Pentium 4 HT Win SMP Results

[GigaHertzAddict]

Pentium 4 640
Single Channel 1GB DDR RAM
Running @ 3.3GHz (206x16)
Crappy SIS chipset

2604: 41 minutes per frame
Not bad really considering it's using single channel and the sis chipset has crappy bandwith to begin with... I estimated about 40minutes, so I was pretty close..:beer:

2610: 54 minutes per frame
Dam! That's long.. My Quad does it in 7 minutes!

[pscout]

you will have to keep an eye on it with the 2610 frame times ... that is close to the deadline i think. You may need to sneakernet some of those wu's to faster boxes.

[GigaHertzAddict]

Yeah, I probably won't run SMP on that system.. I was just testing it out and letting others know what kind of results to expect.. Is there a way for me to use 100% of the CPU when folding a single protein on a HT system? Rather than folding to proteins I rather fold one, just faster..

[WYTOUT]

On an HT system you aren't really only using 50% of the CPU, it just looks that way in taskman. You are pretty much using 100% of the processor's instructions that are normally utilized by FAH. The rest of the available unused instructions are not really helpful to crunching wu's. Being able to use the "rest" of the processor would net you a very small gain. On an HT it is best to run just one fah client.

Keeping HT enabled, those leftover instructions that your client isn't using actually can help you because other things that your computer runs that might be able to use those instructions will be utilizing those portions of the cpu, leaving more of the fah-utilized instructions available to continue crunching your wu's. So really, the one client running is going to give you the best times for one WU on an HT processor.

[Lord_Anubis]

P4 3.0E HT @ 3.6
dual channel 2x512 DDR 3200
Asus P4P800SE

Windows SMP folding:
p2651 1760.0 points 33mins 8sec = 765ppd

p2608 1385.0 points 33mins 53sec = 588.6ppd

p2652 1148.0 points 34mins 1sec = 486 ppd

p2610 1523.0 points 56mins 2sec = 391.4ppd

After the July price drops I hope to replace this rig & a few others, but for now she makes the SMP deadlines.

[benbaked]

P4 3.0E HT @ 3.6
dual channel 2x512 DDR 3200
Asus P4P800SE

Windows SMP folding:
p2651 1760.0 points 33mins 8sec = 765ppd

p2608 1385.0 points 33mins 53sec = 588.6ppd

p2652 1148.0 points 34mins 1sec = 486 ppd

p2610 1523.0 points 56mins 2sec = 391.4ppd

After the July price drops I hope to replace this rig & a few others, but for now she makes the SMP deadlines.

Does a P4 w/ HT make more PPD running SMP then the regular client? It kind of looks like it, or at least its real close.

[pscout]

Does a P4 w/ HT make more PPD running SMP then the regular client? It kind of looks like it, or at least its real close.

probably more than double ... a lot of it due to the bonus on smp.

Important to note tho that cache probably makes the difference between making the deadlines and not.

So a well oc'ed prescott will likely do ok, but a northwood probably not.