| Subject: Re: The effect of L2 cache on SETI times. |
| From: Martin |
| Date: 10/11/2003, 16:02 |
~misfit~ wrote:
[...]
I have an XP2200+ and my g/f has a Barton 2500+. Both run around 1.8GHz
stock but the Barton has twice the L2 cache. (256KB and 512KB respectively)
There is very little difference in WU times between these two machines, both
using the same client.
So from this (albeit very limited) comparison it seems L2 cache doesn't make
a huge difference to WU times. *Unless* any L2 over 256KB doesn't make a big
difference. It could concievably be different between 128 and 256 as in the
Celly/PIII previous discussion.
You get a
*big* speedup when the FFT operations fit nicely into the
cache. This happens for 1024kBytes or more of cache.
Just thought I'd mention it. I expected my g/f's PC to do units a lot faster
than mine.
Note that the Barton has an 'enhanced' rating factor taking account of
its larger cache. The actual clock speed is correspondingly lower than
for an 'equivalent performance' AthlonXP or Intel CPU.
The new Opterons with their 1MB caches crunch s@h rather quickly!
I've just O/Ced her machine to 200FSB for 2.2GHz. 0.1v vcore increase as it
didn't like Prime95 at stock vcore. Just running prime again now but so far
[...]
Interesting...
Thanks,
Martin
--
---------- Give a man a fish and you have fed him for today.
- Martin - Teach him how to fish and he won't bother you for weeks!
- 53N 1W - - Anon
----------