www.asus.com
I have some further data points on this. With an NH-D15 the board could deliver enough wattage but thermal limiting reduced the synthetic benchmark scores. My 13900K scored 2% better at 235 watts than at 275 where the CPU hit 99 degrees quickly and throttled. Throttling also occured at 235 but not as quickly. The R23 score of 31584 at 195 watts was nearly identical to the 275 watt score while drawing 100 watts less from the wall.
In BIOS v2.0 I can see difference ~20 degrees of C between PCH temperature reported in BIOS and in SuperDoctor.
How about something in the middle for performance per temp: leave e-cores enabled but disable hyperthreading? Depending on how much your use case will or won't benefit from HT of course.
My reason for leaving the E cores disabled is my primary OS is going to be Windows 10 LTSC IoT edition which has limited ability to schedule properly from what I understand. I think I read Windows 10 won’t be getting any more scheduler improvements and this is based on a slightly older version of Windows 10 as well. It would probably be fine but my main reason for upgrading this system in the first place is single core performance so I really don’t need the E cores anyway.
That’s a pretty terrible score compared to 38000+ on the reviews I saw. Also I managed a 34451 with 999 watt limit set for Pl1/2. Cooler was a NH-D12L with the single included fan but I did have another fan blowing on the VRM area.
I haven’t seen a single review under 38000 so a delta that wide (~15%) seems unreasonable to me. And the X13SAE doesn’t have any options for manually adjusting the CPU voltage in BIOS.
i don't think there is an automatic way. the scheduler wants also not to priorises one single core that then is early worn out.
www.intel.com
I don't know much about linux but if it's like most things I do know about, it depends on the distro. According to this, Ubuntu supports them pretty well and this was from February.
they are pretty good for background tasks. with alder -lake its the first time processors have two types of cores...
This is true--it's in the 0th percentile of results. Perhaps running Server 2022 is part of the problem. Oddly, the single-core r23 benchmarks are slightly better than average and thankfully that's most important for my workloads.
with POR enabled and two DIMMs per channel i expect this downgrade. maybe older BIOS has POR disabled by default.
I finally found the menu where I could change the DIMM speed from auto to a max of 4400. That was good for an extra 3.5% on my multi-core score.
