D2OL question for you veterens....

[GeoffS]

I have 2 D2OL crunching-dedicated PCs...

Node Information

Rank...ID.......Cands..Confs..Avg Cands/Day....Best Total............Docking Time
4741..11601...35......700...........17.5...............nci_4012_0283...1 days, 02 hr, 39 min, 26 sec
5827..11594...10......200..............5.................nci_1160_1517...1 days, 04 hr, 32 min, 08 sec

Node 11601 is a Tualatin Celeron 1.0G on an MSI694T mobo w/128Mb ram running Win98SE working on Smallpox.
Node 11594 is an FCPGA P3-1100E (non-Tualatin!) running on an Asus P3V4X w/128Mb ram running Win98SE working on Anthrax.

Neither machine is running anything other than D2OL. Any idea why the P3-1.1G is running so poorly? No config tweeks have been done to the engine on either machine other than to close down the agent. Is this like SETI that some tests take longer than others? Do some CPUs do this better than others?

Thanks!
Geoff

 

[Mardeth]

I dont know the differences between CPUs but the time to complete a WU can have a great difference. It depends on what disease your crunching on.

 

[GeoffS]

Ahh.. the first is working on Smallpox and the second is working on Anthrax.

Geoff

 

[TAandy]

Rotatable bonds and number of atoms make a big difference to the time it takes to complete a unit.
DSpy found here tells you how many atoms and rotatable bonds a candidate has and how long it will take to process.

 

[mastertech01]

Andy has the longest history with it I think, and I have seen that his analysis is pretty much right on target. Ive had some cook off in a few minutes. I had one that took 90 minutes with a P4 2.53 GHZ! OWCHY!

 

[Baldy18]

I had one that took 10 minutes per conformer on a PIII450 and then when it hit about the 13th conformer it took 50 minutes for that one and then went back to cranking out 10 minute conformers. Nobody was using it at the time.

 

[mastertech01]

Originally posted by: Baldy18
I had one that took 10 minutes per conformer on a PIII450 and then when it hit about the 13th conformer it took 50 minutes for that one and then went back to cranking out 10 minute conformers. Nobody was using it at the time.

Good Googly Moses!

:Q

 

[TAandy]

Originally posted by: Baldy18
I had one that took 10 minutes per conformer on a PIII450 and then when it hit about the 13th conformer it took 50 minutes for that one and then went back to cranking out 10 minute conformers. Nobody was using it at the time.

Conformers usually complete in roughly the same time from start to finish of a candidate. The time usually only varies if something more important takes over as I`m sure you already know. Could be your virus checker doing a scheduled scan or something like that ?

 

[bfour]

Usually on a DC project, the programs usually release the processor for higher priority tasks, UNLESS you have an option to manually configure the client to take over a higher priority, as in DF, which you can change, but from what you are saying that is defintely on the odd side!

 

[Baldy18]

Yeah I'm not sure what happened with that one but you may be right about the virus checker. If that is it then I can live with it as long as those nasty viruses stay away.

 

[GeoffS]

Virus checker on a D2OL machine... that's kinda funny....

Nope.. these machines are all dedicated to cracking or crunching or whatever you call what doing D2OL is called That's what confuses me... all but 3 of the nodes are doing nothing, and I mean nothing, else than D2OL... installed the OS and installed D2OL and let them fly.

Geoff

 

[GeoffS]

And it's not really catching up, either

Node Information

Rank...ID.......Cands..Confs..Avg Cands/Day....Best...................Total Docking Time
4075..11601....70.....1,400......23.33...............nci_1111_0655..2 days, 00 hr, 40 min, 32 sec
4653..11594....40.......800.......13.33...............nci_1097_0227..2 days, 00 hr, 14 min, 36 sec

Geoff

 

[mastertech01]

Geoff, some days you end up with a pileup of hard ones to crunch. I have a P4 and a PIII wimpy puter running and some days the two combine for what the P4 can do in one day. It just works out that way. Some days they combine to do 25 more.

 

[mastertech01]

By the way Geoff, your lousy little 18 nodes are tearing deep holes in the individual stats world wide! You are running 6th in the world and everyone above you are running many more nodes than you are. In fact I believe DPC_LawnMooherMen doesnt even belong in the individual stats as that is actually a Team of crunchers using a single name with over 175 nodes..

 

[Baldy18]

Originally posted by: GeoffS
Virus checker on a D2OL machine... that's kinda funny....

Yeah it's my fathers machine so you never know what he is going to get into.

I don't have the luxury of having dedicated crunchers.(ok I do have one dedicated cruncher but sometimes it becomes a LAN box)

 

[GeoffS]

mastertech01 - 6th? Whoa! I hope that by no later than Monday I'll have a dedicated XP1600+ and a dedicated XP1700+ running also... running into the same problem I had with RC5... not enough cases! Here's a fairly recent pic of my cubicle at work (there are actually 2 more in there presently), and I have 3 in other cubes and 3 at home

Geoff

 

[mastertech01]

Geoff, HERE YOU ARE. Sweet pic Geoff! Im in the same boat, gotta get a few cases ASAP.

 

[Daaavo]

The question I've got is, what is the better processor for running D2OL?? An AMD or a P4???

 

[Baldy18]

I'm not sure but I don't think the processor make that much of a difference at the same MHz level.

I hear they have/had? a good deal going on with cheiftec cases in the hot deals forum.

 

[mastertech01]

Originally posted by: Baldy18
I'm not sure but I don't think the processor make that much of a difference at the same MHz level.

I hear they have/had? a good deal going on with cheiftec cases in the hot deals forum.

Thanks Baldy18, Ill check out that deal..

Dave I think the OS is more the performance factor in D2OL, as it seems Linux is supposed to be a large difference.

 

[GeoffS]

CPU.............Node....Cands....Conf.....Avg Cands/Day
XP1600+....11590.....143.......2,860.........47.67
Cel 1.4G.....11591.....121.......2,420.........40.33
XP1600+....11581.....117.......2,340.........39
Cel 1.4G.....11587.....112.......2,240.........37.33
P4-1.5G......11643......80.......1,600..........26.67

All are running either Win98SE or Win2K with a small hd and at least 128Mb ram. Dunno if this is conclusive, but it is my (limited) experience.

Geoff

 

[TAandy]

There is some D2OL benchmark info here if you`re interested, but ignore the references to Eaglechilds site as he`s now changed it to digital camera stuff.

 

[MereMortal]

You really have to run things for a while to get a good comparison, due to the differences in candidate size. But the following gives some indication of how things are (over a week time frame):

CPU.......................OS.......Avg Cands. / Day
XP2200@1890....RH8.0...........107
MP1800@1533...RH8.0.............88.44
MP1800@1533...RH8.0.............86.67
MP1900@1600...RH7.3.............64.75
MP1800@1533...XPPro.............54.44
TB1300@1400....XPPro.............52.33
MP1800@1533...XPPro.............51.89

The MP1900 might be anomalous in that is running against a large job 1.5+GB an has been niced. It seems that on my dualies, the linux client is running about 60% faster than the windows client.

 

[GeoffS]

:Q WOW! :Q

 

[Baldy18]

Dang my nodes suck!

CPU.......................OS.......Avg Cands. / Day
XP1800@1533....Win2K...........62.23
PIII933@1000.....XPPro............17.74
PIII500..................Win2K............15.81
PIII450..................Win98............14.24
PII400....................Win98...........13.36