From: Dave.Deavours@f303.n232.z1.fidonet.org (Dave Deavours) Subject: Which is better for Linux: 486DX2/66 or 486DX/50 Date: Fri, 28 May 1993 12:46:18 -0800
CDA> There's nothing to discuss. Unix performance is dependent
CDA> on available RAM and disk speed, not processor speed.
Hmmm. I'm sure a lot of people will take exception to that statement.
Let's see. How does a kernel recompile (make clean;make all) sound as a
good benchmark of overall system performance? Or better yet, on a
loaded system under Xwindows? Or maybe with 3 or 4 users on all using
GCC? I have a 386-40 with 16 megs RAM, 32 megs swap (and a fast, ie,
<12 ms drive). How often do you think it goes to disk during a compile?
How about on a heavly loaded Xwindows system? I'd put the cache hit
rate at about 90%. I never hit swap so I know it's not a memory
bottleneck (free will report from 1 to 4 megs available). Now it's not
a disk bottleneck or a memory problem--it's a CPU bottleneck.
CDA> Fast processors are cool, but they're really just needed
CDA> to produce a linear shift in speedup for Mickeysoft
CDA> WinDoze and other DOS nonsense, which all runs in real
CDA> mode and hence, sucks.
First off, I get a good laugh out of people running DOS on a 486-66. In
a DOS environment you're right--the CPU power is just a speed fix. But
under a multi-tasking, multi-user OS the CPU power becomes a serious
issue. Let me put it to you this way: Would you like to compare an
AT&T 3b2 from 1985 with your 486-33? Why not stick with the 3b2? The
horespower under the hood is put to better use with a OS such as Linux,
but a time starved process will still be slow, even if the drive is fast
and the memory is available. When I upgrade it'll be to a 486-66 w/256
or 512K Sram cache, 16 or 20 megs RAM (depending on how well the
motherboard handles the ram over 16 megs).
Dave
* Origin: PBS BBS Node 3 (1:232/303)