View Single Post
  #9  
Old January 15th 04, 12:44 AM
~misfit~
external usenet poster
 
Posts: n/a
Default

kony wrote:
On Wed, 14 Jan 2004 11:41:19 -0800, "Cari"
wrote:

Here's a comparison...

Celeron 2.0Ghz..... 512mb PC133RAM

One Setiathome work unit.... 15 hours

P4 2.0Ghz..... 512mb RAM PC133 RAM

One Setiathome work unit 5 hours

P4 2.0Ghz 1Ghz PC2100 RAM

One Setiathome work unit 3.2 hours

No, I'm never even contemplating a Celeron procesor ever again.

Cari
www.coribright.com



Not a realistic comparision of the CPUs though, only of SETI
performance, which is atypical.


SETI times are not atypical of CPU benchmarks. I have a book where I write
down benchmarks of every system I build. They include SETI times, CPUMark
and PC Mark 2002 among others. The CPU benchmarks show a distinct
correlation with SETI times. I suppose that you could say all synthetic
benchmarks are atypical.

It would be fair to consider SETI
performance only if that is the only thing the system ever does... as
soon as it does something else, too, the results drastically differ...
SETI just happens to exploit some of the larger L2 cache, but only if
it STAYS in the cache.


I (well, my g/f and I) have a Barton clocked at 2.2GHz and a Tbred at
2.1GHz, 512Kb and 256KB of L2 cache respectively, and the SETI times are
within 10% of each other, as are CPU Mark (210 and 195) and PC Mark 2002 CPU
benchmark (6,789 and 6,334). Even running SETI CLI while the machines are
being used for other things (including gaming) this 10% rule seems to remain
constant.

I benchmark every system I build, run Prime95 for 24 hours, run Prime and
SETI concurrently for 24 hours, then run SETI alone for 24 hours. Nothing
goes out of my door until it has done that error-free and maintained a
reasonable CPU/case temp. I *could* build a system and have it out the door
in hours but I hang onto them for nearly a week, until I'm satisfied, before
I let them go. That includes systems built from recycled parts or upgraded
sytems.

One caveat, I haven't built a P4-based Celeron machine. And won't. The last
Celeron system I built was a Tualatin. With the P4-based Cellys I believe
the slowness has a lot to do with cache latency as well as cache size.

Using a system normally will show much less performance penalty from a
Celeron,


If you consider gaming 'normal' then I'd have to disagree with you.

but certainly one doesn't belong in a system being built to a
mid or high-end price-point, but frankly, I wouldn't even consider a
P4 except for a high-end build, as the Athlons are much better bang
for the buck.


Agreed, especially the Barton 2500+, 90% of them run fine on a 200FSB using
the standard HSF (as long as case-cooling is set up properly) for XP3200+
speed. Some need a slight vcore boost, some will do 2.2GHz with no vcore
adjustment. I've only had one that wouldn't do 2.2GHz Prime95-stable at a
reasonable vcore.

The only time I would consider a P4 is for a dedicated video-editing machine
(SSE2). Not the sort of thing I build machines for ayway.
--
~misfit~