14700K CB R23 scores curve vs. power limits between 15W and 260W

charonme

charonme

Member
Joined
Nov 4, 2023
Messages
169
i7-14700K, IccMAX=307A, MSI LLC4, AC_LL=52, DC_LL=32, offset=-0.15V, CEP=on
4minute CB R23 scores while running intel XTU and HWinfo64, performance power profile on win11, normal priority
i7-14700K_PLvsCBR23_4min_MSILLC4_AC52DC32-0.15V_CEPon_IccMAX307A_TVB_AutoRing_aug2024.png

Above 210-230W the IccMAX and TVB begins to take effect and starts to flatten the curve more and at 270W PL I got the same scores as 260W because of thermal throttling with my cooling setup.
With CEP=off the scores were consistently around 326 points lower
 
Sort by date Sort by votes
Excellent work! It clearly illustrates what we already thought we were observing, which was a quickly diminishing efficiency as PL2 exceeded approximately 180W-200W. [Interesting that Intel Baseline=188W for the i7’s.] I think your graph proves the point - and at even lower PL2.

Do you have any thoughts on why CEP enabled scores would be higher? I had posed a question earlier in the week that perhaps CEP affects recovery time (positively or negatively) when bouncing of the power limiters. But I had no way of proving it.
 
Last edited:
Upvote 0 Downvote
thanks! No idea about CEP, it's even possible it just underestimates the power measurements for some reason, so without CEP it PL1-throttles sooner, but the difference is too tiny for me to measure with a clamp ammeter on the EPS cable. Maybe someone with an ElmorLabs power meter could try :) Also there might be AC_LL and PL settings that could make that difference larger, maybe I'll try to look for some.
It could also be some bug that will go away once I update to a newer bios and microcode (maybe someone with a newer bios or microcode could try to replicate my results?)
 
Last edited:
Upvote 0 Downvote
I'm now running latest 129 bios. (7D91vHD) - my previous tests were with the v2 beta 129 bios.

I tried your settings:

i7-14700K, IccMAX=307A, MSI LLC4, LLC AUTO, AC_LL=52, DC_LL=32, offset=-0.15V, CEP=on
With PL1 and Pl2 at 253W

CB R23 (4 mins) was 22000!

Just to make check I'd not inadvertently changed another setting. I switched to

i7-14700K, IccMAX=307A, MSI LLC4, LLC AUTO, AC_LL=52, DC_LL=32, CEP=off
With PL1 and Pl2 at 253W

CB R23 was 33635

Note Undervoltage protection is off, and Enhance Turo is disabled, everything else is default. I'm not using XTU.
 
Last edited:
Upvote 0 Downvote
whoa interesting, what board is that? Mine is PRO Z790-P wifi DDR4 and an old 2023 bios and microcode
I believe AC_LL 52 should be pretty high enough at MSI LLC4 not to trigger CEP, even AC_LL 18 wasn't triggering it on mine at most power limits
But it's quite possible they changed the CEP behavior with newer microcodes
 
Upvote 0 Downvote
charonme said:
whoa interesting, what board is that? Mine is PRO Z790-P wifi DDR4 and an old 2023 bios and microcode
I believe AC_LL 52 should be pretty high enough at MSI LLC4 not to trigger CEP, even AC_LL 18 wasn't triggering it on mine at most power limits
But it's quite possible they changed the CEP behavior with newer microcodes
Click to expand...
I'm on an MAG-Z790-TOMAHAWK-WIFI from November 2023, DDR5.

i should rerun my tests - since the results I posted in the other thread were with the last beta bios.. This seems to suggest they could have changed something with the final bios?
 
Upvote 0 Downvote
I'd made an error previously..... LLC had not been changed from auto. So with LLC to 4 - I match your graph on the newest bios/microcode.

i7-14700K, IccMAX=307A, MSI LLC4, AC_LL=52, DC_LL=32, offset=-0.15V, CEP=on
With PL1 and Pl2 at 220W

CB R23 (4 mins) was 34060.
 
Upvote 0 Downvote
Back
Top