search

ThomasKaiser / sbc-bench

Simple benchmark for single board computers
BSD 3-Clause "New" or "Revised" License
627 stars 78 forks source link

Inquiry about ATF on Rockpro64 and freqs #76

Closed neofeo closed 10 months ago

neofeo commented 10 months ago

Hi TKaiser, I am a bit disappointed with my Rockpro64 on armbian lying to me even with ATF enabled. Could you explain me why this happened??

Screenshot_20230813-182151_Chrome

ThomasKaiser commented 10 months ago

Where's the ix.io link for this?

neofeo commented 10 months ago

http://ix.io/4DnZ

ThomasKaiser commented 10 months ago

That's the culprit:

vdd_cpu_b: 1288 mV (1500 mV max)

vs.

2400 MHz   1290.0 mV

Voltage feeded to the A72 cores is limited to 1288 mV as such every DVFS OPP that exceeds this won't be activated.

Asides that the OPP table is weird anyway (the voltage increases should happen in tiny steps at low frequencies and have to get the larger the higher the frequency (see rk3399-4.4.211-FriendlyElec_NanoPi_M4.txt for example or any other vendor OPP table there) and this is the wrong place to ask about device-tree basics especially if it's about something that weird as 'overclocking'.

neofeo commented 10 months ago

Thanks for the feedback!

neofeo commented 10 months ago

I changed my entire opp tables, not they are a lot better, very stable at 2.2/1.8 , didn't noticed that seems thst only the first microvolt valye its being taken in consideration. My low power cores were trying to reach 1.8 with 1.2v and that wasn't enough, it was required 1.25 or so. Also, who manage that limit on the voltages, bc it mentions maximun values but at the end it limits them way lower, I am not planning to go to much farther than 1.3v, it's risky, I know, but I will need 1.31 or 1.32 to reach 2.4 ghz.