Re: athlon 64 + FC4 == SLOW
From: General Schvantzkoph (schvantzkoph_at_yahoo.com)
Date: 08/29/05
- Previous message: cmk128_at_hotmail.com: "athlon 64 + FC4 == SLOW"
- In reply to: cmk128_at_hotmail.com: "athlon 64 + FC4 == SLOW"
- Next in thread: cmk128_at_hotmail.com: "Re: athlon 64 + FC4 == SLOW"
- Reply: cmk128_at_hotmail.com: "Re: athlon 64 + FC4 == SLOW"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sun, 28 Aug 2005 18:01:13 -0400
On Sun, 28 Aug 2005 06:17:27 -0700, cmk128 wrote:
> Hi
> I am using athlon 64 3400+ (2.4GHz), i got bogomips 4784.12 in
> /proc/cpuinfo, it is same as my P4 2.4GHz. The problem is, my amd 3400+
>
> run as slow as P4 2.4Ghz on fedora core 4 x86-32 version. why?
> You may say my fedora is not 64 bits, so my amd 64bit cpu runs slow,
>
> but my athlon 64 runs much faster than P4 on my 32 bits windows.
>
>
> thanks
> from Peter (cmk128@hotmail.com)
What in particular is running slow? I hope you aren't using bogomips as
your metric, the bogo in bogomips stands for bogus. Bogomips is a simple
timing loop, it's not meant to be a benchmark. I'm not surprised that a
2.4GHz P4 and a 2.4GHz A64 have the same bogomips number. Bogomips is just
a for loop. The branches in a for loop are completely predictable, if you
do for(i=0;i<1000;i++) and guess that the branch will be taken you'll be
right 999 out of 1000 times. As a result the deep pipeline that cripples
the P4 on real programs (where branches are much less predictable) doesn't
matter. What you've found out is that bogomips is a measure of the clock
rate, not of the effective instruction rate.
- Previous message: cmk128_at_hotmail.com: "athlon 64 + FC4 == SLOW"
- In reply to: cmk128_at_hotmail.com: "athlon 64 + FC4 == SLOW"
- Next in thread: cmk128_at_hotmail.com: "Re: athlon 64 + FC4 == SLOW"
- Reply: cmk128_at_hotmail.com: "Re: athlon 64 + FC4 == SLOW"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
