AMD Ryzen 9 7900X and AMD Ryzen 5 7600X in review: Back to the fastest gaming CPU crown with Raphael!

[Redaktion]

Almost two years after the release of the Ryzen 5000 processors, AMD now unleashes its current line-up aka Raphael with a modern Zen 4 architecture into the race for the performance crown of the fastest CPU. Intel's Core i9-12900K often only sees the taillights with second-fastest model in our review, the Ryzen 9 7900X. You can find out how big the lead is and what AMD's smaller Ryzen 5 7600X can do in this detailed review.

https://www.notebookcheck.net/AMD-Ryzen-9-7900X-and-AMD-Ryzen-5-7600X-in-review-Back-to-the-fastest-gaming-CPU-crown-with-Raphael.657698.0.html

[NikoB]

Something tells me that the population of the world is no longer up to games ...

[NikoB]

And yes, it is fast again only in multithreading. In a single Intel thread, everything is also faster even with the old generation. Therefore, a neat test for a single thread was not given here, so that it would not be clearly evident.

AMD has nothing special to offer the buyer in 2022-2023, except for discounts in price compared to Intel processors even in 2021. But Raptor Lake is faster ...

[yus]

Single thread data with Cinebench R15 is presented, 324 for 7900X, 315 for 7600X, 286 for 12900K. It is also presented with R20, and R23.

[yus]

And yes, it is fast again only in multithreading. In a single Intel thread, everything is also faster even with the old generation. Therefore, a neat test for a single thread was not given here, so that it would not be clearly evident.

AMD has nothing special to offer the buyer in 2022-2023, except for discounts in price compared to Intel processors even in 2021. But Raptor Lake is faster ...

Single thread data with Cinebench R15 is presented, 324 for 7900X, 315 for 7600X, 286 for 12900K. It is also presented with R20, and R23.

[Russell]

Zen was big..
Zen 2 was revolutionary, especially in laptop space..
Zen 3 was definitely an upgrade...

But Zen 4?
Something tells me amd needs another Jim Keller moment if they don't want to be reduced go Rocketlake standards...

[NikoB]

Zen was big..
Zen 2 was revolutionary, especially in laptop space..
Zen 3 was definitely an upgrade...

But Zen 4?
Something tells me amd needs another Jim Keller moment if they don't want to be reduced go Rocketlake standards...

The biggest disappointment (I think it will also be in mobile chips) is the fake Display Port 2.0. They (AMD) themselves explicitly declared support for only UHBR10 (40Gbit/s) mode, but not 20 (80 Gbit/s). Thus, Zen4 will be no support for 8k monitors on laptops (and on PCs) again. As for the GTX4xxx, I don't know yet, especially since so far only tops are being sold, which most people don't need for work.

Again, we are waiting for either junior discrete chips like 4050 or for Zen4+ or what Intel will actually offer in Raptor Lake for built-ins gpu, while I don't really understand what they really did there.

In general, of course, it is clear what the problem is, on Zen4 the memory is again too slow. Only ridiculous 70-80GByte/s, but it is necessary for confident support for 8k - 120+ in the built-in gpu. But discrete card are ready for 8k and DP2.0 for at least 7-8 years. And it's a shame that all this is being introduced so slowly...

In general, desktop processors now formally support the Display Port 2.0 protocol with 40 Gbit/s, i.e. halved from the full version and have support for HDMI 2.1 with FRL, but without an explicit indication of the limit mode (unlike DP2.0). Whether there is support for 12Gbit/s mode for each line out of 4 is not known, otherwise, again, HDMI 2.1 may turn out to be fake in AMD's implementatiion, not full-fledged with 48 Gbit/s bandwidth.

AVX512 also runs in 2 cycles, not one. On the other hand, Intel generally cut out its hardware support outside of server chips.

From my point of view, both Zen4 and Raptor Lake are a passerby. We have to wait for the next generation from both AMD and Intel, because only there will apparently be full support for DP2.0 and HDMI 2.1, as well as AV1 hardware encoding. Yes, and by that time it will apparently be possible to bring the memory 100+ GByte/s in bandwidth.