The first rumors we saw about the possible performance of Ryzen 7000 processors based on the Zen 4 architecture turned out to be untrue. During the official presentation, AMD confirmed that this architecture will allow an increase in single-fiber performance of up to 15% compared to the Ryzen 5000, a percentage that already included both IPC and increased clock speeds. Since we had two performance-enhancing values, it was clear the CPI increase should be less than 15%.

Despite the fact that this data came from AMD itself and that it was made clear that a state-of-the-art processor had been used, some sources that had been involved in marketing rumors for some time were still determined to say that Zen 4 was going to increase IPC among 20 % and 40%, which are values ​​that were real nonsense and which only became apparent in the end, as AMD has confirmed that this new architecture will offer improvements between 8% and 10% IPC compared to Zen 3. The 5% -7% that remains to reach the 15% that AMD gave will be due to an increase in clock speeds.

Will it be enough to overcome Raptor Lake-S? It was a topic that we touched on in depth in this article, honestly I’m still inclined not to. And I think it’s very likely that Intel’s new generation of consumer processors will eventually win the crown of single-threaded performance. However, I doubt multi-threaded performance, especially when the Raptor Lake-S retains the Gracemont architecture on highly efficient cores.

Continuing with the above, the Ryzen 9 7950X should have 16 Zen 4 cores and 32 threads, while the Intel Core i9-13900K should 8 Raptor Cove cores high power, which would increase the power of both the IPC and its higher clock speed, and 16 cores Gracemont high efficiency that would only improve performance by increasing their beat. This would add a total of 24 cores and 32 threads.

Intel Core i9-13900K would have more cores than Ryzen 9 7950X, but 16 of those cores would have high efficiency and would have lower IPC and frequency than the Zen 4 kernel, so I have doubts about which of the two could be the most powerful in multithreading. In any case, as I told you at the time, the most important thing is that we have an epic battle with Intel and AMD in top form.

Zen 4 also promises significant efficiency improvements

And this will be possible not only by jumping into 5nm TSMC node, but also the adjustments that AMD will present at the microarchitecture level. In this sense, we can highlight, for example, the increase in L2 cache per core, which goes from 512 KB to 1 MB, and also the increase in memory bandwidth per core, which increases by 125% compared to the previous generation. . Finally, AMD also confirmed the integration of new ISA instructions to support AI and AVX-512.

The Ryzen 7000 processors, based on the Zen 4 architecture, will be launched later this year. Be able to use them we will need a motherboard with socket AM5 and RAM type DDR5, because this generation will not support DDR4 memory. At the time, Intel was widely criticized for making the leap in memory of DDR5, even though it retained the ability to use DDR4 with Alder Lake-S, so it is clear that AMD’s decision will also lead to criticism, although it may happen that its The most loyal fans do not perceive this as a problem. It is the usual double measuring rod for each of them.

Before I finish, I want to share with you the latest update of AMD’s plan for the CPU sector and HEDT for ordinary consumers. We can see it in that Zen 4 will have three versions, a regular consumer, a regular consumer with a 3D stacked cache and a version of HEDT that will be used in the upcoming Threadripper processors. Zen 4c will have no application in the mainstream consumer market and Zen 5 is still scheduled for 2024.