Thanks to Alan for some heretofore unknown (to me) insight:
# > compile was measurably faster than my PII/333's, but since I work with
# > a huge amount of stuff on my desktop, perhaps that affected my `feel'.
#
# It might do actually. You might also want to play with the MTRR stuff
# to get the best out of X11 with such chips. SMP is way more sensitive
# to cache/bus stuff.
# >
# > As I said, it is my first impression, if it turned out to be a
# > misimpression, I will certainly make a retraction and correction. But
# > I am curious-- are double overclocked Celerons really THAT good?
#
# The cache is a bit small, but the cache is a lot faster and on a 100Mhz
# main memory bus the two often cancel out.
For those in the gallery who might be reading this ...
I experimented a little bit- `MAKE=make -j2' seems to make for
a typically faster compilation on the dual Celeron. Alan is probably
right, and the apparent sogginess of the Celerons is just an illusion.
The fact is that the dual Celeron 300A, before I overclocked,
`felt' slower than the several dual PPro boxen (mostly 200/512, though
there was a overclocked 233/256 box) that I had used over the last 2-3
years. The reason seems to be slightly longer context-switch time. I
think Alan's `soggy' adjective is an apt description for it.
I re-read again the MTRR documentation files and discovered my
old Accelerated X server to be one that already used the MTRR:
[root@terrorist linux]# more /proc/mtrr
reg00: base=0x00000000 ( 0MB), size= 256MB: write-back, count=1
reg05: base=0xed000000 (3792MB), size= 8MB: write-combining, count=1
I must admit to being at a loss, as to what else to try with
MTRR. Again, thanks to those who already responded for your helpful
comments. Being a newbie at tweaking, I welcome further suggestions,
and any commentary that you might offer!
Regards, B.Y.
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
