BCLK OC not reflected in some readings

[Animag771]

I recently overclocked my 5700X to run at 102 BCLK but the only readings that show the change in HWiNFO are Bus Clock (102) and Core Clock (4,947 max). None of the other readings seem to reflect the change.

Core Ratio (max) = 48.5 (should be 49.4 or 49.5 rounded up)
FCLK = 1,767 (should be 1,802)
UCLK = 1,767 (should be 1,802)
L3 Clock (max) = 4,850 (should be 4,947)
Frequency Limit Global = 4,850 (should be 4,947)

All of these readings appear to be using a BCLK of 100 instead of 102 for their multiplier.

Am I crazy? What am I missing here?

 

[Martin]

Those values come from internal SMU metrics that sometimes might not reflect the actual BCLK, especially when the eCLK mode is active.

 

[Animag771]

Thank you for your response Martin!

Is there a way around this? Before I overclocked the BCLK I was able to use the L3 Clock to determine if I was clock stretching under single-core loads because the L3 actually showed the effective clock (I think) but now I have no idea if I'm clock stretching under single-core loads.

Also is there a reason these values are derived directly from the SMU instead of being multiplied by the BCLK measurement within HWiNFO? I'm sure there is a reason but now I'm curious to know why?

 

[Martin]

Those values are read from the SMU because there's no other way to read them.
Why don't you rely on the Effective Clock?
You could also setup a custom multiplier in HWiNFO settings for the L3 clocks based on actual BCLK. So for 102 MHz, set it to 1.02

 

[Animag771]

Those values are read from the SMU because there's no other way to read them.
Why don't you rely on the Effective Clock?
You could also setup a custom multiplier in HWiNFO settings for the L3 clocks based on actual BCLK. So for 102 MHz, set it to 1.02

I realize they are read from the SMU but I was wondering why they aren't multiplied by the BCLK after being read. This would make it accurate even if the SMU readout is incorrect. Although you're much more knowledgeable of how these things work, this was just how I thought it would work.

Effective clock doesn't work for single-core loads, at least not for me. It shows something like 300MHz effective when a single core is running at 4,947.

I feel stupid.... I didn't even think about setting a custom multiplier for the L3 clocks. At least then I'll be able to see the true clock speed in the RivaTuner overlay.