Hi, I went through my stats and pulled out an average work unit from each computer and went to google to do the math to turn thousands of seconds into hours and here's what I got.

P4 1.6 5.2 hours per unit
cel 433 16 hours per unit
pII 300 21 hours per unit
PIII 500 11.2 hours per unit (512 L2)
PIII 600 11.3 hours per unit (i've not checked L2 but I bet it's smaller)
P4 2.2 2.8 hours per unit

anyone else want to do this and post some stats... I'd be interested in hearing how some other cpus do...

---

<img src="http://teamstarfire.org/boinc/summary.php?name=Dwarlock&amp;team=Kansans%20Searching%20for%20a%20Planet%20With%20People%20More%20Interesting%20Than%20This%20One&amp;seti=7832952&amp;cbg=grey">

Intel Celeron D 2.93GHz
512MB Ram
256KB L2 Cache

SETI: 3.5 Hours
Einstein: 8.5 Hours

I don't think my crunch times are very good. Would be interested in seeing how other Celeron D's are doing.

Dig

---

Pentium M 2.0GHz 134 Results, Average 2:05
AMD XP 2600+ 190 Results, Average 2:40
AMD XP64 3000+ 190 Results, Average 2:25

Pentium M rocks!

---

> Pentium M 2.0GHz 134 Results, Average 2:05
> AMD XP 2600+ 190 Results, Average 2:40
> AMD XP64 3000+ 190 Results, Average 2:25
>
> Pentium M rocks!
>
>

i agree - i have a tablet PC with a Mobile Pentium and it kicks rear

---

http://www.martin-karch.de

> > Pentium M 2.0GHz 134 Results, Average 2:05
> > AMD XP 2600+ 190 Results, Average 2:40
> > AMD XP64 3000+ 190 Results, Average 2:25
> >
> > Pentium M rocks!
> >
> >
>
> i agree - i have a tablet PC with a Mobile Pentium and it kicks rear
>

Now, Now we can't be discussing WorkUnits. Workunits cause cheating and unnatural competition. Workunits are illegal to talk about in Boinc.

---

> anyone else want to do this and post some stats... I'd be interested in
> hearing how some other cpus do...

Ok, how about this?

You can also see the difference a MB and dual channel memory make in the processing of work. I had two identical processors (bought on the same day) with one getting a good MB and the other, well, lets just say that I took the MB bundled as the CPU was cheaper than I could get w/o the MB (loss leader sale bundles) ...

---

Paul, thanks. Nice to be able to make a comparison. My times don't seem quite so bad now... about average i guess. :)

Dig

---

> Paul, thanks. Nice to be able to make a comparison. My times don't seem
> quite so bad now... about average i guess. :)

I hope to have a second Macintosh later this year. I am waiting for the next generation to come out. We were hoping for a dual dual-core to be the top of the line, but current speculation is that we will just see an increase from 2.5 GHz to 2.7 GHz in the top end.

Since mine is a 2.0 GHz, well, I can't justify an early purchase ... but, I do want to upgrade the workstation so ...

The Xeon, has yet to complete a CPDN Work Unit so I am still waiting for that to happen. It has one with just over a week to go so, I will be posting a time for that probably 3 Sundays from now ... the time it is projecting is about 20 Days total time per model so the Xeon will have the best time it looks like for now.

I don't have the CPDN running on the Mac as the "hanging" and multiple processes on the Macintosh seem to kill the model when you stop the "zombie" processes ... and with this happening on a regular basis, well, it does not make sense to have lots of models errored out. When that bug gets squished I will start to run models on the Macintosh again ...

---

hmmmm, I'd noticed the differance in motherboards on a few systems as well. as I've now chacked and discovered that my P3 500 and p3 600 do both have the 512k l2 I'm just betting that the Abit BX6 MB the 500 is in is making the differance (it's always out proformed the average before)

---

<img src="http://teamstarfire.org/boinc/summary.php?name=Dwarlock&amp;team=Kansans%20Searching%20for%20a%20Planet%20With%20People%20More%20Interesting%20Than%20This%20One&amp;seti=7832952&amp;cbg=grey">

i've noticed a change in floating point speed and integer speed in the same processors running Seti and Einstein.

Floating Point Speed SETI EINSTEIN
XP2400+ 512MB 1837.24 2355.36
Duron 1.8 512MB 1680.26 2461.94
XP2800+ 256MB 1877.41 2496.63

Integer Speed
XP2400+ 512MB 3048.47 4350.57
Duron 1.8 512MB 3105.37 4213.18
xp2800+ 256MB 3280.36 4403.16

---

> i've noticed a change in floating point speed and integer speed in the same
> processors running Seti and Einstein.

You will also see a fluctuation if you run the benchmarks several times in a row.

Note:

You should have no other processes running in the background like instant messaging, web browsers, and after starting don't open or close windows or move the mouse.

Even with there precautions you will have minor changes to fairly significant changes in the benchmark results ...

---

actually Paul i was referring to the difference between benchmarks for
SAH and Einstein...
pretty huge differences
sorry my tables dissolved....

---

> actually Paul i was referring to the difference between benchmarks for
> SAH and Einstein...
> pretty huge differences
> sorry my tables dissolved....

The benchmarks have nothing to do with which projects you are attached to or running at the moment.

The reasons that the two projects show different benchmark scores is that they are using two different runs of the benchmark.

You should not be seeing that large of a difference between runs which was a point I tried to make back in the days of the Beta. You can see part of the debate and the tests we ran.

As the scheduler contacts are made all of the pages will eventually show the same numbers.

If I look at my Dell Dual-Xeon I had:

FP: 1496
Int: 1827

On the web pages. Running the benchmark program again twice in a row I got:

FP: 1500 then 1501
Int: 1835 then 1846

Which shows the variability in the benchmark scores. I argued for a running average for the benchmarks to eliminate the variablility, and even for a project wide averaging of the scores across like platforms.

Though I do have problems with the specific implementation I am not against the process nor the use of a benchmark. This is a better system than count the "beans" (Work Units) because of the variability of the work and processing times of specific Work Units.

For example SETI@Home, LHC@Home, and Predictor@Home all have Work Units with varying processing times. With SETI@Home's being driven by the Angle Range, LHC@Home's by turn count, and Predictor@Home's by evil demons (or something

---

> actually Paul i was referring to the difference between benchmarks for
> SAH and Einstein...
> pretty huge differences
> sorry my tables dissolved....

The benchmarks have nothing to do with which projects you are attached to or running at the moment.

The reasons that the two projects show different benchmark scores is that they are using two different runs of the benchmark.

You should not be seeing that large of a difference between runs which was a point I tried to make back in the days of the Beta. You can see part of the debate and the tests we ran.

As the scheduler contacts are made all of the pages will eventually show the same numbers.

If I look at my Dell Dual-Xeon I had:

FP: 1496
Int: 1827

On the web pages. Running the benchmark program again twice in a row I got:

FP: 1500 then 1501
Int: 1835 then 1846

Which shows the variability in the benchmark scores. I argued for a running average for the benchmarks to eliminate the variablility, and even for a project wide averaging of the scores across like platforms.

Though I do have problems with the specific implementation I am not against the process nor the use of a benchmark. This is a better system than count the "beans" (Work Units) because of the variability of the work and processing times of specific Work Units.

For example SETI@Home, LHC@Home, and Predictor@Home all have Work Units with varying processing times. With SETI@Home's being driven by the Angle Range, LHC@Home's by turn count, and Predictor@Home's by evil demons (or something).

---

Dwarlock,

> anyone else want to do this and post some stats... I'd be interested in hearing how some other cpus do...

AMD 2500+ : 3.5 hrs/wu
AMD 2000+ : 4.0 hrs/wu

---

AMD 64 3200+ - 2.25hrs on average

---

Wiki de BOINC

<img src="http://www.boincstats.com/stats/banner.php?cpid=51a3eaaef8df276544f56140a8a65413">

> Paul, thanks. Nice to be able to make a comparison. My times don't seem
> quite so bad now... about average i guess. :)

If only I was working I might be able to afford additional computers for more statistics ... I just posted my results for the week, over 800 Work Units since LGC@Home and Predictor@Home seem to have issued LOTS of short runing Work Units.

The times have not changed much ... though with the short turn-around times the average time came down just a little bit.

---

AMD XP2500+

Seti : 2h40m
Einstein : 6h35m
Climate: 570h

---

98SE XP2500+ @ 2.1 GHz Boinc v5.8.8

And God said"Let there be light."But then the program crashed because he was trying to access the 'light' property of a NULL universe pointer.

amd 2400xp 3hr 27m
amd 2400xp-m 2hr 52m
amd 2800xp 2hr 40m
duron 1800 @2150mhz 3hr 2m
duron 1800 3hr 22m
p3 1ghz 6hr 45m
p4 2.66 3hr 40m
p-m 715 1.5 2hr 32m
p4 1.6 4hr 20m
p2 400mhz 14hr 32m
k6-2+550 20hr 12m

all times are for seti

---

Pentium4 2.0GHz 2hr 15min
AthlonXP 3200+ 2hr 40min
Pentium4 2.4Ghz 4hr 30min
Pentium4 1.5GHz 6hr 45min
AMD K-6 473MHz 23hr 50min

---

7,049 S@H Classic Credits