P4 Willamette vs Celeron Northwood...

[Lancelot]

(also posted this in the Intel CPU section, but since this concerns SETI and our team


Maybe this helps the ongoing debate about P4-Celerons and 'how-bad-the-P4-Willamette-is-when-compared-to'. Average SETI WU times tell it all, both rigs running WinXPpro SP1 and SETI CLC3.03. The Celeron system used to be my main rig before I upgraded to the system in my signature. The Willamette system is what I recently upgraded my kid's rig to, coming from a P3-667. Both have been running SETI for a few weeks, the Willamette will continue to do so in aid of the overclockers.com team ofcourse!

P4 Willamette 1.7Ghz 400FSB, Asus P4B-MX i845, 512MB Infineon PC133-222. CPU running default, RAM @3:4 133Mhz CL2-2-2-6. Average WU time: 4 to 4.5 hours

Celeron Northwood 2.0Ghz 400FSB, Asus P4PE i845PE, 512MB Samsung PC2700-333. CPU OCed to 2.66Ghz 533FSB, RAM @4:5 166Mhz (DDR333) CL2.5-3-3-6.
Average WU time: 5 to 6 hours

I have always been fairly optimistic about the P4-based Celerons but I had no idea they were doing this bad. Eventhough the Willamette has twice the L2, she isn't OCed and has to deal with SDRAM for crying out loud!

 

[TC]

The P4 is heavily dependent on the cache to keep the pipeline full.

 

[eaglescouter]

Ouch, that's not pretty when a Celeron 2.66Ghz takes longer to crunch a wu than a 1.7Ghz P4.

Are you sure that there is not some limiting factor in the Celeron rig? Possibly:
a. Some other services running in the background. Spyware?
b. Something set in the Bios at less than optimal settings?
c. Does Sandra report that the memory is a bottleneck for the CPU?

Something just seems wrong with this picture.

 

[Lancelot]

Nope, the Celeron's RAM was even set at 4 bank interleave and 8 burst length. My bro-in-law runs a Celeron 2.4 @3.2Ghz with 1GB PC2700 for video-rendering and SETI still takes about 5 hours per WU on that rig. Like TC said: SETI is heavily cache dependant...

 

[Dk Jedi Allianc]

My first rig was a Celly 566 with 384 Pc-133 ram. Took 11 Hours. Then Oc´ed it to 708 (with box cooler) then it was doing 9,5 hours.

 

[Greg M]

It's not necessarily SETI that is cache dependent, it's that the P4 architecture is cache dependent. The way the P4 works with its deep pipelines and high clock, if you don't keep the pipelines full, you're spinning your wheels with wasted cycles while data has to be pulled from RAM instead of the cache. The Celeron was never a good idea, and with P4 architecture, it's a flat-out terrible idea.

 

[psyshack]

Ive stayed away from cellys at all cost..

 

[cooldrum3]

Ive stayed away from cellys at all cost..

Or lack of cost for that matter

 

[jazztrumpet216]

Ive stayed away from cellys at all cost..

These P4 Cellys anyway. I liked the Tualerons...