the cards released today. The main design ideology behind the HD 6870, as AMD put it, is to give you Radeon HD 5800 series performance at sweet-spot prices.

We don’t have a problem with AMD introducing the 6 series here – the changes they’ve made, even if not extreme, at least justify that. But there’s a very real issue of creating confusion for buyers of the 5800 series now by introducing the 6800 series. The performance may be close and the power consumption lower, but make no mistake, the 5800 series was faster.

Ultimately this is not our problem; this is AMD’s problem. So we can’t claim harm per-say, but we can reflect on matters. The Barts cards being introduced today should have been called the 6700 series. It would have made the latest rendition of the 700 series more expensive than last time, but at the same time Barts is a very worthy upgrade to the 5700 series. But then that’s the problem for AMD; they don’t want to hurt sales of the 5700 series while it’s still on the market.

 

[mkppo]

Huh. Somebody needs to update Wikipedia then.

Not sure which wiki link you're looking at but this wiki page has a nice table outlining all the architectures and instruction sets.

Evergreen was a large die (Terascale) Polaris was small die (GCN).

The HD 7000‘s were the rebrand cards. They could be Terrascale or GCN based upon their model number.

R700 was when they started the small die strategy with the 4850/4870/4870X2. Evergreen was the successor and continued the same strategy but with larger dies (256mm > 334mm) with the 5850/5870/5970. Then came nothern islands which was VLIW4 for only the 69xx but the other models were rebrands of Evergreen. There was also a naming change, with 5870 effectively being replaced with 6970. All of these architectures used the Terascale instruction set with slight modifications depending on the iteration. They were all very successful too, especially the 5xxx generation.

Then Southern Islands came along with the HD 7xxx's and all used the GCN instruction set for almost the next decade. Heyday of GCN was the 7000 and 290x launch times when they were much faster and had a better/more forward looking arch than nv at the time. Polaris was released after a few years when GCN was long in the tooth and was a small die priced cheaply to tide them over till RDNA. They did release two more flagships though with HBM of which the last GCN, Radeon VII, was the pick of the HBM bunch and a swansong to the first arch AMD designed after taking over from ATI.

 

[QUANTUMPHYSICS]

Buying a 5090 is the only way I can buy without looking at benchmarks and KNOW that I've gotten the best of the best.

5080 gets 200 fps in 4K on Counterstrike 2 and 70 fps in 4K on Cyberpunk. I was expecting it to blow Cyberpunk away but I guess 4K at 120fps isn't happening till the 8080 for that game.

I don't play many of these new crappy games. I typically just play Battlefield 2042, Chess and Counterstrike.

Microcenters here in NYC claim they each have just 4 of the 5090 but more than 50 of the 5080.

Looks like most people will be walking away with 5080.

 

[Hyderz]

I got a question…
How likely are we to see 50 series super cards coming next year?

 

[Baba]

I got a question…
How likely are we to see 50 series super cards coming next year?

100%. Look at the pattern of previous generations. Nvidia loves the mid-cycle refresh. Also, 24Gb/3GB dies are almost ready. Imagine 24GB 5080 Supers for the same price. People will flock to them. There has to be some kind of core increase as well to bump up the performance. It's almost as if Nvidia releases meh cards initially only to cause a stir later with the Supers. Similarly to the 4080 Super release.

Microcenter just updated their site with some 5080 models. It's not terrible.

Zotac NVIDIA GeForce RTX 5080 GAMING SOLID Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,149.99


MSI NVIDIA GeForce RTX 5080 VENTUS 3X PLUS Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,169.99


Gigabyte NVIDIA GeForce RTX 5080 GAMING Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,199.99


ASUS NVIDIA GeForce RTX 5080 Prime Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,199.99


MSI NVIDIA GeForce RTX 5080 GAMING TRIO Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,229.99


MSI NVIDIA GeForce RTX 5080 SUPRIM SOC Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,279.99


ASUS NVIDIA GeForce RTX 5080 TUF Gaming Overclocked Triple Fan 16GB GDDR7 PCIe 5.0 Graphics Card
Available on Jan 30,2025
Our price $1,349.99