4 ms·

It's weird how AMD managed to make RDNA 3 such a letdown given RDNA 2 fought with 3090/3090 Ti neck to neck in raster performance whereas 7900XTX is at the 4080

by bitL 4y ago

It's weird how AMD managed to make RDNA 3 such a letdown given RDNA 2 fought with 3090/3090 Ti neck to neck in raster performance whereas 7900XTX is at the 4080 level. They use a much better node compared to RDNA 2 as well but are stuck at similar frequencies and gained only around 30% on average while 4090 doubled 3090 performance.

oynqr 4y ago
Nvidia hiked the TDP even higher with the 4090 to keep the performance crown.
- mardifoufs 4y ago
  Not really, the 4090 is still miles ahead even at lower power consumption than the top end AMD cards ( 7900xtx).
  - teach 4y ago
    Do you have a source for this? I thought I remembered that the 7900 XTX trades blows with the 4090 in some games even with both cards at stock settings. (Though of course the 4090 completely blows it out of the water in other games.)
    - zamadatix 4y ago
      Are you sure you're not thinking the 4080? It has very similar performance but a lower TDP https://tpucdn.com/review/amd-radeon-rx-7900-xtx/images/relative-performance_3840-2160.png https://tpucdn.com/review/amd-radeon-rx-7900-xtx/images/rela... But regarding the 4090 you can drop 100W and barely lose any performance at all e.g. https://i.redd.it/d7lq97nttkv91.png https://i.redd.it/d7lq97nttkv91.png. I assume it's at least somewhat similar for the 7900 XTX as well. The power scaling capping out so hard was really quite the disappointment given the new 600W connector.
    - giobox 4y ago
      You shouldn't need someone to fetch you a source - the performance gap is big and well documented in every single review of the 4090 or 7900XTX - hardly much of a surprise really given we are comparing a 999 dollar part to a 1600 dollar one. Just pick literally any review from the top of a google search with benchmarks. The 7900 XTX will outperform the $1200 4080 in some games using rasterized rendering, but not conclusively across the board. With ray tracing turned on, the 4080 however generally always comes out ahead in benchmarks, so the 4080 and 7900XTX can definitely be described as trading blows. There will be almost no benchmark a 7900XTX trades blows with a ~$1600 4090 on (perhaps with exception of performance per dollar...) - the 4090 will substantially outperform a 7900XTX in most things so long as you aren't CPU bound etc - its generally some 30 percent faster than the 4080 which the 7900XTX is roughly equivalent to. If money is no object for a buyer, the gap between the fastest Nvidia gaming part and the fastest AMD gaming part is pretty big at the moment, assuming you can stomach the MSRP of the 4090. The 4090 is really in a performance league of its own right now. The 7900XTX's biggest feature today is that its MSRP is 200 dollars less than the 4080, although both cards can be argued to be priced too high this generation relative to the performance offered and price points of prior gens. Rumors abound the 4080 is going to get a price cut soon.
    - paulmd 4y ago
      > I thought I remembered that the 7900 XTX trades blows with the 4090 in some games even with both cards at stock settings. (Though of course the 4090 completely blows it out of the water in other games.) Yes, that's the way averages work ;) 4090 is 127.7% of 7900XTX performance averaged across all reviewers at 4K in non-RT titles. So a good game for a 7900XTX is rubbing shoulders with the 4090... and a good game for a 4090 is doubling the framerate of a 7900XTX. 4090 is not lower power than a 7900XTX though. Which is to be expected because it's a significantly faster card. However, the 4090 does undershoot its specified TDP quite a bit, it's 418W average in gaming (on the next page) yielding 107% perf/w of a 7900XTX (so it's slightly more efficient too, just delivers a lot more frames). https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-rx-7900-xt-xtx/launch-analyse-amd-radeon-rx-7900-xt-xtx-seite-2 https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-r...
  - EugeneOZ 4y ago
    450 W vs 355 W is not "miles away". They both are hogs.
    - paulmd 4y ago
      Meta-analysis: https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-rx-7900-xt-xtx/launch-analyse-amd-radeon-rx-7900-xt-xtx-seite-3 https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-r... across all launch reviews, 7900XTX averages 351W during gaming, while 4090 averages 418W for 127.7% of the performance (in raster-only 4K, on previous page), working out to 107% of the perf/w. 4080 is actually even more efficient, 297W actual power for 96.9% of 4K performance, giving 115% of the perf/w of the 7900XTX. 7900XT actually falls a bit farther behind, at only 96% of the perf/w. Compared to a 3080 10GB (which is not a power-hog, that's one of the original fairly-efficient configs) being 70% of the perf/w, that makes the 4080 115/70 = 164% of the perf/w, they are significantly more efficient than previous gens, not at all "power hogs". 6800XT (probably the most efficient card of last gen) is 74% of the perf/w of 7900XTX and that puts the 4080 at 155% of the perf/w of the 6800XT. Here's the 4070 Ti data but it doesn't really change anything (conveniently the 4070 Ti efficiency works out just about exactly the same as 7900XTX): https://www.3dcenter.org/artikel/launch-analyse-nvidia-geforce-rtx-4070-ti/launch-analyse-nvidia-geforce-rtx-4070-ti-seite-4 https://www.3dcenter.org/artikel/launch-analyse-nvidia-gefor... And please don't whip out the "X watts is too much", arguing that products that you don't want to purchase personally shouldn't be manufactured is a shitty thing to do. If you want a 300W card, buy the 4080. If you want a 270W card, the 4070 Ti is right there. Lower-end cards will be launched over time, it's not going to be high-end-only forever. Ada Lovelace is by a large margin the most efficient generation that has ever launched. The problem is clearly pricing and not efficiency, people just listed to a bunch of dumb techtubers last summer and earnestly believed that a 900W TBP card was going to be a mainstream product and once gamerzzz latch onto an NVIDIA conspiracy theory they're sure as hell not going to renounce it. Efficiency bad, it doesn't matter what the numbers say. Oh, efficiency is good? Well still too much power. Oh there's cards available at the same power as last gen? Listen here buddy NVIDIA bad, did I stutter?
      - EugeneOZ 4y ago
mnd999 4y ago
Any good sources for this? Most of the RDNA3 articles I’ve seen are pre-launch hype saying it’s going to be amazing. Definitely interested in some solid tests.
paulmd 4y ago
The gap has remained pretty consistent since Pascal/Vega. NVIDIA is roughly 20-30% ahead "architectural lead" (which I think is roughly approximated by perf-per-transistor perhaps with some wiggle room for node-quality and featureset, eg RDNA2 lacks tensors and has much slower RT etc). That can mean smaller dies/lower cost, or it can mean lower clocks/higher efficiency from a similar-sized die, or in the case of Ampere/Turing, older nodes+bigger dies at lower cost. Now that they've snapped back to TSMC N5 family (it's a custom 4N node, not to be confused with N4, and it's not clear whether the basis for 4N is N4 or N5P, but both N4 and N5P are N5 family) this has reappeared as performance wins and higher efficiency. NVIDIA used older nodes on Ampere/Turing while AMD moved ahead to TSMC N7. NVIDIA is just so far ahead architecturally that they can roughly match AMD's efficiency despite a full-node gap in technology and going against (specifically) the best node in the industry. The older node let them reduce costs, which is part of the reason why AMD has been perceived as reluctant to play the price wars during Turing/Ampere. NVIDIA was playing with older cheap nodes and AMD was basically using the best, most expensive node available just to compete (not even significantly beat), and a GPU needs a nice big chunk of it that could easily be several high-margin CPU dies instead. Nobody is ever going to publicly attribute a positive intent to NVIDIA ever, but, Ampere and Turing were an earnest effort to keep the cost down for consumers. 2080 Ti was the same $999 MSRP as the first GK110 card 10+ years ago. 3080 was so exquisitely priced that AMD sighed and dragged their feet and finally marked theirs 50 bucks less and threw in the towel and didn't barely produce it for the first year. 3060 Ti should have been the next RX 480 level value-proposition that keeps mainstream gamers happy for the next 5 years. Mining ruined everything - yet again. The two negative consequences of this were efficiency and top-end performance. Samsung 8nm is a 10+ node, basically a full node behind TSMC N7, and obviously that has perf/w consequences. NVIDIA's architectural advantage mostly held them over but they still slid in a little under RDNA2's efficiency, when if they'd been on N7 themselves they could have easily been 20%+ more efficient. And it resulted in huge dies, because the perf/mm2 is much worse on a 10+ than on N7. So they didn't really have a top-end - a 3090 is really more comparable to what they could have gotten out of a full-die x104 on N7, but 754mm2 is as big as they could go on Samsung 8. That is part of why the 4090 towers above the rest of the lineup in a way the 3090 didn't. 30-series didn't really have a high-end part, they had "castoffs" (full dies) from the mainstream-enthusiast part that happened to have a little extra headroom. A little bit of product strategy - you put your cutdowns at the high-value points of your lineup. GTX 970 and 980 Ti being cutdowns, 1080 Ti being a cutdown, etc. In CPUs, that's where AMD puts the 3600 and 3900X, and Intel puts the 5820K, etc. With Ampere, NVIDIA went as big as possible, then put the 3080 as being the GA102 cutdown, to make it a great value. Same for 3060 Ti. And the 3090 and the 3070 are the bloopers that result from the full-die chips being more expensive. But with the AD102 die being 754mm2 already (fucking enormous), that's as good as they could do on Samsung 8nm, so they kind of abandoned the true top-end stuff. That was the strategy on 8nm - big dies on a shitty node, and make the 3080 a cutdown to take advantage of it. And on Turing it was the 2080 Ti (because 12FFN was the same node as Pascal was on, it's a 16nm with no optical shrink just the "your mom sized" reticle limit) plus pushing some of those "non-raster" performance-scaling approaches like DLSS. They also needed to bring in a custom memory tech (GDDR6X is NVIDIA-exclusive) to keep costs down - Samsung 8nm has terrible SRAM density compared to TSMC N7 so the "infinity cache" strategy is off the table (or at least you'd need to stack it, and the packaging step drives up costs a lot too). 512b buses are tough to route these days and a lot of people don't think it's possible anymore with tighter integrity requirements (again, maybe it'd be possible with fanout packaging which, quelle surprise, drives up costs). But PAM4 lets you move a lot more data per clock edge using basically the same IC form-factor. The only real alternative would have been HBM2 which would also have driven up costs. So NVIDIA traded a custom memory tech to allow them to use a shitty node that they could get cheaply and produce these massive, fairly low-clocked dies on what amounts to an overgrown smartphone node, all in the name of cost reduction and bending MSRPs downwards ($399 3060 Ti, $699 3080). And miners royally fucked it all up, yet again. But anyway to go back to the original point - this gap has always existed, AMD just appeared competitive because they were using a better, more expensive node, and they took it on the chin in margins as a result. And now that NVIDIA has moved back to a leading node (very leading in fact) the gap has reappeared. People just didn't want to do the math on comparing products across two nodes and realize the gap was always there, but people like me have been saying something along the lines of "RDNA1 isn't all that great considering it's N7 vs 16nm and NVIDIA still is slightly ahead on perf/w" and "RDNA2 is moving forward but it's still barely beating NVIDIA even though they're using a garbage Samsung 10+ node" quite frequently and we got shouted down. I agree that it seems like RDNA3 missed somewhere though. The efficiency gains are massively lower than what AMD promised in their earlier slides, the performance seems disappointing. Maybe our expectations were just mis-calibrated, and maybe the marketing slides were just cooked, but it does seem like it missed some target by 20-30% and nobody external knows what the specific reason is.

by bitL 4y ago

oynqr 4y ago

Nvidia hiked the TDP even higher with the 4090 to keep the performance crown.

mardifoufs 4y ago
Not really, the 4090 is still miles ahead even at lower power consumption than the top end AMD cards ( 7900xtx).
- teach 4y ago
  Do you have a source for this? I thought I remembered that the 7900 XTX trades blows with the 4090 in some games even with both cards at stock settings. (Though of course the 4090 completely blows it out of the water in other games.)
  - zamadatix 4y ago
    Are you sure you're not thinking the 4080? It has very similar performance but a lower TDP https://tpucdn.com/review/amd-radeon-rx-7900-xtx/images/relative-performance_3840-2160.png https://tpucdn.com/review/amd-radeon-rx-7900-xtx/images/rela... But regarding the 4090 you can drop 100W and barely lose any performance at all e.g. https://i.redd.it/d7lq97nttkv91.png https://i.redd.it/d7lq97nttkv91.png. I assume it's at least somewhat similar for the 7900 XTX as well. The power scaling capping out so hard was really quite the disappointment given the new 600W connector.
  - giobox 4y ago
    You shouldn't need someone to fetch you a source - the performance gap is big and well documented in every single review of the 4090 or 7900XTX - hardly much of a surprise really given we are comparing a 999 dollar part to a 1600 dollar one. Just pick literally any review from the top of a google search with benchmarks. The 7900 XTX will outperform the $1200 4080 in some games using rasterized rendering, but not conclusively across the board. With ray tracing turned on, the 4080 however generally always comes out ahead in benchmarks, so the 4080 and 7900XTX can definitely be described as trading blows. There will be almost no benchmark a 7900XTX trades blows with a ~$1600 4090 on (perhaps with exception of performance per dollar...) - the 4090 will substantially outperform a 7900XTX in most things so long as you aren't CPU bound etc - its generally some 30 percent faster than the 4080 which the 7900XTX is roughly equivalent to. If money is no object for a buyer, the gap between the fastest Nvidia gaming part and the fastest AMD gaming part is pretty big at the moment, assuming you can stomach the MSRP of the 4090. The 4090 is really in a performance league of its own right now. The 7900XTX's biggest feature today is that its MSRP is 200 dollars less than the 4080, although both cards can be argued to be priced too high this generation relative to the performance offered and price points of prior gens. Rumors abound the 4080 is going to get a price cut soon.
  - paulmd 4y ago
    > I thought I remembered that the 7900 XTX trades blows with the 4090 in some games even with both cards at stock settings. (Though of course the 4090 completely blows it out of the water in other games.) Yes, that's the way averages work ;) 4090 is 127.7% of 7900XTX performance averaged across all reviewers at 4K in non-RT titles. So a good game for a 7900XTX is rubbing shoulders with the 4090... and a good game for a 4090 is doubling the framerate of a 7900XTX. 4090 is not lower power than a 7900XTX though. Which is to be expected because it's a significantly faster card. However, the 4090 does undershoot its specified TDP quite a bit, it's 418W average in gaming (on the next page) yielding 107% perf/w of a 7900XTX (so it's slightly more efficient too, just delivers a lot more frames). https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-rx-7900-xt-xtx/launch-analyse-amd-radeon-rx-7900-xt-xtx-seite-2 https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-r...
- EugeneOZ 4y ago
  450 W vs 355 W is not "miles away". They both are hogs.
  - paulmd 4y ago
    Meta-analysis: https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-rx-7900-xt-xtx/launch-analyse-amd-radeon-rx-7900-xt-xtx-seite-3 https://www.3dcenter.org/artikel/launch-analyse-amd-radeon-r... across all launch reviews, 7900XTX averages 351W during gaming, while 4090 averages 418W for 127.7% of the performance (in raster-only 4K, on previous page), working out to 107% of the perf/w. 4080 is actually even more efficient, 297W actual power for 96.9% of 4K performance, giving 115% of the perf/w of the 7900XTX. 7900XT actually falls a bit farther behind, at only 96% of the perf/w. Compared to a 3080 10GB (which is not a power-hog, that's one of the original fairly-efficient configs) being 70% of the perf/w, that makes the 4080 115/70 = 164% of the perf/w, they are significantly more efficient than previous gens, not at all "power hogs". 6800XT (probably the most efficient card of last gen) is 74% of the perf/w of 7900XTX and that puts the 4080 at 155% of the perf/w of the 6800XT. Here's the 4070 Ti data but it doesn't really change anything (conveniently the 4070 Ti efficiency works out just about exactly the same as 7900XTX): https://www.3dcenter.org/artikel/launch-analyse-nvidia-geforce-rtx-4070-ti/launch-analyse-nvidia-geforce-rtx-4070-ti-seite-4 https://www.3dcenter.org/artikel/launch-analyse-nvidia-gefor... And please don't whip out the "X watts is too much", arguing that products that you don't want to purchase personally shouldn't be manufactured is a shitty thing to do. If you want a 300W card, buy the 4080. If you want a 270W card, the 4070 Ti is right there. Lower-end cards will be launched over time, it's not going to be high-end-only forever. Ada Lovelace is by a large margin the most efficient generation that has ever launched. The problem is clearly pricing and not efficiency, people just listed to a bunch of dumb techtubers last summer and earnestly believed that a 900W TBP card was going to be a mainstream product and once gamerzzz latch onto an NVIDIA conspiracy theory they're sure as hell not going to renounce it. Efficiency bad, it doesn't matter what the numbers say. Oh, efficiency is good? Well still too much power. Oh there's cards available at the same power as last gen? Listen here buddy NVIDIA bad, did I stutter?
    - EugeneOZ 4y ago

mnd999 4y ago

Any good sources for this? Most of the RDNA3 articles I’ve seen are pre-launch hype saying it’s going to be amazing. Definitely interested in some solid tests.

paulmd 4y ago

The gap has remained pretty consistent since Pascal/Vega. NVIDIA is roughly 20-30% ahead "architectural lead" (which I think is roughly approximated by perf-per-transistor perhaps with some wiggle room for node-quality and featureset, eg RDNA2 lacks tensors and has much slower RT etc). That can mean smaller dies/lower cost, or it can mean lower clocks/higher efficiency from a similar-sized die, or in the case of Ampere/Turing, older nodes+bigger dies at lower cost. Now that they've snapped back to TSMC N5 family (it's a custom 4N node, not to be confused with N4, and it's not clear whether the basis for 4N is N4 or N5P, but both N4 and N5P are N5 family) this has reappeared as performance wins and higher efficiency. NVIDIA used older nodes on Ampere/Turing while AMD moved ahead to TSMC N7. NVIDIA is just so far ahead architecturally that they can roughly match AMD's efficiency despite a full-node gap in technology and going against (specifically) the best node in the industry. The older node let them reduce costs, which is part of the reason why AMD has been perceived as reluctant to play the price wars during Turing/Ampere. NVIDIA was playing with older cheap nodes and AMD was basically using the best, most expensive node available just to compete (not even significantly beat), and a GPU needs a nice big chunk of it that could easily be several high-margin CPU dies instead. Nobody is ever going to publicly attribute a positive intent to NVIDIA ever, but, Ampere and Turing were an earnest effort to keep the cost down for consumers. 2080 Ti was the same $999 MSRP as the first GK110 card 10+ years ago. 3080 was so exquisitely priced that AMD sighed and dragged their feet and finally marked theirs 50 bucks less and threw in the towel and didn't barely produce it for the first year. 3060 Ti should have been the next RX 480 level value-proposition that keeps mainstream gamers happy for the next 5 years. Mining ruined everything - yet again. The two negative consequences of this were efficiency and top-end performance. Samsung 8nm is a 10+ node, basically a full node behind TSMC N7, and obviously that has perf/w consequences. NVIDIA's architectural advantage mostly held them over but they still slid in a little under RDNA2's efficiency, when if they'd been on N7 themselves they could have easily been 20%+ more efficient. And it resulted in huge dies, because the perf/mm2 is much worse on a 10+ than on N7. So they didn't really have a top-end - a 3090 is really more comparable to what they could have gotten out of a full-die x104 on N7, but 754mm2 is as big as they could go on Samsung 8. That is part of why the 4090 towers above the rest of the lineup in a way the 3090 didn't. 30-series didn't really have a high-end part, they had "castoffs" (full dies) from the mainstream-enthusiast part that happened to have a little extra headroom. A little bit of product strategy - you put your cutdowns at the high-value points of your lineup. GTX 970 and 980 Ti being cutdowns, 1080 Ti being a cutdown, etc. In CPUs, that's where AMD puts the 3600 and 3900X, and Intel puts the 5820K, etc. With Ampere, NVIDIA went as big as possible, then put the 3080 as being the GA102 cutdown, to make it a great value. Same for 3060 Ti. And the 3090 and the 3070 are the bloopers that result from the full-die chips being more expensive. But with the AD102 die being 754mm2 already (fucking enormous), that's as good as they could do on Samsung 8nm, so they kind of abandoned the true top-end stuff. That was the strategy on 8nm - big dies on a shitty node, and make the 3080 a cutdown to take advantage of it. And on Turing it was the 2080 Ti (because 12FFN was the same node as Pascal was on, it's a 16nm with no optical shrink just the "your mom sized" reticle limit) plus pushing some of those "non-raster" performance-scaling approaches like DLSS. They also needed to bring in a custom memory tech (GDDR6X is NVIDIA-exclusive) to keep costs down - Samsung 8nm has terrible SRAM density compared to TSMC N7 so the "infinity cache" strategy is off the table (or at least you'd need to stack it, and the packaging step drives up costs a lot too). 512b buses are tough to route these days and a lot of people don't think it's possible anymore with tighter integrity requirements (again, maybe it'd be possible with fanout packaging which, quelle surprise, drives up costs). But PAM4 lets you move a lot more data per clock edge using basically the same IC form-factor. The only real alternative would have been HBM2 which would also have driven up costs. So NVIDIA traded a custom memory tech to allow them to use a shitty node that they could get cheaply and produce these massive, fairly low-clocked dies on what amounts to an overgrown smartphone node, all in the name of cost reduction and bending MSRPs downwards ($399 3060 Ti, $699 3080). And miners royally fucked it all up, yet again. But anyway to go back to the original point - this gap has always existed, AMD just appeared competitive because they were using a better, more expensive node, and they took it on the chin in margins as a result. And now that NVIDIA has moved back to a leading node (very leading in fact) the gap has reappeared. People just didn't want to do the math on comparing products across two nodes and realize the gap was always there, but people like me have been saying something along the lines of "RDNA1 isn't all that great considering it's N7 vs 16nm and NVIDIA still is slightly ahead on perf/w" and "RDNA2 is moving forward but it's still barely beating NVIDIA even though they're using a garbage Samsung 10+ node" quite frequently and we got shouted down. I agree that it seems like RDNA3 missed somewhere though. The efficiency gains are massively lower than what AMD promised in their earlier slides, the performance seems disappointing. Maybe our expectations were just mis-calibrated, and maybe the marketing slides were just cooked, but it does seem like it missed some target by 20-30% and nobody external knows what the specific reason is.