CPU SPEED

NoahPhense · Post by **NoahPhense** » Sat Dec 30, 2006 7:07 pm

I know this topic has been hit A LOT of times. But I need to address it
again. Here is some code: (questions below)

Procedure.l GetCpuMhz() 
  Global int64val.LARGE_INTEGER 
  !FINIT 
  !rdtsc 
  !MOV dword [v_int64val+4],Edx 
  !MOV dword [v_int64val],Eax 
  !FILD qword [v_int64val] 
  Delay(1000) 
  !rdtsc 
  !MOV dword [v_int64val+4],Edx 
  !MOV dword [v_int64val],Eax 
  !FILD qword [v_int64val] 
  !FSUBR st1,st0 
  int64val\highpart=0 
  int64val\lowpart=1000000 
  !FILD qword [v_int64val] 
  !FDIVR st0,st2 
  !fistp qword [v_int64val] 
  ProcedureReturn int64val\lowpart 
EndProcedure 

; MessageBox_(0, Str(GetCpuMhz()*2), "CPU Speed", #MB_ICONINFORMATION)

Debug GetCpuMhz()*2
Delay(1000)
Debug GetCpuMhz()*2
Delay(1000)
Debug GetCpuMhz()*2
Delay(1000)
Debug GetCpuMhz()*2

End

I don't get steady results on every run. Is there something better?
Are there different formulas out there for calculations?

Like:
- grabbing cpu speed from bios
- grebbing cpu speed from windows
- realtime calculation

- np

Trond · Post by **Trond** » Sat Dec 30, 2006 7:31 pm

http://www.dotnetheaven.com/Uploadfile/ ... 55911bf856

NoahPhense · Post by **NoahPhense** » Sat Dec 30, 2006 7:46 pm

That's interesting. But it doesn't help with laptops that control processor
speeds.

I need an accurate realtime processor speed.

- np

netmaestro · Post by **netmaestro** » Sat Dec 30, 2006 7:49 pm

Getting the CPU speed from the registry is safest. Code similar to the proc you posted won't give reliable results on multicore machines, but if you look at the registry key

HKEY_LOCAL_MACHINE\HARDWARE\DESCRIPTION\SYSTEM\CentralProcessor\0

you will get good information. Then you can check

HKEY_LOCAL_MACHINE\HARDWARE\DESCRIPTION\SYSTEM\CentralProcessor\1

to see if there's another core.

Trond · Post by **Trond** » Sat Dec 30, 2006 8:24 pm

First of all, you probably don't. Second, if you use the processor the speed may change.

This gives the correct result here, but it's not perfectly accurate.

Code: Select all

Procedure.q GetCycleCount()
  !rdtsc
  ProcedureReturn 
EndProcedure

Procedure CpuSpeed()
  Protected A.q
  Protected B.q
  A = GetCycleCount()
  Sleep_(100)
  B = GetCycleCount()
  ProcedureReturn (B-A) / 100000
EndProcedure

Debug CpuSpeed()

Obviously the longer you wait, the more accurate.

TerryHough · Post by **TerryHough** » Mon Jan 01, 2007 4:15 pm

@NoahPhense

Did you review topic
http://www.purebasic.fr/english/viewtopic.php?t=3811 ?

I hadn't looked at this since 2003, but the routine I use is based on the topic above and seems to be very consistent. (Jack's version)

For example, 20 sequential checks on this machine all indicate 1784 Mhz.

The registry reports 1785 Mhz, and the My Computer properties indicate 1.79 Ghz.

netmaestro · Post by **netmaestro** » Mon Jan 01, 2007 4:37 pm

This should be as accurate as any:

Code: Select all

Procedure GetCPUSpeed()
  QueryPerformanceFrequency_(@ulFreq.q) 
  ProcedureReturn ulFreq/1000000
EndProcedure

Debug GetCPUSpeed()

akj · Post by **akj** » Mon Jan 01, 2007 6:49 pm

Netmaestro's code does not work on my 300MHz PC as it returns a value of 1.

Here's another routine that does work for me:

Code: Select all

Procedure.l CpuSpeed()
; Returns CPU speed in MHz
Protected A.q, B.q
!RDTSC
!MOV dword [p.v_A+4],Edx ; A hi
!MOV dword [p.v_A],Eax ; A lo
Sleep_(100)
!RDTSC
!MOV dword [p.v_B+4],Edx ; B hi
!MOV dword [p.v_B],Eax ; B lo
ProcedureReturn (B-A)/100000
EndProcedure

netmaestro · Post by **netmaestro** » Mon Jan 01, 2007 6:57 pm

Some older processors don't support QueryPerformanceFrequency, for those it won't work, true enough.

Azul · Post by **Azul** » Tue Jan 02, 2007 3:07 am

netmaestro wrote:Some older processors don't support QueryPerformanceFrequency, for those it won't work, true enough.

How new it should be?

I got '3' with AMD Athlon(tm) XP 2400+ (2GHz)

update: WinXP SP2

netmaestro · Post by **netmaestro** » Tue Jan 02, 2007 3:19 am

Not sure about specific hardware, which configurations support high-resolution frequency counters and which don't, but when you do:

Code: Select all

Result = QueryPerformanceFrequency_(@ulFreq.q)

if you find a non-zero value in Result, the high-resolution frequency counter is supported by your hardware, if it's zero your hardware doesn't support it. I'm not too strong in hardware, so the support for this could be coming from the motherboard / bios rather than the processor, I dunno. If somebody knows, a post would be appreciated.

CadeX · Post by **CadeX** » Tue Jan 02, 2007 3:48 am

netmaestro wrote:Some older processors don't support QueryPerformanceFrequency, for those it won't work, true enough.

My processor is quite insulted by that, seeing as its an AMD 64bx2 2.2GHZ Core processor. Amd Hammer Family, That is not old.

I also get a "3" returned.

I did notice that everyone with issues of that code has an AMD.

Derek · Post by **Derek** » Tue Jan 02, 2007 12:53 pm

I got 1862 with my dual core core2 which runs at 1.86 per core, but as was mentioned, it's not an AMD.

Flype · Post by **Flype** » Tue Jan 02, 2007 1:55 pm

netmaestro wrote:Some older processors don't support QueryPerformanceFrequency, for those it won't work, true enough.

i have AMD AthlonXP 2800+ (2.08 Ghz) and QueryPerformanceFrequency returns here 3.

rsts · Post by **rsts** » Tue Jan 02, 2007 2:32 pm

Of those posted, Trond's give 1799 on my amd athlon 1.8 and NoahPhense
3600 or 3599.

Neither of the others seem to work properly. netmaestro returns the expected amd 3 and I get an assembler undefined symbol on akj's.

No complaints. just reporting my experience in the interest of science

cheers