Folding@Home

Jake the Dog said:
i have the following machines crunching for me for some or most of the time:

1 x P4-1600
4 x P3-800
1 x quad P3-700
2 x P3-667
1 x P3-500
1 x P2-450

i can't seem to get my XP2000 crunching :crap:
Click to expand...

You have quite the farm of crunchers Jake, despite the XP2000+ not working at the present time. You moved by the Bartender faster then a frog fart in a funnel cloud.
 
My, what big frames you have...Sheesh!
Code: 
[02:20:19] Protein: p180_enghomeoGB
[02:20:19] - Run: 120 (Clone 81, Gen 0)
[02:20:19] - Frames Completed: 0, Remaining: 10
[02:20:19] - Dynamic steps required: 500000
[02:20:19]
[02:20:19] Writing local files:
[02:20:19]
[02:20:19] parameters          work/wudata_01.prm
[02:20:19] - Writing "work/wudata_01.key": (overwrite) successful.
[02:20:19] - Writing "work/wudata_01.xyz": (overwrite) successful.
[02:20:19] - Writing "work/wudata_01.prm": (overwrite) successful.
[02:20:20] - Writing "work/wudata_01.key": (append) successful.
[02:20:20]
[02:20:20] PROJECT="work/wudata_01", NSTEPS=500000, DT=2.0000, DTDUMP=100.000000
, TEMP=308.00
[02:20:21] TINKER: Software Tools for Molecular Design
[02:20:21] Version 3.8  October 2000
[02:20:21] Copyright (c)  Jay William Ponder 1990-2000
[02:20:21] portions Copyright (c)  Michael Shirts 2001
[02:20:21] portions Copyright (c)  Vijay S Pande 2001
[03:33:53] Finished a frame (1)
[04:47:22] Finished a frame (2)
 
What's with the great drop-off in our stats? I still have the same number of machines crunching, more or less, but my output has dropped way off. So has evereyone else's, it seems. I'm not seeing any of the dreaded "unable to send results" messages, so where are all those spare electrons going?
 
Check out my frames:

Code: 
[06:10:30] Protein: p180_enghomeoGB
[06:10:30] - Run: 73 (Clone 17, Gen 1)
[06:10:30] - Frames Completed: 0, Remaining: 10
[06:10:30] - Dynamic steps required: 500000
[06:10:30]
[06:10:30] Writing local files:
[06:10:30]
[06:10:30] parameters          work/wudata_02.prm
[06:10:30] - Writing "work/wudata_02.key": (overwrite) successful.
[06:10:30] - Writing "work/wudata_02.xyz": (overwrite) successful.
[06:10:31] - Writing "work/wudata_02.prm": (overwrite) successful.
[06:10:31] - Writing "work/wudata_02.key": (append) successful.
[06:10:31]
[06:10:31] PROJECT="work/wudata_02", NSTEPS=500000, DT=2.0000, DTDUMP=100.000000
, TEMP=308.00
[06:10:31] TINKER: Software Tools for Molecular Design
[06:10:31] Version 3.8  October 2000
[06:10:31] Copyright (c)  Jay William Ponder 1990-2000
[06:10:31] portions Copyright (c)  Michael Shirts 2001
[06:10:31] portions Copyright (c)  Vijay S Pande 2001
[07:59:37] Finished a frame (1)
[09:48:45] Finished a frame (2)
 
Hey folks,

Even though these WU's have very long frame times the number of frames have dropped to 10. It all evens out.
 
Please, I piss on your wussy 500,000 step units. Phear my 5 million bad boy!

Code: 
[August 19 05:24:07] Working on Unit 01
+ Working ...
[05:24:07] Folding@Home Client Core Version 2.47 (June 14, 2002)
[05:24:07]
[05:24:07] Proj: work/wudata_01
[05:24:08] nsteps: 5000000 dt: 2.000000 dt_dump: 1000.000000 temperature: 300.00
0000
[05:24:08] xyzfile:
[05:24:08] "   201  p145_1L2Yunf
[05:24:08]      1  CH3   13.422499   39.088881   13.745021     7     2..."
[05:24:08] keyfile:
[05:24:08] "NOVERSION
[05:24:08] ARCHIVE
[05:24:08]
[05:24:08]
[05:24:08] printout           1000
[05:24:08] writeout           1000
[05:24:08]
[05:24:08] integrate  ..."
[05:24:08]
[05:24:08] Hashes matched on file work/wudata_01.dyn
[05:24:08] ARC file integrity verified
[05:24:08] Restarting from checkpointed files.
[05:24:08]
[05:24:08] Protein: p145_1L2Yunf
[05:24:08] - Run: 41 (Clone 34, Gen 1)
[05:24:08] - Frames Completed: 38, Remaining: 62
[05:24:08] - Dynamic steps required: 3100000
[05:24:08]
[05:24:08] Writing local files:
[05:24:08]
[05:24:08] parameters          work/wudata_01.prm
[05:24:08] - Writing "work/wudata_01.key": (overwrite) successful.
[05:24:08] - Writing "work/wudata_01.xyz": (overwrite) successful.
[05:24:09] - Writing "work/wudata_01.prm": (overwrite) successful.
[05:24:09] - Writing "work/wudata_01.key": (append) successful.
[05:24:09]
[05:24:09] PROJECT="work/wudata_01", NSTEPS=3100000, DT=2.0000, DTDUMP=100.00000
0, TEMP=300.00
[05:24:10] TINKER: Software Tools for Molecular Design
[05:24:10] Version 3.8  October 2000
[05:24:10] Copyright (c)  Jay William Ponder 1990-2000
[05:24:10] portions Copyright (c)  Michael Shirts 2001
[05:24:10] portions Copyright (c)  Vijay S Pande 2001
[05:44:38] Finished a frame (39)
[06:05:04] Finished a frame (40)
[06:25:30] Finished a frame (41)
[06:45:56] Finished a frame (42)
[07:06:22] Finished a frame (43)
[07:26:48] Finished a frame (44)
[07:47:14] Finished a frame (45)
[08:07:39] Finished a frame (46)
 
Groltz said:
Pradeep said:
Please, I piss on your wussy 500,000 step units.
Click to expand...

The exclamation was over how abnormally large the frames are in the p180_enghomeoGB protein. Not the size of the WU.
Click to expand...

Yeah I was just talking FoldingTrashTalk ;) I have come to the conclusion tho that p4's suck at folding compared to AthlonXPs, especially when comparing the fpu units.
 
And at long, long last, we blast past Macrumours and into the top 40 - Team #38 at present, and likely to stay in that spot for quite some time, as the steps to the next few teams are quite large ones.

Special notice for Cliptin, who has re-taken second spot from me once again with a wonderful burst of speed, and to Raven: hi Raven, we don't know who you are but thanks for joining us.
 
I'd just like to add note of sympathy or one particular machine of mine, known as TM1280. TM1280 is one of my dedicated folding machines. I've taken to snaffling any spare hardware around the place and building it into machines. This lets me test out the lying-around hardware, folds some proteins, adds significantly to Tannin's power bill, and drives Kristi nuts because all the powerpoints are taken up. But it means that anytime we want a second-hand system (which is once or twice a week) it's already built and tested; it just needs to have the network card removed and have a suitable hard drive inserted ready for installing whatever OS it is going to ship with.

I use old, small hard drives to fold on, mostly in the 420MB through to 2.1GB size ranges, and when the system sells, pull out the drive ready to be inserted into another machine as soon as I get time to build one. The machines I name according to the only common factor that is preserved: the hard drive.

Anyway, poor TM1280 is a Quantum Fireball TM, 1.2GB in size, and he has been folding for a while now; originally in an XP 1700, more recently in a Thunderbird 1000. Yesterday he finished a work unit and downloaded a fresh one - the infamous P180. On a Thunderbird, that's a long, slow crunch - an 8 pointer. Unfortunately for TM1280, Kristi went and sold his motherboard today. So as not to waste the space, I dragged out another processor and reconstructed the machine, plugged it back into the network. Alas, I have now discovered what TM1280 is working on .... and with his new Pentium II 333 CPU, he is going to take quite a while to finish it!
 
Tea said:
...and to Raven: hi Raven, we don't know who you are but thanks for joining us.
Click to expand...

Likewise.

We are accumulating an awful lot of zeros in the bottom third of our roster.

Roster.gif
 
I'm convinced that there is something wrong at the Stanford end, Groltz. Tea has more or less the same amount of total horsepower that she had a couple of weeks ago when she was turning in 90-odd points a day, but now she is getting 30 and 40 point days.

And I am certainly still crunching, albeit with just a single machine at present, seeing as Kristi has snaffled my other K6-III. So where are my results? I have a 24-hour network conection, no known service issues, and my network config remains just as it was back when Tea was turning in the 90 pointers.

Cliptin seems immune to all of these problems. I have no idea why.
 
Tannin said:
Cliptin seems immune to all of these problems. I have no idea why.
Click to expand...

I only seem to be immune. Baed on this discussion on the FAH help forum, I have determined that it would be best for you to download the new beta to configure it to not try to use SSE code. It seems there is a problem with implementation of this code on AMD procs.

Half of my machines are P4s so I only seemed to not be affected.
 
I don’t know if this is bad, but I haven't updated my folding client because it works. I have yet to have a single problem with crunching and connecting. I'm running 3.01 I believe.

I'm sad to say I will be loosing some crunching; one of the machines I was borrowing has to be given back, so my numbers will decrease. :cry:
 
I haven't had any recent trouble in submitting WU's.

Been running the new beta now for about 10 hours (2 WU's) with no problems.

--Steve
 
Pradeep said:
I have come to the conclusion tho that p4's suck at folding compared to AthlonXPs, especially when comparing the fpu units.
Click to expand...

On my UD client, my Athlon 1.5 GHz is given a performance score of 150, while a Pentium 4 1.5 GHz is assigned as 100. Does that mean the XP is 50% faster than the Williamette P4 for the same clock frequency?

On another note, good to see you F@H guys are in such an exciting dogfight. SR's UD team seems to be stagnanating with members like Tannin and Prof Wizard not active lately. Since upgrading to the XP 1800, I seem to be flying up the charts, but much of it is due to the inactivity of the rest of the UD team.

Oh well, at least you F@H guys are doing something equally productive, unlike hundreds of other computer websites whose teams are mostly crunching SETI@Home data.
 
e_dawg said:
On my UD client, my Athlon 1.5 GHz is given a performance score of 150, while a Pentium 4 1.5 GHz is assigned as 100. Does that mean the XP is 50% faster than the Williamette P4 for the same clock frequency?

On another note, good to see you F@H guys are in such an exciting dogfight. SR's UD team seems to be stagnanating with members like Tannin and Prof Wizard not active lately. Since upgrading to the XP 1800, I seem to be flying up the charts, but much of it is due to the inactivity of the rest of the UD team.
Click to expand...

The quick answer is No. The performance scores are related to ram, Network card, CPU MHz, and HD space and have little to do with actual performance. At one time I actually solved the multiple equations to actually get the coeficients, but I didn't keep the data. However, I will refer you to a post I made earlier here that shows the discrepancies between the UD's capability scores vs. actual performance.

P5-133XL said:
If you think the P133 is me, you're wrong. My P133 does United Devices, not F@H. It is my observation that UD gives much more credit for low-end machines than F@H. The UD algorithum gives credit (even if it does not use it) to ram, network card and HD space and discounts CPU capability.

My P133 has a UD capability score of 47 (a unit commonly takes 106 hours) and generates 6 points/CPU hour
My PII-266 has a UD capability score of 77 ( a unit commonly takes 31 hours) and generates 9.5 points/CPU hour
My PIII-500 has a UD capability score of 101 ( a unit commonly takes 21 hours) and generates 16 points/CPU hour
My Athlon 1900 has a UD capability score of 175 (a unit commonly takes 6 hours) and generates 31 points/CPU hour

Scaling everything to P133=1
P133 = Productivity=1, score=1
PII266 = Productivity=3.4, Score=1.6
PIII5000 = Productivity=5.0, Score = 2.7
Athlon1900 = Productivity=17.7, Score = 5.2

One can see that the UD score per CPU hour given does not even come close to the actual productivity of the individual machines. Thus, if you have low-end machines and one is trying to optimise scoring between the two systems, use the Athlons in F@H and give everything else (especially low-end machines) to UD.
Click to expand...
 
Heh, well my CPU is now rated at 141 instead of 150. We'll see what value it comes up with tomorrow. One thing that bugs me is the rating for the network card. I am always at 70, even though I have a 100 Mbit card. Is it because I didn't force the card to use full duplex?
 
You know, I just realized the significance of that statement. UD is an Intel venture. You don't think they would deliberately penalize crunchers who don't use Intel NIC's do you?
 
That's not true. I have Netgear 10/100 cards in both my machines, and UD rates it anywhere from 117-122. I think it's a combination of network card and internet bandwidth available. The CPU rating also changes depending on if you're using anything CPU intensive when it checks it.

Steve
 
It has been some time; but I was also using a Netgear NIC with a broadband connection. I only registered a 70 with nothing running but UD. I have heard it said that Intel NICs are good; and so I chalked the discrepancy up to cheap hardware not something malicious from Intel. The rest of my machine was old but you'ld think that wouldn't affect it.
 
Finally, back in fifth spot. It's just temporary (until my next ISP hickup), but damn it feels good. Thank you Doug to have slow down to allow me to pass you even if it might only be for a short period of time.

Hey, no matter if Xtreme Systems passes us or not, we should be proud of what we have already achieved. Top 40 in just a few months and with a fraction of the active members most other teams have, that's great.

The few, the proud, the SF F@H members.
 
THE FRIDAY EVENING TABLES OF WOE

Team stats sorted by weekly total: (SF is 25th overall in weekly output)
Table7.gif


Member stats sorted by weekly total:
Table8.gif



------Steve
 
Mr. Monstre

Not to worry. I turned off my Dual AThlon system for about 30 hours. Since the room that computer is in is directly across the hall from the thermostat for my house, I wanted to see how much it is causing the air conditioning in my house to turn on. Indeed, I now know why I can set the thermostat for my house at 77F yet it seems a lot cooler than that in most rooms of the house. The dually rig heats up my room which dumps that heat into the hall which trips the t-stat on. Either way, my house is nice and comfy as it is so I don't mind. When the dually was off, I found myself turning the t-stat down to about 75F to be comfortable.

My dad's Slot-A 700mhz Athlon has been out of comission for over a week. The mainboard (Abit KA7) died and the replacement (Epox 7KXA+)is coming in on Monday. In the meantime, I put my 1.2Ghz T-bird system in his case for him to use. His power supply is only rated for a 1Ghz T-Bird so I have it underclocked to 700mhz just to be on the ultra-safe side.

My father-in-laws Celeron 950 machine does not seem to be on much lately. I'll have to talk to that bitch...his computer is basically mine anyway. :-)

My P3 machine has sucked big time ever since I put Linux on it (it seems). It did not register any points in the whole time my dually rig was down. I put WinXP back on the P3 rig so hopefully it will do a little better now. (The Win32 client seems like a better performer).

What a cluster-F#$%!

I should be running on most cylinders early next week! Thanks for noticing my loss in production though! :-) It's good to point these things to motivate and ensure there are no problems!

C
 
It's been a few days already, but our fellow Adcadet has restarted to contribute to our team. He even past P.W. in the meanwhile. I'm glad we have one less "0" at the bottom of our stats table.

I haven't seen him post for quite some time though.
 
e_dawg said:
Pradeep said:
I have come to the conclusion tho that p4's suck at folding compared to AthlonXPs, especially when comparing the fpu units.
Click to expand...

On my UD client, my Athlon 1.5 GHz is given a performance score of 150, while a Pentium 4 1.5 GHz is assigned as 100. Does that mean the XP is 50% faster than the Williamette P4 for the same clock frequency?
Click to expand...

I was comparing the F@H performance. Compared to even a single process running on my dual XP 1.47, this P4-M 1.6 is probably half as fast.
 
Pradeep said:
e_dawg said:
Pradeep said:
I have come to the conclusion tho that p4's suck at folding compared to AthlonXPs, especially when comparing the fpu units.
Click to expand...

On my UD client, my Athlon 1.5 GHz is given a performance score of 150, while a Pentium 4 1.5 GHz is assigned as 100. Does that mean the XP is 50% faster than the Williamette P4 for the same clock frequency?
Click to expand...

I was comparing the F@H performance. Compared to even a single process running on my dual XP 1.47, this P4-M 1.6 is probably half as fast.
Click to expand...

How significant is the M to your test? ie. Does the M mean anything other than it's BGA or something
 
Pradeep said:
I was comparing the F@H performance. Compared to even a single process running on my dual XP 1.47, this P4-M 1.6 is probably half as fast.
Click to expand...

I was testing a regular P4 1.6 and it was able to crunch SETI at about the same speed as my Athlon XP 1800+. Maybe F@H is more FPU oriented...
 
Cliptin said:
[How significant is the M to your test? ie. Does the M mean anything other than it's BGA or something
Click to expand...

It's a mobile CPU. But I have it set to max performance, no battery saving going on.
 
Congratulations to Cliptin, our new #1! Some serious folding was required to overtake The JoJo, and Cliptin just keeps on going better and better.

Meanwhile, I've slowed to a walk. There is something seriously wrong with the Folding client at present. A week or two ago I had more machines folding than I have ever had before, but they were producing half as many points at best - often less than half as many.

Cliptin says that there are issues with the new core and SSE/3DNow optimisations, which the beta has supposedly fixed. I read up on that (thanks Clippy) but it seems that the issues are more or less restricted to K6 era CPUs - and only one or two of my folders were K6-based. There was a Celeron and a P-III, all the rest were Thunderbirds, XPs or Durons.

So I don't know what's wrong with the folding client, but whatever it is, it stinks. We have been really busy this last week or two and Kristi has sold a good many of my foilders on me anyway. I could build more, but why? It's not as if they are going to generate any significant return until Stanford un-break the client software, and I'm damned if I'm going to frig about loading yet more bloody betas on all those different machines just to see if Stanford have their program working properly again. One or two machines, sure, I'll fiddle about and try to get them working. But ... er .. about 16 machines, no way.

Net result is, until Stanford fix their software problems, I'm not going to build any more folders. Kristi would probably only sell them on me anyway.

Sigh
 
You must log in or register to reply here.
Back
Top