[ILUG] Thoughts on a compile box

[lbedford at wbtsystems.com]

On Thu, Apr 27, 2000 at 03:40:02PM +0100, Kenn Humborg wrote:
> Oh, I'm not afraid of the actual work of putting it together.
> That's no problem.  What worries me is "well, this particular
> chipset, with this particular BIOS and that particular RAM
> just don't get on..."  With the higher clock speeds, _everything_
> gets harder to match correctly.  That's why I'd like to buy a
> machine from someone who can say "Well, I've sold 10 of those
> motherboards with that RAM and that CPU, and that hard disk,
> and it all works fine" (and has a reputation for not being
> flexible with the truth!).
In which case why not hunt down either PIII-450's/PII-400's (you
can get them in some places), stick it on a Gigabyte/Tyan
Dual Slot 1 board. It won't fall over, and will generally be
reliable. You can get Dual Intel boards, but they're more
expensive. I don't think the ram speed hit for PC-100 SDRAM
over RDRAM is going to bother you, as the list price
for 256Mb RDRAM from Kingston is $1600 :)
> 
> > You should get a 750MHz Athlon as specced for well within your
> > budget - you
> > could go faster, but I don't think the extra performance is worth
> > the extra
> > cost.
> 
> So, dual 500MHz Celeron vs 750MHz Athlon...  Which will compile
> faster?  My gut feeling says the SMP box, but I could be wrong.
> Please enlighten me.
I suspect for compilation you'll find the celeron faster. Don't
quote me, I just suspect. You're not going to be using
the 3Dnow! instructions, and it's the FPU on the Athlon which
is great (better than Dual Celeron anyway). 
There's also the business of a 750Mhz Athlon will have a 100/133
Mhz Bus speed, which the Celeron's won't have (unless you overclock
them), so memory speed will be a bit limiting. I don't know
what compiling is like for cache hits, but the celeron's only
have 128k, which hampers them in Seti by most accounts.

L.
-- 
Liam Bedford						System Administrator
WBT Systems, Block 2, Harcourt Centre, Harcourt St.	01-4170153
print pack"C*",split/\D+/,`echo "16iII*o\U@{$/=$z;[(pop,pop,unpack"H*",<>
)]}\EsMsKsN0[lN*1lK[d2%Sa2/d0<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<J]dsJxp"|dc`