On 14 Oct 2001, at 7:30, Jean-Yves Canart wrote:

> According to latest benchmarks (http://www.mersenne.org/bench.htm),
> AthlonXP seems to be slower than the Thunderbird.  Does anybody have a
> technical explanation ?

The Athlon XP lies about its speed. Remember the old Cyrix trick? 
Well AMD have gone for the same - pick a benchmark that suits 
you, then claim your chip is "1800+" if it runs _that_ benchmark a 
wee bit faster than the opposition's chip running at 1800 MHz.

I think an Athlon XP 1800+ is in fact running at 1533 MHz.

IMHO this is as bad a marketing scam as attatching the XP label, 
which is obviously designed to hoodwink consumers into thinking 
that an upgrade is neccessary if the want to run Windows XP - 
thus enabling AMD to benefit from Microsoft's imminent marketing 
blitz.

The unfortunate thing about this is that I'm afraid that most 
consumers will swallow at least part of the lies :(

Having said all that, AFAIK the basic core of the Athlon XP is much 
the same as the Thunderbird: it is fabbed at 0.13 microns instead 
of 0.18 microns, which means keeping it cool should be a bit 
easier. This should not affect the basic chip efficiency one way or 
the other. Running Prime95, I would expect an Athlon XP running 
at 1.4 GHz to be exactly the same speed as a Thunderbird 1.4GHz 
(266 MHz FSB) in the same board. N.B. you may need a BIOS 
upgrade to support the Athlon XP due to the changed core voltage.

There is one small architecture change between Thunderbird and 
Athlon XP. Athlon XP now supports Intel SSE operations - but still 
not SSE2 as used in the Pentium 4. This probably doesn't help 
Prime95 since prefetch, which _is_ relevant, was already supported 
by all variants of Athlon CPUs. But it might be worth telling a 
system running Prime95 on an Athlon XP that it's actually running 
on a Pentium III just in case that's any faster than the native Athlon 
code.

With the large multipliers used by current processors, the 
efficiency of the chipset and the memory have quite a large effect 
on the performance of a CPU. Moving a CPU between systems can 
therefore result in changes in performance as measured by 
benchmarks. You really need to check that the chipsets and the 
memory configurations are the same before you can compare 
benchmark timings in any meaningful way. Note particularly that 
e.g. 256 MB can be made up of one bank of 256 MBit, two banks 
of 128 MBit or four banks of 64 MBit RAM chips; expect a 
performance difference of 5% - 7% between these configurations 
even if the chip access speeds & timings are identical. More banks 
are faster.
> 
> Do we have to consider now Intel/P4 as the best platform (at least for
> prime95)?

For LL testing using mprime/Prime95:

That seems to have been the case for some months now; it's 
SSE2 which makes the difference. A P4 running Prime95 is well 
over twice as fast as a T'bird running at the same clock speed.

I would expect the same to be the case for any other application 
which has been specially coded to take advantage of SSE2. There 
probably are examples, especially in the area of multimedia, but I 
can't think of any offhand.

For general work (including trial factoring):

Any variant of Athlon will be significantly faster than P4 running at 
the same clock speed. At clock speeds up to 1.2 GHz, PIII may be 
a little faster than Athlon. Especially the new "Tualatin" PIII 
processor with 512KB cache.

However (given suitable software) other processors may be _much_ 
more efficient in terms of work done per clock cycle than any of the 
mainstream PC processors. PowerPC, Alpha and especially 
Itanium spring to mind. AFAIK none of these run above 1GHz yet, 
though they can be found as the basis of some impressively 
powerful workstations. 

AMD do have a point that raw CPU speed is no longer a good 
indicator of overall system performance, let alone the performance 
on a particular benchmark task, which can be further influenced by 
extraneous factors like the efficiency of the graphics card.

Regards
Brian Beesley
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

Reply via email to