Wrong bus speed and multipliers for AMD Phenom II 940 about libcpuid HOT 9 CLOSED

@x0rg - any ideas?

from libcpuid.

@kbios: recent changes on RDMSR were to fix that.
With stock clocks, the bus clock should be 200MHz and CPU multiplier x15 (because the stock clock is 3GHz). So yes, this is very buggy. 😢

Please paste cpuid_tool --rdmsr and cpuid_tool --rdmsr-raw outputs as root.
Your CPU is overclocked at 3,4GHz, right?

from libcpuid.

# ./cpuid_tool --rdmsr
  cur. multi.: 1700 MHz
  max. multi.: 1700 MHz
  core volt. : 1.35 Volts
  base clock : 853.75 MHz
# ./cpuid_tool --rdmsr-raw
CPU is AuthenticAMD AMD Phenom(tm) II X4 940 Processor, stock clock is 3415MHz.
msr[0xc0010061]=00 00 00 00 00 00 00 30 
msr[0xc0010063]=00 00 00 00 00 00 00 00 
msr[0xc0010064]=80 00 00 1a 20 00 20 12 
msr[0xc0010065]=80 00 01 c3 30 00 30 07 
msr[0xc0010066]=80 00 01 b6 40 00 40 02 
msr[0xc0010067]=80 00 01 72 40 00 40 40 
msr[0xc0010068]=00 00 01 72 40 00 40 40 
msr[0xc0010069]=00 00 01 72 40 00 40 40 
msr[0xc001006a]=00 00 01 72 40 00 40 40 
msr[0xc001006b]=00 00 01 72 40 00 40 40

Exactly, I overclocked to 3.4 by setting the multiplier to 17

from libcpuid.

Ok, I'm working of this. With my local changes, the computed bus clock is 200,88MHz.
I'll push a fix in a few moment. This issue is related to used method to determine how many Pstate are presents.

from libcpuid.

Thanks for fixing this, @x0rg!

from libcpuid.

@kbios: if you can, please update libcpuid to latest git revision and re-paste the cpuid_tool --rdmsr output to confirm fix.

@anrieff: NP. I'm now watching libcpuid, I will be notified of all conversations, I think it will be helpful for future issues related to MSRs. ;)

from libcpuid.

Wow that was fast! I can confirm that with the latest commit the bus speed is correct:

# cpuid_tool --rdmsr
  min. multi.: 0
  cur. multi.: 0
  max. multi.: 0
  core volt. : 1.35 Volts
  bus clock  : 200.88 MHz

from libcpuid.

And it shows another bug with CPU multipliers...

from libcpuid.

Maybe the %d format, can be replaced by a %.2f. Interesting bug.

EDIT: Done in 660cde1.

from libcpuid.