Daily Update.

[DaveX]

My DF stopped uploading today with the introduction of the new client. Anybody else having u/l problems?

[Groth]

The hospital that sponsors DF has been having network problems on the weekends for a while now. The speculation is script-kiddies, but who knows. Ain't a big deal, just let the client buffer results. The problems will get ironed out shortly.

[Arcturius]

Is everybody still disgusted with the upload troubles? We're down to 20 active users. :shrug:

[satanicoo]

Im only folding because it doesnt bother me.
Maybe its time to say goodbye.

[jaydee]

My power bill went from $70 to $20 once I stopped. That coupled with the issues keeps me from returning.

[Groth]

I'm still folding away. We need some loud, enthusiastic cheerleader types to stir up interest.

As for worth it, CASP6 (a world wide protein folding competition) should be a good indicater. In CASP5, Distributed Folding was solidly middle of the pack on all the proteins (most of the teams were inconsistent performers; good at some, bad at others) -- pretty good for a novice program with little funding. It'll be interesting to see if they've improved or not.

Damn, Arcturius! You're gainin' on me fast!

[Arcturius]

Originally Posted by Groth

I'm still folding away. We need some loud, enthusiastic cheerleader types to stir up interest.
Damn, Arcturius! You're gainin' on me fast!

Only about 40 more days until a new protein...hopefully it will be another small one and I can _really_ catch up...

Did anyone ever test to see if output from HT CPUs was greater with 1 or 2 processes per CPU?

[Groth]

If we get a small protein, I won't be the only one you threaten. Actually, there's no way in hell I can hang on to fourth for more than another week -- two crunchers have been shut down to give their memory to the main machine, and the main machine has been busy crunching thermal sims.

Been ages since I owned anything Intel, don't know about HT. You might want to ask in the Official Project Forum.

[Arcturius]

Originally Posted by Groth

If we get a small protein, I won't be the only one you threaten. Actually, there's no way in hell I can hang on to fourth for more than another week -- two crunchers have been shut down to give their memory to the main machine, and the main machine has been busy crunching thermal sims.

Been ages since I owned anything Intel, don't know about HT. You might want to ask in the Official Project Forum.

Thermal sims? Watcha up to?...

Sadly, there were only a handful of posts on there about HT...some said 5% increase, some said 70%, some said performance penalty...but nobody seemed to be running Linux. :shrug:
*sigh*
Looks like I'll have to do the testing myself. ;p

[Groth]

Originally Posted by Arcturius

Thermal sims? Watcha up to?...

Nothing useful yet, but I'm hoping. thread

Make sure you publish them HT test results!

[satanicoo]

putted a 2.4 at 3.25 and crunshing.
no HT unfortunately.

Hopefully will have a CPU like that, and the i can tell you that.

[Arcturius]

I wanted to find out whether HT actually provides a boost for the current DF client under Linux (somebody else can test Windows, I don't feel like dealing with GUI overhead and a bunch of useless processes taking up resources )

Plan: I figured I needed to complete roughly four full sets to get results I could be reasonably confident in. 1 set/client 4x clients, 2 sets/client 2x clients (HT still on), 2 sets/client 2x clients (HT off).
If somebody has input/criticism on my methods/assumptions lmk.

So...I started with the run of 4x clients on 4x virtual CPUs under Fedora Core 1 (2.4.22-1.2115.nptlsmp)

Total run time: 258575s (71:49:35)
Client 0: full cycle, 0:2931; 37931
Client 1: full cycle, 0:1754; 36754
Client 2: 248:29; 34829
Client 3: full cycle, 0:461; 35461
Total: 144975 structures
Avg rate: 33.64 structures/min

Working on 2 sets/client 2x clients (HT still on) right now...hopefully I'll finish before the CASP protein switch.

[Arcturius]

Also, does anyone have any idea why production went through the roof, starting Friday? I haven't seen anything about a switch to a smaller protein, and the clients I checked were still working on proteins with 145 residues... :shrug:

[Groth]

I don't think anything happened Friday. Looks you've got a good start on the hyperthreading question.

We did change to the first of the CASP6 proteins today. 126 amino acids long, with a larger Generation-0.

With CASP6 going on now, we'll probably be switching proteins every Tues.

[Arcturius]

The 2x clients on 4x virtual CPUs under Fedora Core 1 (2.4.22-1.2115.nptlsmp) run is obviously slower, so I have stopped it early.

Total run time: 157167s (43:39:27)
Client 0: 241:54; 34154
Client 1: 220:72; 32072
Total: 66226 structures
Avg rate: 25.28 structures/min

About to start 2 sets/client 2x clients (HT OFF). Good thing I saved a copy of my DF directory so I can keep running with -if. ;p

[Arcturius]

2 sets/client 2x clients on 2 CPUs (HT OFF) under Fedora Core 1 (2.4.22-1.2115.nptlsmp) run:

Total run time: 283030s (78:37:10)
Client 0: full cycle, 233:71; 68371
Client 1: full cycle, 198:31; 64831
Total: 133202 structures
Avg rate: 28.24 structures/min

I just realized that my first run spilling over into the 'easy' gen0 structures could skew the results some, so I have a 'handicapped' number below, where I made my calculations using the same total run time, but did not give credit for any of the gen0 structures.

In the end...it looks like HT gives a boost of about 14.9-19.1% (when used properly). When used improperly, it yields about a 10.5% performance defecit (this assumes an otherwise idle machine, very likely in Linux, highly unlikely in Windows, due to GUI overhead).

Well, I'm off to start a few more clients on some Xeons... ;p

33.64 HT ON, 1x client per virtual CPU
32.45 HT ON, 1x client per virtual CPU (handicapped)
28.24 HT OFF, 1x client per physical CPU
25.28 HT ON, 1x client per physical CPU

[jaydee]

Time for a update I guess...

New protein today. 119 my software just updated to it...

[Nightfall]

Finally... were done with CASP, now back to our normal protein schedule

[Groth]

I take it you don't consider variety to be the spice of life?

[Nightfall]

I just prefer not to be doing updates every week, it's time consuming to keep track of 15 machines to see if they all auto-updated every tuesday and then manually update them if they don't. I just assume to let them fold and forget about it Well that, and watch my points rack up

[Groth]

If only my points would rack up that easily. I lost a mobo and UPS to lightning last week, so I have even less hope of staying ahead of you.

[Nightfall]

Originally Posted by Groth

If only my points would rack up that easily. I lost a mobo and UPS to lightning last week, so I have even less hope of staying ahead of you.

I won't catch you for another 3 months though! Sorry to hear about your loss of hardware. Always sucks when you loose things due to causes you can't control

[Groth]

Eh, no big loss. Dismantled the UPS, got a serious 12V transformer (50 amps). Mobo is undergoing a Frankenstein treatment to become a test platform.

I peaked this spring at six computers folding, now I'm down to two and half. In three months I'll be generating negative points.

[Nightfall]

Originally Posted by Groth

Eh, no big loss. Dismantled the UPS, got a serious 12V transformer (50 amps). Mobo is undergoing a Frankenstein treatment to become a test platform.

I peaked this spring at six computers folding, now I'm down to two and half. In three months I'll be generating negative points.

We are the only ones who would take a fried mobo and try and do something with it. A [H]'er would prolly just toss it in the garbage

[Nightfall]

About to hit 100 million mark... woohoo