Multi Core VS Single Core Effective Clock

[Ryeleigh]

So I undervolted my Ryzen 5 5500 and found that when doing a quick test with Cinebench R23 with Multi core test my Effective Clock can stay within 10 MHz difference, but when doing single core test, the effective clock go to more than 100 MHz difference, why does it differ so much? does it mean its core stretching and I should lower my undervolt settings?

 

I use HWInfo 64 with Snapshot CPU Polling to monitor the clock speed.

 

I use ASUS EX A320M Gaming so I use VDDCR CPU Voltage Offset as the method to undervolt.

I know I shouldn't use A320 Boards for newer chip but I'm on a strict budget, don't even have the budget to buy 5600 (was previously on 2200G), so it is what it is.

 

Thanks in advance for any response.

[jaslion]

I mean thata just how boost clocks And power budgets work? when doing a single core test only a single core is really loaded so they normally will boost higher as there is power budget left.

 

Modern cpu's have multiple performance states where depending on x amount of cores being loaded by x% it will boost higher or lower to stay within spec.

 

So basically working as intended.

[Ryeleigh]

7 minutes ago, jaslion said:

I mean thata just how boost clocks And power budgets work? when doing a single core test only a single core is really loaded so they normally will boost higher as there is power budget left.

 

Modern cpu's have multiple performance states where depending on x amount of cores being loaded by x% it will boost higher or lower to stay within spec.

 

So basically working as intended.

sorry, my fault for not being clear, because I undervolted what I mean with 100 MHz Difference is, the effective clock actually run 100 MHz lower than what the core clock (perf) reported

 

so Core Clock (perf) reported 4,250 MHz

Core Effective Clock reported 4,245 MHz on all core when doing multi core test

and reported 4,100 MHz on the particular core that doing the work when doing single core test

[jaslion]

3 minutes ago, Ryeleigh said:

sorry, my fault for not being clear, because I undervolted what I mean with 100 MHz Difference is, the effective clock actually run 100 MHz lower than what the core clock (perf) reported

 

so Core Clock (perf) reported 4,250 MHz

Core Effective Clock reported 4,245 MHz on all core when doing multi core test

and reported 4,100 MHz on the particular core that doing the work when doing single core test

I mean that can happen with undervolting. It is overclocking after all. So oddities can happen.

[Ryeleigh]

7 minutes ago, jaslion said:

I mean that can happen with undervolting. It is overclocking after all. So oddities can happen.

yes, I'm aware, hence my question.

Quote

does it mean its core stretching and I should lower my undervolt settings?

Because to my knowledge, there's a limit when it is considered not stable and not recommended to keep that particular settings.

 

it's especially concerning because single core is lower than multi core when usually the opposite happen like what you said on your first reply.

[jaslion]

1 hour ago, Ryeleigh said:

yes, I'm aware, hence my question.

Because to my knowledge, there's a limit when it is considered not stable and not recommended to keep that particular settings.

 

it's especially concerning because single core is lower than multi core when usually the opposite happen like what you said on your first reply.

Whats the voltages in multi? Could be its getting some extra juice which allows the cpu to go a bit higher.

[Ryeleigh]

11 hours ago, jaslion said:

Whats the voltages in multi? Could be its getting some extra juice which allows the cpu to go a bit higher.

yeah, the voltages when doing multi is higher

[Ryeleigh]

So I just decided to not doing any undervolt at all because I found the behavior is so random even when using minimal amount of offset.

 

One time doing below 10 MHz difference, then next few restart the difference is 50 MHz, wait a few hours then test again it is still 50 MHz, restart again the difference is 30 MHz, and now the difference is almost 100 MHz again all while using the same offset. 

[TatamiMatt]

6 hours ago, Ryeleigh said:

So I just decided to not doing any undervolt at all because I found the behavior is so random even when using minimal amount of offset.

 

One time doing below 10 MHz difference, then next few restart the difference is 50 MHz, wait a few hours then test again it is still 50 MHz, restart again the difference is 30 MHz, and now the difference is almost 100 MHz again all while using the same offset. 

Instead of wondering about clock speeds and voltages, undervolt your CPU and run a cinebench for 10 mins, then change undervolt and retest, choose the higher score as it will mean higher performance.

 

Are you manually undervolting or running a PBO offset curve?

[Ryeleigh]

4 minutes ago, TatamiMatt said:

Are you manually undervolting or running a PBO offset curve?

A-type board doesn't have PBO so it's manually using VDDCR CPU Voltage Offset