Folding@Home

[Groltz]

In the beginning, it became widely known that Athlons could outcrunch P4's in FAH. In the last few months I have wondered if this still holds true. P4 CPU's and chipsets have matured and evolved since those accusations started making their rounds.

Have the improvements in the P4 chipsets/CPUs boosted their ability to crunch in FAH? Any speculation on the upcoming models? (Springdale and Canterwood chipsets)

In today's Inquirer it says that AMD "will start sampling 2800+ and 3000+ Barton desktop CPUs in just a few days" I wonder how these will be at the task of folding. The only architectural change is a larger L2 cache, right?

The AMD Opteron/Athlon-64. Anyone's guess I suppose. Folding performance is heavily reliant on FPU performance...true? These new generation AMD CPU's might be very impressive in this regard.

It is unfortunate that a standardized FAH benchmark has not been authored that hardware sites could include when they review CPU's. With as many FAH teams as there are, many of them belonging to computer hardware sites, I am surprised this hasn't happened.

 

[blakerwry]

i just got a northwood based celery 2.0ghz.... I will test it out soon enough.. I'm hoping it will beat my 1700+

 

[SteveC]

i just got a northwood based celery 2.0ghz.... I will test it out soon enough.. I'm hoping it will beat my 1700+

It won't.

It is unfortunate that a standardized FAH benchmark has not been authored that hardware sites could include when they review CPU's. With as many FAH teams as there are, many of them belonging to computer hardware sites, I am surprised this hasn't happened.

It is surprising that a FAH benchmark hasn't been made yet. How hard would it be to make a reliable one?

 

[Groltz]

It is surprising that a FAH benchmark hasn't been made yet. How hard would it be to make a reliable one?

It shouldn't be very hard at all.

In the meantime, I still have an original "ProteinA" WU that I saved for checking CPU folding ability. :wink:

 

[Fushigi]

Have the improvements in the P4 chipsets/CPUs boosted their ability to crunch in FAH?

I can't say anything about the past 9 months, but my Dell Latitude P4M 1.6GHz (announced late March '02; I got mine in April '02) is roughly half as fast as my T-bird 1.4GHz when it comes to folding. Both machines are at the latest BIOS. Neither is overclocked. The T-bird is even saddled with a crappy ALI chipset. Both have sufficient RAM. The T-bird is W2K SP3; the P4M is XP Pro. I know it's not exactly an apples to apples comparison, but the P4M's extra 200MHz of clock should make up for most differences in laptop vs. desktop performance. Yet it's still only about 1/2 as fast.

- Fushigi

 

[CougTek]

I don't think we'll drop in the overall ranking that much. Our strong point is the consistency at which we produce. Remember how quickly teams Hotmail and University of Alabama past us? Well guess what, we are now over-producing them both and we've already past team Hotmail.

Many of the teams that currently out-produce us are probably like fireworks. They'll slow down quickly IMO.

See! SEE! Hong Kong and our eternal rival Team Austech will fall behind us. I should be paid for my premonition services. Am I good or am I not? No need to tell me, you're all lucky to team with me ;-)
[/ego trip]

 

[Handruin]

I don't think we'll drop in the overall ranking that much. Our strong point is the consistency at which we produce. Remember how quickly teams Hotmail and University of Alabama past us? Well guess what, we are now over-producing them both and we've already past team Hotmail.

Many of the teams that currently out-produce us are probably like fireworks. They'll slow down quickly IMO.

See! SEE! Hong Kong and our eternal rival Team Austech will fall behind us. I should be paid for my premonition services. Am I good or am I not? No need to tell me, you're all lucky to team with me ;-)
[/ego trip]

In deed we are, glad to be part of the folding team with you Coug! Good call on the overall status of our team.

So what is our next plan to keep us moving along?

 

[P5-133XL]

For some unknown reason productivy has increased. We were operating at the 3800-4200 points per week. now we are operating at the 5400 level. I don't think it is more participants because my per week score has gone from the 210 to the 260 range for the same computer (one Dual 1900+ MP Athlons).

I know that F@H was working on sending more valuable WU to computers that benchmark better and less valuable WU's to slower computers. Perhaps they have instituted that and we have a collection of relatively higher horsepower machines that get more valuable WU's. Perhaps it is simply that we are being lucky and getting better WU's. Something has changed but it is not us and it is not the competition.

 

[CougTek]

So what is our next plan to keep us moving along?

Maybe Pricewatching a powerful 350U$ cruncher for Mercutio's lab, so he can buy some twenty boxes more and install F@h on all of them?

For some unknown reason productivy has increased.

Not that unknown. It is true that we get more valuable WUs than before, but we also have two or three additional "regular" contributors than before. Mercutio and Kristy, along with Fushigi and Ghostweel. Our Bartender has recently (and temporarily) found new ressources too. I'm not even sure if Blakerwry had joined the team when we were on a sub-4000 points/week pace.

 

[Tea]

It's mid-January already and I'll be adding a few machines before too long, when the brief Ballarat summer is over and the horrible long winter sets in. About May, I guess. Till then, I'll be pumping the electrons through the air conditoner instead.

 

[blakerwry]

yeah, I 1st had 2 computers going (a pII 400 and a Duron 1.3) I have since updated to a PIII650, Duron 1.3, AthlonXP 1700+ and will shortly add my Celeron 2gHz...

 

[Fushigi]

T-bird 1.4GHz
P4M 1.6GHz
P3 500MHz
P2 333MHz
P3 1GHz x 5

The P2 & P3s should be considered temporary as these are just under-utilized servers at work. If the server loads pick up (or anyone notices ) I'll have to terminate them.

None are wholly dedicated to Folding and the T-bird & P4M don't have 24x7 access so they sometimes have to wait to upload & get new WUs.

Now if they'd come out with a PowerPC-based Unix client, I could run it on our AS/400s -- 6 CPUs across the 2 systems.

- Fushigi

 

[blakerwry]

Fushigi, from what I've seen of folding@home performance tests they effect the overall system speed < 1% when using the console version of folding@home.

 

[Fushigi]

I am, of course, only using the console version on all of the above. I know it doesn't seem to impact overall system performance very much. It's more a matter of if anyone says anything about the systems running at such a high utilization. Some of these servers are monitored with Tivoli and I'm not sure what info they're checking for. The ones that are monitored are all dual systems so none will exceed 50% because of Folding.

These are corporate resources, after all, and I do have a responsibility to use them in the best interests of the company.

- Fushigi

 

[Groltz]

Just as quickly as it appeared, the Total Team Adjustment has dissappeared from the Statman page.

Just recorded this screenshot:

 

[Pradeep]

I received four genome work units yesterday, which kept my dial up duallie up and crunching for much longer than normal. So this could have increased production if other people are also connected via lowly dial up.

 

[Handruin]

I now see the corrupted user stats:

http://www.storageforum.net/forums/images/fubar.jpg

 

[Fushigi]

From the stats page:

    F@H
    Rank
12 2,455 Fushigi 323 2,115.36 2.35 432.99 52.39 12.80 13.99 12.80 12.80 
13 2,420 SteveC  447 1,978.09 2.20 148.46 17.98 17.98 0.00  0.00  0.00

I'm curious how SteveC has a higher F@H rank than I do even though my score is higher. Is the rank calculated against something other than score?

On another note, my wife's Athlon 800 is now Folding.

- Fushigi

 

[SteveC]

I'm curious how SteveC has a higher F@H rank than I do even though my score is higher. Is the rank calculated against something other than score?

There's another person with the username SteveC folding for a different team. I can't access the page right now, but if you look here, you'll see my results on top, and his results on the bottom. This page will show my correct F@H ranking, but the Statsman page can't differentiate between two users with the same username.

 

[The JoJo]

YESSS!!! Over 100+ units per day for me! Wohooo.......

 

[Cawood Ledford]

Well done Jo. Third place seems to be yours as long as Clocker does pick up his pace.

 

[Mercutio]

Welcome to our site, Cawood.

 

[Cliptin]

Good job JoJo!

 

[The JoJo]

Thanks, but I hope as many as possible try to pick up their pace .

 

[The JoJo]

And welcome aboard cawood!!! I hope you fold for us? Please?

 

[blakerwry]

woohoo we moved up a place... unfortunately the university of alabama is gaining on us... just last week they were down ~4000 points.. now they are less than 1000 points behind... hopefully most of their computers will be off (and not idle) this weekend and will give us a chance to regroup and attack when they least expect it.

 

[blakerwry]

question: Has anybody else seen this behavior?

http://anime-jennie.com/hdd/folding.gif


In the bottom right of the screen you will see the data grabbed from the unitinfo.txt file in the folding@home directory.... this shows that I downloaded my work unit at 14:24 on Jan 31st(i openned unitinfo.txt and verified that this is waht it said)... however, you will notice that on my clock it is only 10:41 Jan 31st ....

Has anybody else noticed that folding@home is reporting incorrect download times for its work units? this might might explain some weird data I was getting when I was comparing my XP1700+ with my Celeron 2gHz....

 

[SteveC]

It uses UTC time (aka GMT), not local time to keep track of dates.

 

[blakerwry]

hmmm... I just looking in my folding@home log and it thinks it is currently 17:00 hours???? I thought myabe it was getting AM/PM mixed... but that would make the time either 11:00 hours or 23:00 hours... 17:00 hours corresponds to 5:00pm WTF?

 

[Handruin]

blakerwry, what utility is that on your desktop? I've seen it before and for the life of me could never find it. Would you happen to have a link for it?

 

[Dozer]

Blake found it a couple of weeks ago:

http://www.storageforum.net/forum/viewtopic.php?t=1638&highlight=coolmon

 

[Pradeep]

I guess a few people are getting the 400 frame genes now? Worth 19.20 each?

 

[Fushigi]

I guess a few people are getting the 400 frame genes now? Worth 19.20 each?

I get these about 40% of the time any more. About 30% each 100 & 200 frame units. Very rough guess, of course.

- Fushigi

 

[Groltz]

I guess a few people are getting the 400 frame genes now? Worth 19.20 each?

I've never seen a WU worth 19.20, but I get ones worth 19.70 constantly.

 

[Pradeep]

Whoops yes 19.7, i misstyped.

I find that if I let all four of my clients upload their results at the same time and benchmark, I tend to get the bigger genome units, whereas if I submit one unit while the other 3 clients are running, I guess the benchmark thinks my comp is slower and gives me a quicker folding protein. I produce much better when I'm running the big genes, damn dial up.....

 

[Handruin]

Blake found it a couple of weeks ago:

http://www.storageforum.net/forum/viewtopic.php?t=1638&highlight=coolmon

:aok:

Thanks Dozer!

 

[blakerwry]

The Folding@home log file reading is a custom job by myself... im sure others have made their own scripts as well...

here's the code:

+---------
| FOLDING  \
+-----------------------------------+
| Current Protein:                  |
| <NAME=FILEDATA FORMAT="C:\Program Files\Folding@home\unitinfo.txt,3"><PIXELSTART=252>|
| <NAME=FILEDATA FORMAT="C:\Program Files\Folding@home\unitinfo.txt,4"><PIXELSTART=252>|
| <NAME=FILEDATA FORMAT="C:\Program Files\Folding@home\unitinfo.txt,5"><PIXELSTART=252>|
+-----------------------------------+

so.... has anybody else noticed that their folding client shows the incorrect time?

 

[blakerwry]

ah... the time shown in unitinfo.txt and in the log is Grenwich mean time... hmph.. seems the same on another windows computer... I'll have to check later if my linux box does that... i dont think so.

 

[blakerwry]

i just got a northwood based celery 2.0ghz.... I will test it out soon enough.. I'm hoping it will beat my 1700+

It won't.

You were right...

I have been benchmarking my celery 2.0a w/ 512MB pc2100 against my Palomino 1700+ w/ 512MB pc2700 and I found that atleast on the 623 protein the athlon was ~20% faster.

I'm sure the pc2700 helped the athlon... I bet if the celery had the pc2700 and the Athlon had the 2100 things would be equal... but it just shows you how slow the northwood core still is compared to the palomino(1.47gHz) clock for clock.

 

[blakerwry]

All measurements were done with the computer at idle after the WU had been completed.