Re: Questions about notebooks and Debian
On Thu, Aug 02, 2001 at 07:05:44AM +0200, Sven wrote:
> On Wed, Aug 01, 2001 at 05:24:51PM +0100, Bastien Nocera wrote:
> > Noah John wrote:
> >
> > > the ibook is definatly a good purchase. from what i gather, some sound
> > > still doesn't work. However, what i really meant to say was:
> > > G3 500 MHz vs. i386.
> > > is about an 800-933 MHz P3
> > > is about a 733 MHz Athlon or P4
That's odd. The P4 needs more clocks to get work done than
an athlon. A 733MHz Athlon would not run at close to the same speed
as a P4 at 733MHz. Look at the SPEC benchmarks. (and note that if
you scale down the proc clock speed without changing memory speed, you
get more bandwidth and less latency relative to the clock cycle, so
the P4 has less advantage due to its bandwidth, but the Athlon has
less advantage due to lower SDRAM latency.)
This is a rough enough comparison that this probably doesn't matter,
but I would have thought you'd put the P4 at 933 and the P3 at 800 if
you were going to use the numbers you did.
> > > is about a GHz celeron (celeron has no b-side cache).
All celerons since the 300a have had 128kB on die L2 cache. I don't
know what set-associativity it has. (This makes a big difference,
BTW.) I think it runs at the proc clock speed, instead of half speed
like pre-coppermine PIIIs.
> >
> >
> > You certainly have benchmarks to prove that ?
> > The G3 in the iBook has only 128k of L2 cache, and it is slower than my
>
> Err, isn't that 256Ko of on die cache, like the PIII and the later athlons ?
The PPC 750 does not have on die L2 cache, but gets almost as good
hit latency by putting the cache-management logic (tag lookup, etc.)
on die, and having a dedicated L2 connection bus.
IIRC, apple chose to run the L2 at half the proc clock speed. (The
chip allows the board designer to set the clock ratio to use for the
L2 bus, from 1:1 on down.)
>
> > iMac G3 400 (which has 512k of this same cache).
>
> Did you do any benchmarking on it ?
>
> > If you want a powerhorse of a laptop, get a TiBook, or an x86 laptop.
> > The x86 will give you a crappy architecture, and not much battery life.
>
> but the TiBook costs twice as much and more, and as said battery life on i386
> is very bad, especially on the athlon/duron based ones.
Battery tech is getting pretty good. My friend got a 1GHz PIII
laptop with 512MB RAM, and he says the battery lasts 3-4 hours IIRC.
I asked him how toasty it got, and he said it wasn't very nice holding
it in his lap...
One important thing to note is that GCC's codegen for PPC takes
longer than for x86. This means that compiling the kernel is not a
relevant benchmark for comparing compute power of x86 vs. PPC. It is,
however, an appropriate benchmark for seeing how long it takes to
compile stuff. If compiling stuff is one of the things that you do
that actually matters how long it takes, then it's something to think
about. (some stuff takes so little time that it doesn't matter which
machine does it faster, like running ls. ls runs fast enough on any
new machine. 0.1ms vs. 0.2ms doesn't matter. Compiling a kernel (or
the project you're working on) is something that takes long enough
that you have to do something else while it happens, so it does matter
how long it takes.)
--
#define X(x,y) x##y
Peter Cordes ; e-mail: X(peter@llama.nslug. , ns.ca)
"The gods confound the man who first found out how to distinguish the hours!
Confound him, too, who in this place set up a sundial, to cut and hack
my day so wretchedly into small pieces!" -- Plautus, 200 BCE
Reply to: