AusWolfMidrange had 40 CUs half a decade ago. Now, there's 60, and soon 64. I don't see any need for "slot power". As long as AMD can keep the consumption at the current 250-ish W mark while offering a performance upgrade, it'll be fine.

I only want them to fix the video playback power consumption. My 7800 XT eats more playing a film than my entire bedroom HTPC while playing a game, which is ridiculous.

At this performance level, 250 W in what will be 2025+, it'll be a joke in bad taste. Unless it sold for $199 or something.

Dr. DroAt this performance level, 250 W in what will be 2025+, it'll be a joke in bad taste. Unless it sold for $199 or something.

That's why I said: "As long as AMD can keep the consumption at the current 250-ish W mark while offering a performance upgrade..."

Considering that even the 6700 XT was a 230 W GPU, I'm not complaining as long as there is a nice performance bump.

Dr. DroAt this performance level, 250 W in what will be 2025+

AusWolfThat's why I said: "As long as AMD can keep the consumption at the current 250-ish W mark while offering a performance upgrade..."

Considering that even the 6700 XT was a 230 W GPU

The 6700 XT is heavily, ridiculously overclocked and overvolted out of the factory. In fact, users report that their undervolted cards eat only around 100W.

Amd/comments/1517zvy




Just saying to ignore the AMD settings, because they are mostly wrong.

ARFThe 6700 XT is heavily, ridiculously overclocked and overvolted out of the factory. In fact, users report that their undervolted cards eat only around 100W.

Amd/comments/1517zvy




Just saying to ignore the AMD settings, because they are mostly wrong.

This doesn't negate my point.

Chrispy_If the rumours of a 30% price/peformance shift in 2024 are accurate, then that will be the case.

Something around the performance of a 7900GRE for 30% less is very promising. I'd sure be happy with a $€£ 399 GRE. It's going to handle 1440p high-refresh and probably 4K60 which is arguably something that price point has never attained.

I think you're right there.

As a 6800XT owner a middling uplift for the flagship needs to have solid markers for me to want to move.

I know it's likely to be a pipe dream (simply because nvidia might decide not to raise the bar)

But I want 7900XTX raster, 4080 super RTX (or at least a decent uplift in RTX performance), power sipping close to or better than what's on offer and cheaper than current offerings.

I feel like they might tick 2 of the boxes, and I would forgo the want of RTX uplift if we were getting a jump in raster but it 'seems' safe to say we're not even raising the raster bar at all.

Chrispy_The only thing left is for CDPR to update FSR in CP2077 to a newer, better version because FSR2.1 suffers with ghosting behind vehicles pretty badly.

They won't do that because they are an nGreedia shop and have already announced that further development has been defunded going forward, as they are putting all their efforts into the next gen version coming in 4 years, which is, IMO crazy. Maybe they didn't get enough money for Phantom Liberty, which was an excellent addon...

BagerklestyneBut I want 7900XTX raster, 4080 super RTX (or at least a decent uplift in RTX performance), power sipping close to or better than what's on offer and cheaper than current offerings.

I feel like they might tick 2 of the boxes, and I would forgo the want of RTX uplift if we were getting a jump in raster but it 'seems' safe to say we're not even raising the raster bar at all.

I'm pessimistic when it comes to AMD claims. All of the "RDNA3" IPC improvements were thoroughly tested by dozens of independent reviewers once the 7600 came out and could be directly compared to the 6650XT. When you clock the core and memory to the same speeds, there's a big fat zero.

Either the monolithic Navi33 is a lie, and is just a rebrand of RDNA's Navi23 with AV1 bolted on, or RDNA3 really is useless with a truly-zero architectural improvement.

So when leaks say Navi48 *could* offer 7900XTX performance, that means it either needs to run at 4GHz, or AMD need to have made a 50% IPC improvement between RDNA3 and RDNA4. I'm a realist so I'm honestly expecting Navi48 to run at 2.8-3.0GHz thanks to a new node and RDNA4 to have a 10-15% IPC improvement. That's good for maybe 7900XT performance.

Hopefully I'm wrong. I'd love to be wrong and for these >30% performance/$ claims to pan out!

Chrispy_I'm pessimistic when it comes to AMD claims. All of the "RDNA3" IPC improvements were thoroughly tested by dozens of independent reviewers once the 7600 came out and could be directly compared to the 6650XT. When you clock the core and memory to the same speeds, there's a big fat zero.

Either the monolithic Navi33 is a lie, and is just a rebrand of RDNA's Navi23 with AV1 bolted on, or RDNA3 really is useless with a truly-zero architectural improvement.

So when leaks say Navi48 *could* offer 7900XTX performance, that means it either needs to run at 4GHz, or AMD need to have made a 50% IPC improvement between RDNA3 and RDNA4. I'm a realist so I'm honestly expecting Navi48 to run at 2.8-3.0GHz thanks to a new node and RDNA4 to have a 10-15% IPC improvement. That's good for maybe 7900XT performance.

Hopefully I'm wrong. I'd love to be wrong and for these >30% performance/$ claims to pan out!

Totally agree. I said so at the time when it came to light RDNA3 was hot and low performance that something had gone very wrong, or we have been lied to. What makes me think the chip is some kind of lie is that AMD never fixed it, and if it was some simple chip errata then surely they would have done a re-spin and used it to kick nGreedias arse by now and used the old, broken chip to launch a low-end range with reduced speed and lower thermals.

I also think Raja screwed AMD and blew a fortune on second-rate chip designers and have left AMD with an architecture that's impossibly expensive to fix and or to scale up passed where it was 24 months ago. This would also explain the awful performance of their "next gen" RDNA4 architecture, which will at best catch up with nGreedias current gen, just not the 4090. And yes raster perf, blah blah, but that's not where the market is going. nGreedia have won the RT war, and more and more games will use it, especially as Sony is going to push it hard with the PS5 refresh. For AMD to have sub-par RT perf with RDNA4 in 2025/6 would be a disaster for them, so let's hope they at least get that right!

At the moment AMD are nearly two generations behind nGreedia. A position I don't think they can come back from, as far as the high-end gaming market is concerned, unless they can match or beat the 4090 perf with RT for substantially less money. They will be battling nGreedias bottom to mid-range 50x0 series with RDNA4, and Intels Battlemage cards, which could possibly outperform all but AMD's highest end cards, for less money!

Let's end by hoping AMD are majorly sandbagging RDNA4 for the last 12 months, and that there will be a performance variant of the chip which takes them way past the 4090.

RX 7900 XTX has 96 RT cores, while RTX 4090 has 128 RT cores.
AMD needs to double the RT cores and/or make them much beefier in order to match nvidia's ray-traced performance.
I don't think it's that dificult, not to say impossible.
Maybe AMD decided to cancel the RDNA projects and start from scratch on a brand new architecture with ray-tracing in its core.

Chrispy_I'm pessimistic when it comes to AMD claims. All of the "RDNA3" IPC improvements were thoroughly tested by dozens of independent reviewers once the 7600 came out and could be directly compared to the 6650XT. When you clock the core and memory to the same speeds, there's a big fat zero.

Either the monolithic Navi33 is a lie, and is just a rebrand of RDNA's Navi23 with AV1 bolted on, or RDNA3 really is useless with a truly-zero architectural improvement.

So when leaks say Navi48 *could* offer 7900XTX performance, that means it either needs to run at 4GHz, or AMD need to have made a 50% IPC improvement between RDNA3 and RDNA4. I'm a realist so I'm honestly expecting Navi48 to run at 2.8-3.0GHz thanks to a new node and RDNA4 to have a 10-15% IPC improvement. That's good for maybe 7900XT performance.

Hopefully I'm wrong. I'd love to be wrong and for these >30% performance/$ claims to pan out!

While I agree that the fastest of the rumoured RDNA 4 SKUs is unlikely to be faster than the 7900 XT, I don't think that RDNA 3 has no improvements compared to RDNA2. RDNA3 improves upon RDNA2, but the compiler isn't good enough to make decent use of its dual issue capabilities. There are examples where RDNA3 improves upon RDNA2 significantly, but they seem to be limited in number which would suggest those represent cases of hand optimized code. For instance, in Cyberpunk 2077, the 7600 is faster than the 6650 XT despite clocking lower: 2525 MHz and 2699 MHzrespectively. Another example is the 7800 XT which is significantly faster than the RX 6800. This isn't solely due to clock speed, because the 7800 XT has only about 10% higher clock speeds.

I don't anticipate any significant improvements to the compiler and we don't know enough about RDNA 4 to determine if it would increase the IPC of each compute unit. Cautious pessimism seems to be the prudent option and that would suggest that almost all of the improvement will be the result of higher clock speeds and more compute units.

The question is not about whether Navi 48 will reach a certain performance level (be it RX 7900 XTX, XT or RX 7800 XT, RTX 4090, RTX 4080, etc.), the question is about its price level.
AMD needs to give us the RX 7800 XT level of performance for 200$. That will be revolutionary and will make the thing sell like no tomorrow/hot cakes.

ARFThe question is not about whether Navi 48 will reach a certain performance level (be it RX 7900 XTX, XT or RX 7800 XT, RTX 4090, RTX 4080, etc.), the question is about its price level.
AMD needs to give us the RX 7800 XT level of performance for 200$. That will be revolutionary and will make the thing sell like no tomorrow/hot cakes.

Not going to be anywhere near $200.

7800XT with a 30% improvement in performance/$ is another way of saying $499 * 1/1.3, which is $385, which they'll "round up" to $399, and then only the base models will cost that, with fancy versions like the TUF and Nitro+ costing $450.

As always, paying for an overbuilt cooler is worse value the cheaper your base product becomes, so on a $399 part, you REALLY want be looking for the best-quality $399 card. There's absolutely no way a slightly quieter cooler is worth an extra 13% when the MSRP models usually include a couple of high-quality, very quiet editions. You can usually count on Asus and Sapphire to make a Dual/Pulse that doesn't suck.

Chrispy_Not going to be anywhere near $200.

7800XT with a 30% improvement in performance/$ is another way of saying $499 * 1/1.3, which is $385, which they'll "round up" to $399, and then only the base models will cost that, with fancy versions like the TUF and Nitro+ costing $450.

As always, paying for an overbuilt cooler is worse value the cheaper your base product becomes, so on a $399 part, you REALLY want be looking for the best-quality $399 card. There's absolutely no way a slightly quieter cooler is worth an extra 13% when the MSRP models usually include a couple of high-quality, very quiet editions. You can usually count on Asus and Sapphire to make a Dual/Pulse that doesn't suck.

My current go to is As Rock. Nice and quiet but also truly 2 slots, with proper cooling. Even cheaper than Gigabyte for the Challenger series cards. Pulse are great if you know you want to put your card on water.

Chrispy_Not going to be anywhere near $200.

7800XT with a 30% improvement in performance/$ is another way of saying $499 * 1/1.3, which is $385, which they'll "round up" to $399, and then only the base models will cost that, with fancy versions like the TUF and Nitro+ costing $450.

RX 7800 XT today ~500$.
You say RX 8800 XT (RX 7800 XT +30%) ~400$.

Would not be disastrously bad, but still nothing special and I don't think it will motivate the sales much.

I don't think it will reach that performance. I am more inclined to believe RX 7800 XT +-5%.

kapone32Pulse are great if you know you want to put your card on water.

They're also great running at stock. Compact, quiet and cheap. Pulse has been my go-to ever since Asus f*ed up the 5700 XT Strix.

Chrispy_The only thing left is for CDPR to update FSR in CP2077 to a newer, better version because FSR2.1 suffers with ghosting behind vehicles pretty badly.

If you're at 4K and your GPU isn't ancient it makes no sense to consider FSR whatsoever. XeSS 1.3 at 50% scaling gets you better IQ and much less stability issues than FSR at any mode. It's my go-to on a 6700 XT, netting me 50 to 75 FPS at otherwise 4K High depending on a location. Not so Intel-favoured in other games though.

I don't expect AMD to release anything beyond 25 percent better bang per buck compared to RDNA3 in RDNA4. And honestly, Intel's upcoming generation is more promising. But "7900 XTX performance at 500 USD and below 350 W" is what we need now for competition to become a thing but we'll never get until that becomes a given in NVIDIA's SKUs.