How to find what blocks the single-core turbo frequency from reaching stock values (Intel TwinLake N355) ?

[alkersan]

Hello. I've recently got a little board with N355 CPU for a homelab. Overall it works fine, but I've found one disappointing pitfall - it doesn't boost beyond 3.0GHz, in single threaded workloads, despite being advertised being able to burst 3.9GHz:

Spoiler: TS Bench 1T

The CPU itself reported to be legit, and doesn't look like some ES:

Spoiler: Intel Processor Identification Utility

The turbo table looks okay, x39 is there (cross checked in BIOS - values are the same):

Spoiler: Turbo multipliers

I though it could be due to power limits, like PL1/PL2, and managed to increase them from the default 9W:12W to 45:45. But that only changed the total power consumption during prolonged multithreaded workloads, apparently because now cores could sustain 3GHz for longer. So this isn't what I was looking, single-threaded max frequency is still capped at 3GHz

Spoiler: HWInfo & 7Zip Test

I suspect that something else is limiting the boost. What it can be?

Thank you

 

[:D:D]

Perhaps this thread would be better served under Throttlestop thread? Any admins concur?

 

[ShrimpBrime]

Looking at the screen shots, Power Limit 2 is set to 12w. Try 15w

PL1 is minimum 9w and PL2 by default is 12w, CPU rating 15w

 

[alkersan]

Looking at the screen shots, Power Limit 2 is set to 12w. Try 15w

PL1 is minimum 9w and PL2 by default is 12w, CPU rating 15w

Yeah, I've tried increasing PL limits to different values, all way up to 40w and 40w. But that didn't affect the maximum boost frequencies, neither on single-core nor multi-core. CPU still was capped at 3GHz in single-threaded workload.

 

[ShrimpBrime]

Yeah, I've tried increasing PL limits to different values, all way up to 40w and 40w. But that didn't affect the maximum boost frequencies, neither on single-core nor multi-core. CPU still was capped at 3GHz in single-threaded workload.

Looks like the EDC is limited on the board. 14.219 amp peak. If the best performance is at defaults, I'd just use that. It's going to be seemingly warm, but the heatsink is small, and I'm curious if the are using thermal pads or paste to sink the chip. But maybe a fan on it couldn't hurt, might boost better. More heat means leakage. I'd try a physical approach.

 

[alkersan]

Seems it's something related to the disabled C-states, as suggested here. Just started to look into that