The best-case projection of performance
AMD’s benchmarks were unfortunately not very comprehensive, at least not for Ultra HD without ray tracing. To be on the safe side, I picked the best case of the games, but I warn against making absolute generalizations. So it’s unlikely to get better, though it will certainly get worse in some games. I beat the poor MSI GeForce RTX 3090 Ti SUPRIM X to almost 500 watts to at least get within the supposed proximity of the NDA-protected GeForce RTX 4080. Based on the extrapolation, we see about 11 percentage points advantage of the GeForce RTX 4090 FE to the Radeon RX 7900 XTX without extended power limit. I’ll come to the power consumption and possible efficiency in the next chapter.
The calculated bars also show that the Radeon RX 7900 XTX should be a bit faster than a GeForce GTX 4080 FE 16GB, if all positive factors and the lack of ray tracing coincide, while the Radeon RX 7900XT, which I had already positioned weeks ago, ends up exactly where I had already estimated it back then (and was also criticized from relevant circles for it): namely below the GeForce RTX 4080 16GB and probably a bit above a still expected GeForce RTX 4070, which is actually the discontinued RTX 4080 12GB.
Tis is exactly the deep-green sandwich I predicted when there was no talk of the XTX. The Radeon RX 4080 16GB and the Radeon RX 7900XTX shouldn’t have anything to do with ray tracing, and depending on the game, Team Red or Team Green should be ahead. The Radeon RX 7900XTX probably inherits, despite decent improvement, the slight weakness in almost all DXR game types.
The frame times can only be hypothetically scaled up. Whether this will really happen can be assumed, but without a card and suitable drivers, it’s like flying blind under the sofa in a helicopter. Speculative, but interesting nonetheless.
The variances are also a place in the sun for the synthetic RX 7900XTX when scaled up. Purely speculative, of course.
Our Unevennes index also nicely shows where the difference between playable, so-so, and buzzkill lies.
Power consumption and efficiency
The mentioned 355 watts TBP should probably be reached under full load in practice and even exceeded a bit (with a fully warmed up card) due to the handling of the power limit, which does not include all voltage converter and board losses. Since NVIDIA monitors this correctly via shunt on all inputs, the already reported 320 watts TBP for the GeForce RTX 4080 FE should even be slightly undercut. After all, NVIDIA reduced the power limit from the original 420 watts, first to 340 watts and then finally to 320 watts, because they probably noticed that they didn’t need more. The rest, of course, is again NDA-given speculation.
If AMD had waived 20 watts (or even more), it would certainly have been much more efficient, but also slower. Nevertheless, even so the AMD card (on the calculator) achieved a good result. I think that the essential goal was simply not to end up right behind two NVIDIA cards, if the GeForce RTX 4080 performs as generally assumed in the end. The fact that AMD deliberately forgoes an overclocking margin within the PCI SIG specifications for the reference card is certainly priced in there, just like the MSRP of 999 USD as a thick annoyance before the official launch of the GeForce RTX 4080 16GB on November 16, 2022.
If you extrapolate the CPU’s power consumption, the distances and values are surprisingly plausible here as well.
Summary and conclusion
Yes, I know that everything is rather speculative. As far as the possible performance of the upcoming Radeon RX 7900XTX is concerned, it will certainly score well in pure raster graphics and various applications that rely on pure processing power, but it will not be able to catch up with the GeForce RTX 4090. AMD has already mentioned this publicly as well. With suitable board partner cards and three 8-pin connectors for the power supply, the regions of a then much more efficient GeForce RTX 4090 FE 24GB could certainly be reached with power consumption values of well over 400 watts. However, you could also overclock them properly once again.
What the board partner cards of the GeForce RTX 4080 can do when you beat them up to an assumed 400 watts, for example, is then another story. That’s just too much speculation for me at the moment, because the usual GPU lottery also comes into play. We can assume that AMD is still struggling a bit with DXR, even though many things have already improved. They will probably not manage it completely. But here, too, the hope of the respective target group always dies last. Let’s see, soon we’ll all be smarter.
101 Antworten
Kommentar
Lade neue Kommentare
Veteran
Veteran
Urgestein
1
Urgestein
Mitglied
Urgestein
Urgestein
Urgestein
Neuling
1
Urgestein
Urgestein
1
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →