Processors

[brucer]

This video is during a geekbench 4 benchmark, does the same during 3dmark Timespy and Firestrike runs.. Why is it dropping to 3.5ghz at times when the base frequency is 3.9ghz? 

Yea I guess it could be from not having any load on that particular core, but something doesnt seem correct to me.... If you look at the utilizations of other cores, there is zero utilization on other cores that are at 3.8ghz, which is still below base frequency, and also some at or above base frequency without any utilization on them..  I'm using cpuid hwmonitor to monitor.. in 3dmark Timespy and Firestrike I use afterburner and its osd..  I'd use Ryzen Master to monitor but its useless when running benchmarks such as Timespy, Firestrike and Superposition.

sorry about the music, I just picked soemthing, better than listening to background noise

[brucer]

I've also used Hwinfo64 to monitor with the same results..

[ajlueke]

Ryzen processors all depend on the c-states like cc6 sleep.  AMD felt it was more efficient to just put cores to sleep rather than run them awake at low frequencies/voltages.  So it runs at a high p2 of 2.2GHz and just sleeps below that. 

This makes the core "invisible" to task manager and third party monitoring applications, as any probe of the clockspeed would wake the core an eliminate the power savings.  Those applications just display the last reported clock speed.  So even if a core displays 2.2 or even 4 GHz but it may not actually be awake.

[brucer]

Then why are cores showing 3.8ghz and 4.25ghz with 0% utilization on them while other cores show 3.5ghz with 0% utilization on them, also with voltage usage being shown on the cores while running the test?

[ajlueke]

A decent read on this topic.

https://www.reddit.com/r/Amd/comments/cbls9g/the_final_word_on_idle_voltages_for_3rd_gen_ryzen/

[brucer]

a decent watch, just push play Radeon 7 “100%” End of Life, Says Puget Systems | AMD Reduced Ryzen 3000 Stock Boost Clocks - YouTub... 

[ajlueke]

There has been quite a bit of discussion on a few forums on this topic.  The consensus seems to be that all the AGESA 1.0.0.3XX variants do not boost to the level that 1.0.0.2 did.  Why?  I don't think we've had an official statement on the matter.  Did AMD find that the old boost behavior was, in fact, unsafe?  Unsure, but something has definitely changed.

[brucer]

did you watch the video? start at about 6:50 into it, if its not already there..

[shinkojiro]

Something Shamino, an ASUS bios engineer, says does not count as an official statement from AMD on the matter. "AMD was being too aggressive with their boosting previously" is his opinion. 

[shinkojiro]

Someone from AMD does need to give updates on whats going on and whats going to be done about the boost clocks. Even tho, I know its just going to be an agesa update that fixes it, they should do a bit more to acknowledge the issue.

About the base clock... youre never likely to see 3.9ghz from a core in actual use unless youre cooling is utter rubbish. The processors are acting more like Modern GPU's with clock frequencies, where they will boost above that base clock very often even with all-core load. The low frequencies are the inactivity downclocks, which 3rd party reporting software cannot report accurately nor report the sleep state. The only reporting software that can show the processors actual state and frequency is Ryzen Master.