Re: Linux on AMD K6

Pavel Machek (pavel@Elf.mj.gts.cz)
Thu, 17 Jul 1997 19:53:03 +0200


> On Mon, 14 Jul 1997, yuri mironoff wrote:
> > On Thu, 10 Jul 1997, Dan Hollis wrote:
> > > CPU Dhrystones
> > > ---------------------- ----------
> > > AMD K6/200 374838.4
> > > Pentium Pro 200 356929.4
> > What options are you compiling with? I get 416K for a 200 MHZ PPro.
> > Aside from that - Dhrystones is a useless benchmark. Try lmbench or
> > do a kernel build (serial and parallel) - now that would be interesting.
>
> Kernel build measures the disk and i/o subsystems more than it does the
> CPU. Thus kernel builds are more useless than dhrystones. I'm only
> interested in measuring the raw CPU power, not the speed of the
> SCSI disks.

No. With 20Meg of ram, disk is almost never accessed when compiling
kernel. Kernel compilations get some 95% of cpu... Well, question is
if you want CPU performance, or CPU & memory performance. make of
kernel tests the second one.

> I'm using the byte bench 2.0 benchmarks with default options, I don't
> touch anything. It was compiled exactly the same on both the ppro/200 and
> the K6.
>
> People might claim dhrystones are "useless", but unfortunately it does
> seem to be a good indicator of speed, despite what people claim.
>
> E.g. we compared a pentium and a powerpc with dhrystones. People claimed
> powerpc should be faster, but no matter what application we ran, it
> wasn't. In fact, each application showed exactly the same speed as
> dhrystone predicted.

-- 
I'm really pavel@atrey.karlin.mff.cuni.cz. 	   Pavel
Look at http://atrey.karlin.mff.cuni.cz/~pavel/ ;-).