[v7.62-5200] Tiny CPU core frequency diffirence

[T[]RK]

Hello Martin!

First i would like to thank you such amazing awesome tool. Without it i feel handless and blind.

PC System:

Hardware:
CPU: AMD Ryzen 7 5800X3D [BOX version];
TiM: Arctic MX-4 (not fresh, not old);
CPU cooler (air): Phanteks PH-TC14PE (Dual Tower, 140mm);
Motherboard: ASUS ROG Crosshair VIII Dark Hero (BIOS v4602);
RAM: 4x 8GB DDR4-4133 Team Group Xtreme
GPU: EVGA GeForce RTX 3080 Ti FTW3 Ultra Gaming;
SSD1: 500GB Samsung SSD 980 Pro (Top M.2 slot);
SSD2: 16GB Intel Optane (Botton M.2 slot, just for garbage flow, page file, chrome cache)
HDD: 500GB Western Digital Caviar Blue (SATA);
PSU: SeaSonic PRIME Ultra 850 Titanium (SSR-850TR)
Software:
OS: Microsoft Windows 11 Pro 22H2 (build 22621.2283) downloaded from Microsoft directly + hash tested before install.
AMD Chipset driver: v5.08.02.027
NVIDIA GPU driver: v537.24
Windows power plan: High Performance
BIOS settings: Default

So, i notice that when i use Boost Tester (which cycle load on CPU cores) i can't get it's maximum frequency for single core - ~4 550 MHz (with disable Spread Spectrum and BCLK = 99,98 MHz it's more close to 4 549 MHz, but anyway).

What i see in HWiNFO64 (with enable "Snapshot CPU Polling" and Polling Period = 500ms by the way) that maximum effective clock around 4 536-4 537 MHz.


Then i close HWiNFO64 and started Ryzen Master and enable Histogram to record peak values. And they are diffirent - 4 548-4 549 MHz. Really close to what should be.



Again, default BIOS settings. Zero overclock\undervolt. Fresh Windows install from Microsoft site. I think i only disable Hyper-V with Blue Stack file - "HD-DisableHyperV_native_v2", but i did check Hyper-V after install windows - it was disable anyway and result before that also was similar.

So, something interfere with CPU?

Also, should i upload Debug Report from HWiNFO64 after Boost Tester (1 pass for 1 core)? It's about 11 Mb for DBG file.

 

[Martin]

That difference seems to be due to SSC which is indicated by BCLK = 99.8 MHz. Maybe disabling SSC in BIOS doesn't work? Core ratios report x45.5 so if BCLK would be 100 MHz, it should match the expectation.
I don't know how Ryzen Master detects the clocks, it might be rounding them up per 100 MHz BCLK.

 

[T[]RK]

That difference seems to be due to SSC which is indicated by BCLK = 99.8 MHz

Nope. Tested in CPU-Z. Clocks monitor showed me higher clock - 4540.80 MHz. Still higher then HWiNFO64. For some reason...

Maybe disabling SSC in BIOS doesn't work?

In Default BIOS "SB Clock Spread Spectrum" is enabled. Here is with disabled.

 

[Martin]

There BCLK=100.0 MHz, so as expected.
With enabled SSC it's 99.8 also as expected.
I don't see any problem.

 

[T[]RK]

I don't see any problem.

As i said, it's tiny.

Here is two more screenshots:

HWiNFO64 with Boost Tester (Spread Spectum disabled):

Maximum effective clock: 4 546.5 MHz

HWiNFO64 with CPU-Z Clocks monitor and with Boost Tester (Spread Spectrum disabled):

Maximum effective clock: 4 550 MHz

I think it's because BCLK start fluctuate and i saw how it raise a bit in CPU-Z from 99,98 to 100,1+. But HWiNFO64 didn't show it.

I also suspect that two of them together may generate wrong data like so:

 

[Martin]

With "Snapshot CPU Polling" enabled, effective clocks are completely calculated by the CPU using an internal mechanism.

 

[T[]RK]

Martin! It took THAT long time, but i find reason why effective clock was lower!

It was SATA HDD connected to motherboard. After i disconnected it and run test again - i got proper results.