Things get particularly interesting again with LinX, which unlike LinpackXtreme also benefits from the E-Cores in terms of performance. Both benchmarks are based on the same Intel Math Kernel Library, but the differences in the implementation of the calculation tasks are decisive here.
Again, the P-cores alone are far thirstier than in the company of the E-cores or with an active AVX-512. But unlike in the Y-Cruncher, the latter are actually the most economical here overall.
However, when it comes to performance measured in floating point operations per second, the E-cores can show their full potential and provide for an increase in computing power of 26%. These are roughly similar ratios as you can also achieve in the readily taken Cinebench R23. While the P cores alone can still benefit noticeably from AVX-512, they are nowhere near closing the gap to Stock.
Converted to the power consumption, it looks much better for the Sunny Cove P cores, even if they don’t quite reach the level of the factory configuration. Here, however, it becomes once again impressively clear what the AVX-512 instruction set alone can do for certain use cases – after all, 27% higher efficiency.
Final thoughts
Whether Alder Lake CPUs are more efficient in the standard configuration or only with P cores for AVX-512 depends very much on the use case. And that the hybrid approach with different cores is still relatively new and immature is also no secret. In time, many improvements can be expected, especially through software updates from Intel and Microsoft. But originally the trigger for today’s article was the literally incredibly low power consumption of the P-cores with AVX-512 and the question whether the efficiency is actually higher. And indeed, with the feature enabled, the efficiency of the P-cores is significantly higher in all benchmarks than without. In fact, the results are so clear that the instruction set can and should always be safely activated – if possible, of course.
It’s extremely impressive how AVX-512 has evolved over the years, from a flaming inferno with little extra power on X299 to an absolute efficiency secret on Z690. It’s no wonder, then, that Intel’s AVX team was anything but thrilled with the alleged planned removal of the feature for Alder Lake. Because there is no doubt that a lot of brainpower and engineering skills of Intel employees have gone into this. So if the same or more performance can be gained from the Sunny Cove cores alone, what are the Gracemont E cores good for anyway? Sure in some use cases like Cinebench or LinX they bring the necessary percentage points to place themselves on a i9-12900K ahead of a 5950X from AMD, but at what price?
Because the silicon area must also be paid, by Intel, but ultimately also by the customer, who also buys the e-cores, whether they wants them or not, not to mention the lower cache or ring clock. The other way around, you could ask why Intel can’t offer AVX-512 on the P cores with active E cores as well. Wouldn’t that be the perfect use for the Intel Thread Director – to distinguish whether a process needs AVX-512 or not and assign it to the right cores accordingly?
As is often the case, more questions remain unanswered at the end of the day than before. But you have to ask them if you’re looking for answers, especially if this peculiarity of the new Intel architecture seems to have been completely overlooked in the hype surrounding the launch so far. These are undoubtedly exciting and special times in the hardware industry, there is a lot to unpack and it remains exciting. So keep your eyes open on our website for upcoming Intel Alder Lake overclocking guides and DDR5 reviews.
49 Antworten
Kommentar
Lade neue Kommentare
1
Veteran
Veteran
Mitglied
Mitglied
1
1
Urgestein
Urgestein
1
Mitglied
Mitglied
Urgestein
Veteran
Veteran
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →