Funny clock speeds on Ryzen 9 5900X

mitkaa

New Member
hello everyone,

Im wondering if anyone can explain to me, why am I getting these odd clock speeds on my Ryzen 9 5900X as shown in HWinfo v6.42 on screenshot below.
And by odd numbers I mean they are not rounding up (e.g 5,074.5MHz on 1st. core).

These readings are changing when I reboot my system, sometimes the numbers are rounded up and sometimes they are off by 2 MHz.

thank you all for answers.

regards,
Mitkaa
 

Attachments

  • clock speed.jpg
    clock speed.jpg
    258.6 KB · Views: 21

Martin

HWiNFO Author
Staff member
Because they depend on (are multiples of) BCLK clock, which is not always exactly 100.00 MHz.
 

TexasAVI03

New Member
I have seen that over the last few revisions. Open and close it & see if it rounds it again.

Are you adjusting your PBO2 Curve Optimizer?
 

Martin

HWiNFO Author
Staff member
You might also try the "Snapshot Polling" option, that will use a different method to report clocks.
 

mitkaa

New Member
I have seen that over the last few revisions. Open and close it & see if it rounds it again.

Are you adjusting your PBO2 Curve Optimizer?
have tried to close and open it many times and it does not round up. Im also using negative 18 on all cores at the moment
 

TexasAVI03

New Member
ASUS motherboard by chance? Which BIOS # if ASUS?

In HWiNFO64, check your processor core list in HWiNFO64 Main>Central Processor drop-down>click on your processor. On the right column look for "Core Performance Order (CPPC)" (Listed twice, both should be same). The last two in the list, for my 5900X was core 8 & 12 (bios number will be 7 & 11). See picture.

Then go into your BIOS change those two to a 19, save & exit, get back into Windows and see if HWiNFO will round up the frequencies now.

I confirmed build v7.01-4420 that I could still make HWiNFO Sensors Core 0-11 Clock show **24.7-9, **49.7-9, or **99*.7-9 MHz variations if I changed one of the cores to an odd (or positive) while the rest of the number, in pairs were matched. In my case, a pair of even numbers on the core or paired odd numbers. Confusing, I agree, see list below.
My current ASUS PBO2 Curve is (number based upon BIOS, add +1 mentally to match what you see in HWiNFO Main but in HWiNFO Sensors the numbers will match the BIOS):
CCD 0 (per Ryzen Master): primary/best core, second best
Core 0-5 (all negative)= 18, 0, 20, 6, 20, 22
CCD 1: primary, secondary
Core 6-11 (all negative) = 22, 30, 23, 22, 25, 30
These are 2 cores I could run as odd numbers to get HWiNFO to show rounded up numbers based upon the .1MHz deviations created in the mV fluctuations in the BIOS. IF I had 1 odd or specific combination of all even numbers I would see the **24.7-9MHz through **99.1MHz on the HWiNFO Sensor page at Core 0 Clock-Clock 11 Clock. [These numbers are based upon my logging & preventing KERNEL-Power errors using HWiNFO to achieve stable 5.1MHz Boost. Your silicon lottery will not be the same as mine. However, running your strongest cores at that offset must mean you have lowered the Core Voltage of the processor. IF you have an ASUS board, on VCore Voltage Auto, you should see some outrageous core voltages in HWiNFO. Supposedly, research yourself, AMD says the 5000 series can handle it. Due to heat, power consumption & my goals for speed across all cores vs. single core speed, my VCore Voltage is manually set to 1.34375.

[Related if you have an ASUS MoBo: Did you set the negative 18 to all the cores under AiTweaker>PBO, where there is only a single option for all or Advanced>AMD OC'ing>Ok the agreement of destruction>Precision Boost Overdrive (aka PBO2)>Curve Optimizer Per Core? Are you getting KERNEL-Power errors (everyone says its a WHEA or black screen reboot) with random black screen restarts back to Windows login screen, when you go idle in Windows for a period of time?]
 

Attachments

  • HWiNFO Central Processors.JPG
    HWiNFO Central Processors.JPG
    84.1 KB · Views: 9

Zooentwerter

New Member
I have seen that over the last few revisions. Open and close it & see if it rounds it again.

Are you adjusting your PBO2 Curve Optimizer?
Same for me. I´m rounding to full MHz and sometimes the numbers are off by 1MHz +/-.
(But all at once either plus or minus)

Screenshot 2021-04-13 170403.png

When I restart the program, it is fixed most of the time.
Spectrum Spread is disabled and the bus clock is shown correctly ;-)

I use the PBO2 curve optimizer on a Asus ROG Strix X570-I Gaming Bios 3602
My offsets are also odd or even without changing the MHz on a particular core.

These are my negative offsets:

CORE#0 = 30
CORE#1 = 30
CORE#2 = 12
CORE#3 = 7
CORE#4 = 30
CORE#5 = 23
CORE#6 = 30
CORE#7 = 29
CORE#8 = 30
CORE#9 = 27
CORE#10 = 30
CORE#11 = 30
CORE#12 = 21
CORE#13 = 30
CORE#14 = 20
CORE#15 = 26

BTW: Got these values by running a AVX2 stress test for 60 sec per Core in OCCT. (Large; Extreme; Variable; AVX2)
(The latest version allows per core stress test with error detection)

Never had a KERNEL 41 after that anymore ;-)

Noticed the low offsets for my best CPPC Cores?
My voltage is set to auto and I think PBO2 already sets the VID accordingly to FIT values.
When you also offset them further you get the power resets.

You might also try the "Snapshot Polling" option, that will use a different method to report clocks.
Snapshot Polling disables the effective clock for threads, which is a bummer.
 
Last edited:
Top