There are three opteron processors with the cpuid that I paste below:
model name??? : AMD Opteron(tm) Processor 6136 (magny core, 8)
cpu MHz??? ??? : 800.000
cache size??? : 512 KB
bogomips??? : 4800.20
model name??? : AMD Opteron(tm) Processor 6174 (magny core, 12)
cpu MHz??? ??? : 800.000
cache size??? : 512 KB
bogomips??? : 4400.13
model name??? : AMD Opteron(TM) Processor 6274? (interlagos, 16)
cpu MHz??? ??? : 1400.000
cache size??? : 2048 KB
bogomips??? : 4399.97
As you can see, although interlagos is the newest one which has higher cpu
clock rate and higher cache size, the bogomips of interlagos is the lowest!
I read that bogomips is not a true measurement, but it senses as a speed meter.
Any idea?
~MN
Mahmood Naderan wrote:
> As you can see, although interlagos is the newest one which has higher cpu
> clock rate and higher cache size, the bogomips of interlagos is the lowest!
> I read that bogomips is not a true measurement,
Indeed it is not.
> but it senses as a speed meter.
It measures the speed of one particular delay loop. Because such a busy
loop is pretty much the silliest thing you could do to a CPU, it does
not make sense to use it to measure overall CPU speed, or to compare
different microarchitectures.
> Any idea?
If you want to know how fast your code will execute, you have to measure
that code. If you are too lazy to do this, search for a benchmark that
represents that code well; for general-purpose code, the gcc part of the
SPEC CPU benchmark probably comes closest.
Regards,
Clemens
Thanks
// Naderan *Mahmood;
----- Original Message -----
From: Clemens Ladisch <[email protected]>
To: Mahmood Naderan <[email protected]>
Cc: "[email protected]" <[email protected]>
Sent: Saturday, June 16, 2012 1:50 PM
Subject: Re: about interlagos
Mahmood Naderan wrote:
> As you can see, although interlagos is the newest one which has higher cpu
> clock rate and higher cache size, the bogomips of interlagos is the lowest!
> I read that bogomips is not a true measurement,
Indeed it is not.
> but it senses as a speed meter.
It measures the speed of one particular delay loop.? Because such a busy
loop is pretty much the silliest thing you could do to a CPU, it does
not make sense to use it to measure overall CPU speed, or to compare
different microarchitectures.
> Any idea?
If you want to know how fast your code will execute, you have to measure
that code.? If you are too lazy to do this, search for a benchmark that
represents that code well; for general-purpose code, the gcc part of the
SPEC CPU benchmark probably comes closest.
Regards,
Clemens