AMD RX 9090 XT and 9080 XT rumors

[vjhawk]

If the 9090 XT can trade blows with the RTX 4090 and the 9080 XT with the RTX 5080, AMD could make some serious inroads depending on the product pricing and availability.

I'm thinking the 9090 should have 32GB and the 9080 24GB to maintain their ram advantage over Nvidia's offering to give an attractive product for customers.

Moore's law video on the leaks (starts at 15:11):

View: https://youtu.be/vgzBNzCw0Rg?t=920

 

[_mockingbird]

Moore's Law Is Dead = snake oil salesman

 

[learners permit]

His imagination is running off at the mouth again but he's not the most annoying tech parrot.

 

[vjhawk]

Let's say the 9090 XT with 32GB does not come out.

Even a 9080 XT with 24 GB of memory and 25% faster than 9070XT performance that trades blows with the 5080 would be welcome if it came out at a reasonable price.

Maybe $799 MSRP for the AMD card and I think a lot of people would buy it.
We know MSRP would probably sell out really quick and the non-MSRP variants will probably rise closer to 1K+.
But when a 5080 on Newegg costs $1399 to 2049 (For the Astral), well those prices are sane in comparison to 5000 series prices.

 

[noko]

Let's say the 9090 XT with 32GB does not come out.

Even a 9080 XT with 24 GB of memory and 25% faster than 9070XT performance that trades blows with the 5080 would be welcome if it came out at a reasonable price.

Maybe $799 MSRP for the AMD card and I think a lot of people would buy it.
We know MSRP would probably sell out really quick and the non-MSRP variants will probably rise closer to 1K+.
But when a 5080 on Newegg costs $1399 to 2049 (For the Astral), well those prices are sane in comparison to 5000 series prices.

Even at $999 solid, if it beats the 5080 and Supra, would make Nvidia look anemic. Probably sell like high end hotcakes.

 

[crazycrave]

Please don't tell Moore's Law about my Intel Arc B570 10GB card as it was never released and a canceled product.

 

[KazeoHin]

Even at $999 solid, if it beats the 5080 and Supra, would make Nvidia look anemic. Probably sell like high end hotcakes.

But could AMD supply them enough to keep the price reasonable?

 

[noko]

But could AMD supply them enough to keep the price reasonable?

Doubt it, but one can be surprised once in a great while.

 

[arthur_tuxedo]

It’s impressive how much performance they are wringing out of the humble mid-range Navi 48. This is what happens when you get handed a golden opportunity to compete at the high-end, but you already forfeited the match. AMD lives, doesn’t learn, doesn’t get Luvs.

 

[LukeTbk]

It’s impressive how much performance they are wringing out of the humble mid-range Navi 48

It is, but is it much different than what they have been doing with RDNA 2... 3... why the opportunity more golden than before ?

AMD has the superior node, the 9060xt is a bigger die (GB206 at 181mm, 199mm for Navi 44) Navi pushing more power (188w vs 166w)

357mm of superior tsmc node Navi 48 does not beat an 83% enabled 378mm GB203 (5070ti) when pushed at 315w, why should the 9070xt not be close to the 5070ti ?

They are not that far behind, true, but isn't that always the case ? Lovelace felt way ahead of RDNA 3 in comparison, but Nvidia had the better node that time, now it is AMD (like for RDNA 2), so they are closer again.

A lot of scenario, say if AMD would have done X-Y-Z, seem to assume they are able to keep it secret long enough of Nvidia and/or that Nvidia has zero reaction/adjustment to those move, that the bigger navi 52 does not mean a bigger GB203 or 384 bits even more cut down GB202 gaming GPU does not exist in that world, that the 5060/5070/5080 super release schedule is not super aggressive if they are really caught by surprise.

 

[arthur_tuxedo]

It is, but is it much different than what they have been doing with RDNA 2... 3... why the opportunity more golden than before ?

AMD has the superior node, the 9060xt is a bigger die (GB206 at 181mm, 199mm for Navi 44) Navi pushing more power (188w vs 166w)

View attachment 735293

357mm of superior tsmc node Navi 48 does not beat an 83% enabled 378mm GB203 (5070ti) when pushed at 315w, why should the 9070xt not be close to the 5070ti ?

They are not that far behind, true, but isn't that always the case ? Lovelace felt way ahead of RDNA 3 in comparison, but Nvidia had the better node that time, now it is AMD (like for RDNA 2), so they are closer again.

A lot of scenario, say if AMD would have done X-Y-Z, seem to assume they are able to keep it secret long enough of Nvidia and/or that Nvidia has zero reaction/adjustment to those move, that the bigger navi 52 does not mean a bigger GB203 or 384 bits even more cut down GB202 gaming GPU does not exist in that world, that the 5060/5070/5080 super release schedule is not super aggressive if they are really caught by surprise.

Maybe NV would have reacted with a stronger lineup at the top-end, but I feel like they already had enough pressure to make Blackwell a legitimate upgrade from Lovelace, and they didn’t do it. I also don’t know what else NV could have done, given that the 5090 is already a 600 Watt card. A 450 Watt RDNA 4 card with double the CUs clocked somewhere around 2.7-2.9 GHz would have performed well above the 5080, like +50% or more, comfortaby above the 4090, and might have come very close to the 5090. At 600 Watts, it might have matched or beaten the 5090. AMD could have charged, say, $1,800 for the 600W, $1,200 for the 450W, and $900 to $1,000 for a cut-down model if needed. Assuming that NV will out-maneuver them and not even showing up to the fight is the kind of mentality that got Radeon into its current lowly position, IMO.

 

[noko]

Maybe NV would have reacted with a stronger lineup at the top-end, but I feel like they already had enough pressure to make Blackwell a legitimate upgrade from Lovelace, and they didn’t do it. I also don’t know what else NV could have done, given that the 5090 is already a 600 Watt card. A 450 Watt RDNA 4 card with double the CUs clocked somewhere around 2.7-2.9 GHz would have performed well above the 5080, like +50% or more, comfortaby above the 4090, and might have come very close to the 5090. At 600 Watts, it might have matched or beaten the 5090. AMD could have charged, say, $1,800 for the 600W, $1,200 for the 450W, and $900 to $1,000 for a cut-down model if needed. Assuming that NV will out-maneuver them and not even showing up to the fight is the kind of mentality that got Radeon into its current lowly position, IMO.

More like letting others lead them by their nose.

 

[LukeTbk]

I also don’t know what else NV could have done,

Nvidia has better performance per die size and per watt, despite a cheaper-weaker node being used, at any die size / power Blackwell should have beaten Rdna 4 no ? Why the smaller 5060ti beat the 9060xt with less watt but it would change at the higher end ?

The 5080 could easily have been bigger if needed, the 5070 is even smaller than the 4070 and the 4070 was quite small.

I think they worked extremely hard to make those navi-48/44 card, FSR 4, etc...and did show up at the fight, they were a bit shock by Nvidia pricing, new DLSS transformer, enough to cancel their announcement.

I am not saying AMD should assume Nvidia will out-maneuver everything they do, I am more saying when we monday morning quaterback what AMD do we need to not only take into account how much time there is between a decision and a product release but also that for any AMD decision we think they should have done, take into account the probable Nvidia response.

I do not believe the narrative Nvidia do not care about AMD talking point, ego alone.... let just say it is not surprising they had a 3090ti when AMD could make a 6950xt that will challenge the 3090 but no 4090ti when that was not the case...

 

[arthur_tuxedo]

Nvidia has better performance per die size and per watt, despite a cheaper-weaker node being used, at any die size / power Blackwell should have beaten Rdna 4 no ? Why the smaller 5060ti beat the 9060xt with less watt but it would change at the higher end ?

The 5080 could easily have been bigger if needed, the 5070 is even smaller than the 4070 and the 4070 was quite small.

I think they worked extremely hard to make those navi-48/44 card, FSR 4, etc...and did show up at the fight, they were a bit shock by Nvidia pricing, new DLSS transformer, enough to cancel their announcement.

I am not saying AMD should assume Nvidia will out-maneuver everything they do, I am more saying when we monday morning quaterback what AMD do we need to not only take into account how much time there is between a decision and a product release but also that for any AMD decision we think they should have done, take into account the probably Nvidia response.

I do not believe the narrative Nvidia do not care about AMD talking point, ego alone.... let just say it is not surprising they had a 3090ti when AMD could make a 6950xt that will challenge the 3090 but no 4090ti when that was not the case...

You might be right, and I might not be accounting for scaling issues that AMD would have run into. I was assuming 9070 XT +80% for a hypothetical double-CU model, and that might not have been possible. But even if the 5090 was still comfortably ahead and even if NV had made a 5080 worthy of the name (they should have done that anyway just to entice upgrades and avoid criticism), I don’t see how it would have hurt AMD to show up to that fight.

AMD’s panic at Blackwell’s announcement is proof-positive that they’re marching to the beat of NV’s drum. AMD’s leadership were the only ones gullible enough to take any of the fake frame comparisons seriously. Once Blackwell got into reviewer hands, they promptly tore NV’s claims apart and the company suffered serious reputational damage, damage that AMD couldn’t properly capitalize on because they were desperately trying to “me too” every mistake NV made, touting their own frame-gen instead of calling out the dishonesty. They were so busy copying NV’s mistakes that they confidently stomped their foot into the same 8 GB sh*t pile that they just saw Jensen step in. The Radeon group has some of the best engineers in the industry, but their “leadership” lets them down time and time again.

 

[LukeTbk]

AMD’s leadership were the only ones gullible enough to take any of the fake frame comparisons seriously.

that really not what was took seriously or made them panic, Transformer model maybe, more so the price announced, that surprised them, the only thing that they would have mostly known before that day, they knew about Frame generation and knew it would be used like for Lovelace for the presentation. $550 for the 5070, that was the surprise and made it uneasy to announce a $650 9070 on stage right after.

and I might not be accounting for scaling issues that AMD would have run into. I was assuming 9070 XT +80% for a hypothetical double-CU model,

I am not sure they need scaling issues either, the 9070xt is about 70% faster than the 9060xt with double the cu(80% at 4k), so it scale well, making it work money wise could be more the hard part, the lesser you sale the higher the total added R&D/units sold cost on each card get, a $100 millions design and tape out, if you sales 1 million card, not so bad $100 per gpu, if you sales only 200,000 that become $500 per gpu before the first manufacturing, marketing, driver support and what not expense and hard to make it work, they do not have known in advance X40-RTX 6000 using that same die sales to fall back on and make it work either.

I don’t see how it would have hurt AMD to show up to that fight.

Outside the very possible just spend more than make money back on the 640mm big 128 CU Navi, there is maybe some truth in AMD talk of a very simple generation helping at keeping cost down, would there Navi44-48 be as good if they had to work on a big Navi52 ? There always limited resource and a opportunity cost that goes into any form of work. Would trying to go all-in while still not having the fastest GPU at the end of it a good use of resource, a 5080 class competitor would have been maybe a better one to do.

There is something, a big downside for AMD to not have a gpu in early 2026 (if that the case, but could very well be the case) that is not faster than a 2022 RTX 4080, 4 years later, a not fully enabled 380mm on a 2020 tmsc node (which some said at the time should be called a 4070, those people are saying that the best AMD gpu 4 years later do not match an Nvidia 4070 of 2022 and are not even that cheap or will start to rethink that stance...).

But the most important is to fully compete in that $250-$700 window marketplace and in that performance range and be really good in that power envelope and target performance needed that will drive the PlayStation 6, in many regards RDNA 4 is a success imo as of now, 9060 non xt, 9070 gre to complete the stack need to be fast enough, volume to stay good, FSR 4 to spread fast enough.... RDNA 4 made it so that RT performance gap is not a talking point anymore (in a TPU review, 9060xt vs 5060ti performance difference, RT on, RT off it is the same, giant gap in say Black Myth are in scenario that are unplayable on both anyway)

 

[evhvis]

It doesn't make sense to release a 9080 XT this late IMO. Even if it is in the cards, it would be unlikely that it launched before September and a lot of the potential 9080 XT customers have most likely already bought GPUs already as AMD didn't give any indication that they would compete at the high-end. I would much rather see them focus on the next gen and improving PT performance as they are quite far behind on that and a bit behind on their software stack (FSR). The improvements to DLSS 4 TF made PT a reality on 4080/5070 ti and up from Nvidia as you can upscale from much lower res and get good results. The 5070 ti with DLSS 4 and Ray Reconstruction can run PT at DLSS Quality at 1440p in Alan Wake 2 and CP 2077 with framerates around 60. The PT performance along with superior upscaling and RT Denoiser was why I chose 5070 ti this gen. AMD needs parity on those type of features to be equal value propositions in the upper mid range and high end market.

I run a 7900 GRE in my secondary as it was great value when I bought it (closer to 4060 ti price than 4070 price), but they haven't released a GPU in almost a decade that I would consider putting in my main due to lack of performance in cutting edge features. I would love to see AMD come back to the ATI 9700 pro days where they are actually pushing performance and features, not just trying to be a value proposition. The GPU market is stagnating due to AMD being content to release inferior GPUs and just price them a little lower to be good value, but it does nothing for the competition in the high end market, which allows Nvidia to release mediocre gen-to-gen upgrades as there is no alternative. Pretty sure the 5080 was supposed to be the 5070 ti and the 5070 ti would have been the 5070 at different price points if AMD had actually gone for it this gen, but AMD going for value options allowed Nvidia to move product numbers and increase profit margin on the products.

 

[yinzerniner]

. The GPU market is stagnating due to AMD being content to release inferior GPUs and just price them a little lower to be good value, but it does nothing for the competition in the high end market, which allows Nvidia to release mediocre gen-to-gen upgrades as there is no alternative. Pretty sure the 5080 was supposed to be the 5070 ti and the 5070 ti would have been the 5070 at different price points if AMD had actually gone for it this gen, but AMD going for value options allowed Nvidia to move product numbers and increase profit margin on the products.

is that really true, or did Nvidia cap their gaming GPU performance since they can make 5x the profit on that silicon from AI sales?

Also Nvidia most likely knew that AMD made big strides with FSR4 and ray tracing, but thought team red would do what they always did, similar performance minus $50. AMD even surprised their retail partners with the suggested MSRPs of the 9070 series, and did rebates for the stock already purchased. That gave AMD a huge pubic opinion push which they’ve continues with the 9060xt launch

If AMD decide to launch a 9080 or 9090 series it would probably make sense to do it only after the Super series cards come out. That way they can see the reaction and act accordingly like what they unknowingly did for the 9070, and purposely did for the 9060

 

[KazeoHin]

AMD's goal is NOT to compete with Nvidia. That's like saying a tick's goal is to compete with it's host.

AMD's goal is to have Nvidia set the pricing expectations, and AMD makes a minimum-viable-product to pick up the occasional customer without competing too much so they don't trigger Nvidia to lower prices.

If Nvidia wants to charge 300+ for an 8GB entry-level card, AMD can make something that barely qualifies as an alternative and charge $240 for it. They make WAY more money on that then they would if they actually tried to earn market share.

 

[vjhawk]

It's AMD's mentality that is letting them down.

They are clearly not aggressive enough in a market where NV is not creating enough supply to meet demand. And what supply they do give is insanely bad bang for the buck.

AMD decided to forfeit the high end to NV. OK, reasonable.

Then why aren't they going more aggressively to capture the low to mid tier market?

Their 9060XT 16GB embarrasses Nvidia's 5060 8GB, Beating it at Ray Tracing in 1080P by an average of 42%.

But even so they are unlikely to sell enough units to capture market share.

They have the 2nd place mentality and despite producing technically good products, just don't know how to capitalize on opportunities like Nvidia's FLOPPED 5000 series release.

The majority of sheep-like mainstream customers and pre-built PC makers will still use Nvidia card although it is inferior this generation.

If NVidia fixes their 6000 series to be anything approaching acceptable performance/dollar value, this abomination of a 5000 series paper launch will end up forgotten as a blip on the radar as if it never happened and Nvidia will still own 90% of the market despite screwing their customers and AIB partners with this current generation.

 

[Gideon]

I don't know what you guys are going on about, but AMD is priced fine against Nvidia at this point. Most of their cards are selling above retail and that usually indicates more demand than supply. They only have so much manufacturing capability and that falls short of being able to supply the entire gaming market. So Nvidia knows they don't have to try to hard and their stuff sells, because AMD is capacity constrained.

 

[Mr. Bluntman]

Tom has been going ham on the 9090 XT stuff lately on MLID to the point where it's since gotten annoying and I just skip his raging RDNA4 boner segments. It's just a balls-out overclocked 9070 XT with a ridiculous TDP and more VRAM; nothing revolutionary.

 

[wicktr0n_]

With every generation and every card, we always seem to run into the "if" statements with AMD.

I hope they eventually get there with UDNA.

 

[yinzerniner]

Tom has been going ham on the 9090 XT stuff lately on MLID to the point where it's since gotten annoying and I just skip his raging RDNA4 boner segments. It's just a balls-out overclocked 9070 XT with a ridiculous TDP and more VRAM; nothing revolutionary.

Yeah his constant "insider sources" bs without verifiable information is tiresome; definition of throwing crap at the wall to see what sticks.

Not saying that a 9080/9090xt wouldn't be fun but it totally depends on execution. AMD have proven they can get some good PR by finally playing the pricing game correctly vs Nvidia but the top end is a hugely different beast. Also, once you get above the $600 level a lot of these cards are used for productivity tasks and AMD is still trounced by Nvidia for these. Hopefully AMD can continue to spread their optimization teams to app developers like they've done for game studios and the adoption of FSR which has sped up in recent months.

 

[LukeTbk]

Not saying that a 9080/9090xt wouldn't be fun but it totally depends on execution. AMD have proven they can get some good PR by finally playing the pricing game correctly

They did burn it a bit on the initial launch too, the 9060xt series reception was quite timid despite the really good offer, because no one believed the $350 price point...

While I cannot go on newegg or bestbuy to buy one (so they were "right"), they seem easy to find at $370-380 and compared to the value of what was selling in the market until that launch:
https://www.newegg.com/d/Best-Selle...XIeq1NOQmmJDoT-1e0fLq2WUGM1JTudaEqzZ6zmILZpLU

($300 3060.... still selling, $360 7600xt 16G, $430 5060tTI 16G), timid reaction for what those gpu were imo...

 

[The Cobra]

I thought the whole point of RDNA 4 was to be a stepping stone to big RDNA 5 chip TO theoretcally compete with the 6xxx series from Nvidia.

 

[LukeTbk]

I thought the whole point of RDNA 4 was to be a stepping stone to big RDNA 5 chip TO theoretcally compete with the 6xxx series from Nvidia.

There is many narrative

) Quick filling step toward a new UDNA (cdna+RDNA combined arch) for what it mean
) A cost cutting operation (I am not exactly sure how much they save and how cheaper they can be because there is not a 500mm version of them, but that was pushed a lot, tape out few sku, work only on 2 chips to save cost)

Never heard the stepping stone to a big RDNA 5 one.

5000 series paper launch

Paper ? Biggest selling launch of any generation of Nvidia after the first x months, 19 available 5070s-5060s available in stock to ship at bestbuy.com right now, they sold many millions of 5000 series cards.

I am not sure about the premise that the 9060xt 16GB at $370 street price is not aggressive enough, feel like a clear winner at that price range

 

[noko]

$369 Asrock 9060XT 16GB, 6gb models at Newegg are available. Supply looks good and some $20 discounts on the 8gb models at $299, lol. I hope the 8gb models don't sell at all unless less than $250, more like $200. All AMD has to do is keep the price down and supply up and they will have a winner here.

 

[evhvis]

I thought the whole point of RDNA 4 was to be a stepping stone to big RDNA 5 chip TO theoretcally compete with the 6xxx series from Nvidia.

The rumors were that it was to be a stopgap until they could unify the architectures. AMD didn't expect Nvidia to come out with such a weak generation though, otherwise they probably would have made a 9080 XT as well. The official version is that they needed to increase market share and would rather compete at the low and mid-range where most cards are sold.

RDNA's weakness is that gaming has moved on from pure raster and shaders and more towards compute power. RDNA 4 is a step in the right direction, but they need more of the compute power that they spend lots of money researching for CDNA so it makes sense to unify the architectures. The move to RDNA was done because AMD had tons of compute power, but couldn't utilize it properly. With stuff like FSR/DLSS etc. the compute power is needed and more specifically AI acceleration which aligns with professional cards.

 

[Gideon]

Paper ? Biggest selling launch of any generation of Nvidia after the first x months, 19 available 5070s-5060s available in stock to ship at bestbuy.com right now, they sold many millions of 5000 series cards.

I am sorry since when did the 5000 series outsell the 1000 series? 5000 series has been pretty much a paper launch, very little stock coming in to any store. I think the 1080 and 1080ti alone have outsold all of the 5000 series combined. Heck the 1080 probably alone surpassed this poor launch by Nvidia.

 

[ZeroBarrier]

AMD didn't expect Nvidia to come out with such a weak generation though, otherwise they probably would have made a 9080 XT as well.

At this point it just feels like Nvidia is somewhat partially sandbagging their consumer GPUs just to keep Radeon in the race.

 

[yinzerniner]

Paper ? Biggest selling launch of any generation of Nvidia after the first x months, 19 available 5070s-5060s available in stock to ship at bestbuy.com right now, they sold many millions of 5000 series cards.

That's kind of true, but again the truth of the matter is somewhere hidden in the stats.

Team green said that the 50 series sold way more than 40 in first couple months, but that's probably because 50 series had twice as many cards launched in the same time period - 50 had 90, 80, 70 ti and 70 while 40 only had 90 and 80. So it had twice as many skus, and also midrange cards, so just doubling the sales figures is the bare minimum they should have bragged about.

It wasn't a "paper launch" but the available cards were severely limited in stock, which has gotten a lot better in the last month or so. AMD is still having 9070 supply issues, but it mightve been because they're obviously very occupied with the 9060xt right now.

But yeah, the 9060 xt 16gb, while the 8gb is a dog. Not quite as lowly as the 5060 though

 

[KazeoHin]

I am sorry since when did the 5000 series outsell the 1000 series? 5000 series has been pretty much a paper launch, very little stock coming in to any store. I think the 1080 and 1080ti alone have outsold all of the 5000 series combined. Heck the 1080 probably alone surpassed this poor launch by Nvidia.

Not even close.

PC Gaming is close to 3x the market size now than during the launch of the 10-series. Which means Nvidia can manufacture 2.5X as many and still be sold out.

The 5000 series is a dud of a launch and Nvidia is purposefully not meeting demand in order to keep profit margins as wide as possible, but that does not mean they aren't selling millions upon millions of units.

 

[Gideon]

Not even close.

PC Gaming is close to 3x the market size now than during the launch of the 10-series. Which means Nvidia can manufacture 2.5X as many and still be sold out.

The 5000 series is a dud of a launch and Nvidia is purposefully not meeting demand in order to keep profit margins as wide as possible, but that does not mean they aren't selling millions upon millions of units.

It's not, 2017 was peak graphic card sales, all it's done is dropped from there. Wish I could find a more updated one, but this is what I could find quickly at work. However I do agree with you that Nvidia is withholding supply, as they would rather make AI cards.

 

[LukeTbk]

I am sorry since when did the 5000 series outsell the 1000 series? 5000 series has been pretty much a paper launch, very little stock coming in to any store. I think the 1080 and 1080ti alone have outsold all of the 5000 series combined. Heck the 1080 probably alone surpassed this poor launch by Nvidia.

that directly from Nvidia last investor call (and they said it to journalist as well before), the statement do not say it outsold the 1000 series during all Pascal lifetime, but during their respective first X days of launch. Pascal was pre first crypto, it got really popular a bit late in its cycle.

There is more than twice as many steam active account in 2025 than 2016:

You have a much bigger profesionnal/ai market now

Why little stock coming in any store ? you can buy a 5070 directly from bestbuy.com ... msrp (https://www.bestbuy.com/site/pny-ge...with-triple-fan-black/6620352.p?skuId=6620352)

It is many things, paper launch it is certainly not, Nvidia made its biggest quarter gaming revenues wise of its history during blackwell launch, what source is there that there is not many of them ? (steam hardware survey which tend to lag a little bit do show millions of units already)

 

[Gideon]

that directly from Nvidia last investor call (and they said it to journalist as well before), the statement do not say it outsold the 1000 series during all Pascal lifetime, but during their respective first X days of launch. Pascal was pre first crypto, it got really popular a bit late in its cycle.

There is more than twice as many steam active account in 2025 than 2016:
View attachment 736144

You have a much bigger profesionnal/ai market now

Why little stock coming in any store ? you can buy a 5070 directly from bestbuy.com ... msrp (https://www.bestbuy.com/site/pny-ge...with-triple-fan-black/6620352.p?skuId=6620352)

It is many things, paper launch it is certainly not, Nvidia made its biggest quarter gaming revenues wise of its history during blackwell launch, what source is there that there is not many of them ? (steam hardware survey which tend to lag a little bit do show millions of units already)

Revenue has nothing to do with units shipped, when gpus cost twice as much or more these days, yeah revenue is up. Steam active accounts also have nothing to do with gpu sales, just more people using steam now, they have monopolized the market there. But it does not mean there is more than twice the people using computers now. Laptop sales is partly why discrete gpu sales have declined, other half is people refusing to upgrade their cards due to the cost. I have listed the only 5000 series to show in the steam survey vs the 1060, a lowly and woefully underpowered card today.

NVIDIA GeForce RTX 5070 - .71% of Steam users
NVIDIA GeForce RTX 5080 - .47% of Steam users
NVIDIA GeForce GTX 1060 - 2.24% of Steam users and had a decrease of .10%

As we can see the 5000 series has sold, but nothing spectacular after almost 6 months of being available to purchase. 5070 has only had about 3 months and is greater than the 5080 series already. As for stock, it's pretty bad, my Bestbuy had 2 cards available 50 miles away. They have sold likely 2 million to 2.5 million cards, which sounds like a lot until you look at the chart I posted and were half way through the year. It's been a very constrained launch, and it's pretty obvious to most anyone that wanted to upgrade, unless they had a Micro Center nearby.

 

[LukeTbk]

5070 has only had about 3 months and is greater than the 5080 series already

Yes ? the 5070 is a 5000 series gpu and expect to oversell the 5080, like the 3070 outsold the 3080, the 4070 outsold the 4080 and so on

Revenue has nothing to do with units shipped,

Of course, but selling for over 3 billions (usd$) is not paper launchy, Nvidia was talking in units I think, (not sure how clear they are), steam active account is about gaming popularity, it is not just monopolizing the market, the China market exploded since Pascal launch and when you look at percentage of steam user with a card, that where how many of those is there matter.

When every 1% is getting close to 2 millions gpu, having 1.78% of steam on them in early may (the 5060 will show up later on, 5090 is probably getting close, they are probably already at 2% of ~180 millions) and how many blackwell gpu sold to non steam gaming machine (5090-RTX 6000 ?) they could have sold over 3 millions already.

It's been a very constrained launch

For the high end sku, not necessarily for the 5070, 5060ti, 5060, how hard they ever were to get ? you can buy them from bestbuy.com you do not even need to go to an bestbuy to get them.

 

[sc5mu93]

No interest in 9080, 9090. Radeon should focus on udna and scalable chipset design. Then use AI hardware/algos to address microstudder and cross package latency. I wanna see issue-free dual gpu and mgpu solutions.

 

[Marees]

Tom has been going ham on the 9090 XT stuff lately on MLID to the point where it's since gotten annoying and I just skip his raging RDNA4 boner segments. It's just a balls-out overclocked 9070 XT with a ridiculous TDP and more VRAM; nothing revolutionary.

It is more of his wishlist than a real leak

He's publicly admitted that he co-hallucinates with chat gpt

For ex: he initially said magnus is ps6 then when pointed out admitted that it's for Xbox

 

[zandor]

I think the most we will get this gen from AMD is a 32GB model with slightly increased clocks and power limits. So an OC 9070XT with more vram. 32GB 9070XTX or something. I think they should do that, assuming NV does as expected and brings out "super" variants with 3GB GDDR7 modules. So 24GB 5080 & 5070Ti Super, etc.

 

[noko]

I am more wondering if the 9000 series will be short lived and AMD beating Nvidia on launch dates for next generation. The 9000 series absence from mobile is pushing OEMs to be almost 100% dependent on Nvidia unless just using an APU. I don’t see that winning for AMD long term.