A leak that comes, in theory, from two well-known OEMs ensures that the first Zen 3 engineering samples are demonstrating a higher CPI increase than expected, a very positive fact to which, in addition, a superior energy efficiency which we have seen in Zen 2.
As our regular readers will remember, the first information indicated that Zen 3 was going to bring profound changes at the microarchitecture level. We do not expect a significant increase in the number of cores and threads in the Ryzen 4000 compared to the current generation, something that frankly seems right to me, since with 16 cores and 32 threads we have more than enough in the general consumer market, but we do there will be a considerable jump in terms of CPI.
There was talk of an improvement in the CPI of up to 17%, but according to this new leak, the first samples have come to offer an improvement in the CPI of up to 20%. Let’s put this data in perspective, right now the Zenzen 3000, based on Zen 2, have an IPC almost identical to that of the Core 9000 and Core 10000 of Intel, although these processors reach higher working frequencies, which allows them to maintain the crown single-wire performance.
That 20% improvement in IPC against Zen 2 would be a real blow for Intel, since it would leave the chip giant without any weapon against AMD. This time not even the high working frequencies of monolithic core designs that Intel uses could offset such a large increase in IPC, which means that Zen 3 based processors will not only have higher multithreaded performance, but will also have higher single-threaded power.
If to all this we add the improvements in terms of efficiency, the conclusion we have is clear, Intel would be in troubleAnd its only way out is to jump into the 10nm ++ process once (Ice Lake uses the 10nm + process, but it’s limited to low-power processors for laptops).
Zen 3 will compete with Rocket Lake-S
That is what the dates we have seen so far tell us. AMD will launch Zen 3 later this year, although the debut of general consumer processors (Ryzen 4000) could be delayed until the first quarter of 2021. Intel, by contrast, will launch Comet Lake-S processors later this month. , and at the end of the year it will present the Rocket Lake-S processors that, in theory, they will coexist with those.
The Rocket Lake-S processors will be a last twist to the 14nm process and will be differentiated with an additional “+”, that is, 14 nm +++. Unlike Comet Lake-S processors, which have reached 10 cores and 20 threads, they will have only 8 cores and 16 threads, but will have Willow Cove cores, the same ones that use the Tiger Lake processors (10 nm ++ of low consumption for laptops), which means that they will bring a significant jump in terms of IPC (between 20% and 25%).
If this is confirmed Intel will have a competitive generation in terms of IPC, Rocket Lake-S, but won’t be able to compete with Zen 3 in core count, since the Ryzen 4000 processors based on this architecture will have configurations of up to 16 cores and 32 threads.
Efficiency and working temperatures are two separate issues. We have already seen the levels that Comet Lake-S reaches, a Core i9 10900K reaches 93 degrees and consumes 234 watts working at stock frequencies, not good values that confirm that the 14nm process has peaked, and that Intel has to abandon it as soon as possible if it wants to be able to continue competing with AMD.
>> break again, too hard.
Several people asked me to translate this post by “Ice Universe”. Here goes:
Disclaimer: I don’t know Zen3’s IPC uplift, only translating * as is *. Please take rumors with a large bowl of salt.
The only thing I endorse is the last sentence: AMD YES! 😀 pic.twitter.com/R7MsPmjPvw
– RetiredEngineer® (@chiakokhua) May 15, 2020