[Hardware] Intel's new gaming CPUs will perform less in RAM, are they limited?

[WilkerCSBD]

This week is hot with the new Intel gaming CPUs, because performance in games aside, the novelties that will include the Willow Cove micro architecture are going to bring tail. If something could boast Intel without too many problems, it was the fact that its BMI reached higher frequencies than their counterparts in AMD, where with just a series voltage in all cases DDR4-3600 could be set, but with Rocket Lake-S ... This seems to be more than limited.

Error, limitation, paradigm shift with a focus on chiplets ... We do not know at the moment what is happening with the modes and the BMI of the new generation Intel processors, but what is certain is that they are a step back in speed compared to Comet Lake-S, maximum exponent so far.

Intel Rocket Lake-S and its 11th Gen CPUs would be limited above 3733 MHz in Gear 1
Intel-Rocket-Lake-S-IMC-2

If you follow us regularly this week you will have followed the evolution of this whole issue of the BMI of the new Intel processors. Without wanting to be repetitive, the Gear 1 and Gear 2 mode will bring a queue, since as standard only the i9-11900K and KF will implement the first (1: 1 divider), while the rest of the range will arrive with Gear 2.

Although as we said we were in the middle of speculation, the performance seems to be different in both modes, more reduced in Gear 2, but that does not invalidate that Gear 1 can be manually set in BIOS like any other setting.

So far everything is correct and it shouldn't matter anymore, except for the fact that setting Gear 1 with a 1: 1 divider, the CPUs noticeably reduce their maximum frequency even with overclocking. That is, with full overclock without triggering the temperatures are already being reported values that until now could be done almost with series voltage as soon as the BMI was really good: 3733 MHz at most.

Gear 2 mode adds latency and the frequency results are not much better
Intel-Rocket-Lake-S-IMC-3

Apparently, changing from Gear 1 to Gear 2 unlocks the frequency in a few more MHz, we can possibly get DDR4-4000 in BMI that really go black, but in return we will have a consideration as in the case of AMD: latency.

Cases of non-engineering samples are already being reported where the latency skyrockets above 10 ns when switching from Gear 1 to Gear 2, even with the speed change inclusive. That is not only worrying, it is that if the values are correct we are talking about a high performance loss that places them in the best of cases at the level of AMD in this section, and in the worst behind.

On the other hand, and needless to say, other cases of improvements with new BIOS are being reported, where the latency would have dropped between 10 and 15% and the writing speed would have risen in much higher percentages, in some cases with obvious problems of + 50%. Therefore, it seems that the platform is not mature in general, since these reports are with several manufacturers, which indicates that drawing conclusions is hasty and we will have to wait for the reviews to have a more reliable scale.

This could explain the performance discrepancy that has been seen in recent weeks between leaks, which we did not publish because we already sensed that something was happening when prestigious websites were releasing less encouraging data that did not make much sense. In any case, Intel would have already hinted that, indeed, its platform would be above that of AMD, now it remains to be seen if that is indeed correct or is a product of marketing.