Am 20 Oct 2004 20:47 meinte Zane H. Healy:
>Sorry for the off-topicness... I googled to find a
comparison
>between these two chips with regards to relative performance and
>can't find anything. Can anyone point to a spot that has some
>definitive comparisons between the two?
Well... The PA-8500 is a true 64-bit CPU, and the
Opteron is a
kludge. What more do you need to know.
Define true and kludge. One could argue the Opteron is a tru 64 Bit
while the PA is a kuldge .. *G* Serious, it's the old game what is
an n-bit CPU - is a 68008 an 8 Bit CPU, a 16 or a 32 Bit one? If
it's the buswidth, then a Pentium is already a 64 Bit CPU... is it
the (work) register width, then it's 32, and looking at the typical
instruction encodeing, it's 16 Bit ... Oh, and don't forget the ALU
size (another problem, which ALU to choose, since modern CPUs have
usualy several)...
If you're looking at the AMD64 vs. PA Ric, I can't see any relevant
difference. Both are working with 8, 16, 32 and 64 Bit integers on
a regular base, both have an opcodescheme that is optimized for the
most common (or at least most common as the designers see it) data
types (8, 32 and 64), and both can perform the usual every day tasks
of a general purpose CPU without minding any special quirks.
Performancewise (as the original question was), SPEC95 is usualy an
acceptable measure. See here for data from 99
http://www.heise.de/ix/benchmarks/SPEC/rh9909/rh9909_5.html
Not realy new, but a quite nice look at SPEC
http://www.heise.de/ix/benchmarks/SPEC/rh9909/default.html
SPEC is also the first place you should look if you want to compare
CPUs performancewise. But they publish the latest results (go on CPU/
CPU2000/Pusblished or for the older to
http://www.specbench.org/cpu95/results/
) I GHz AMD Athlons had rates arround 45 here, so when just calculateing
the PA-Result linear, we come to the same region... now, the AMD64
seams to be roughly 3-4 times as fast as the old 1 GHz Atlon at the
same clock level ... so get your fugures here.
Performancewise I think the Opteron is top notch at the moment.
And personaly, I don't care what bit pattern initiates a certain
operation in my Server, as long as the price/performance ratio is
as good as possible - also a reason why I'm now switching to mini-ITX
boards as workstations. Not realy fast, but nice all in one, for a
real low price.
And as usual YMMV - means, it realy depends on your task. Isn't that
the big number game between AMD and Intel? Just look what kind of
'real world' example Intel uses to show their advantage. The P4 has
a rather long pipeline, which allows high speeds when goin thru
predictable sequences (and a big malus then not), so for jobs like
MP3 or Video encodeing here raher long, FPU intense and often repeated
sequences on linear datasets are to be performed, it oursuns AMDs
Athlon by sheere MHz and is only limited by the bus bandwith for data
(that's why Intel did realy soup up the bus over the last versions).
Gruss
H.
--
VCF Europa 6.0 am 30.April und 01.Mai 2005 in Muenchen
http://www.vcfe.org/