From: gilbert@inviso.com (Gilbert Callaghan) Subject: SUMMARY: Linux performance: WOW! (was: OUCH!) Date: Tue, 29 Jun 1993 20:13:35 GMT
In article <C96oC5.366@inviso.com> gilbert@inviso.com I wrote:
>It is on a local network with several other 486/33's running Interactive
>Unix. It's running 0.99pl10 with net-2 and everything works great.
>
> BUT... IT'S SLOW!!!
>
>Here are comparisons of Linux and ISC compiling MicroEMACS with gcc.
>This is the output of 'time make':
>
> Linux ISC
>
> real 8:34.0 real 2:37.1
> user 8:07.1 user 1:23.5
I got a LOT of helpful replies on this one, so I thought I'd write
a summary.
First off, several people suggested that in order for the test to be
fair, the ISC machine must also be running gcc. I guess I didn't make
it very clear that BOTH machines were running gcc.
Anyway, most people suggested that I try, as a test, a kernel compiled
to use only up to 16 Meg. Then I re-tried my compile test: 2:48.1
Yes! Just a tad slower than ISC; that's ok.
Or so I thought - then it got even better. I went into my CMOS and found
that the cache was completely DISABLED! SO I enabled it (128k cache) and
ran my compile test in 1:47.0 minutes! Far better than ISC can do.
Naturally I started wondering if the cache was enabled on the ISC box.
So I took it down and found that it WAS.
One last test. Since caching is now enabled, I re-tried it with the
32Meg kernel. I got a very slow 6:34.0 result.
Personally, I'm happy now. I can run at blazing speed with 16 Meg, which
is enough for me. There does still seem to be a problem, though. It
appears that this has nothing to do with caching, as caching speeds up
both the 16M and the 32M kernels. And I have received mail from other
people who are experiencing the same thing. Hopefully we can unravel
this mystery soon.
Perhaps it's a motherboard problem; I'm not sure.
Thanks for all the replies.