639 replies to this topic

[dnamechanic]

... Now my brother bought a PS3, so it's running 24/7.

Great, that will be a bice boost for the team.

Am I going to be listed twice?

Not sure. Seems like this sort of thing has happened before. Looks like EOC is combining the points from both entries and counting as one contributor. It might clear up at Stanford stats in a few days. An inquiry could be posted at the Stanford Folding Forums.

[dnamechanic]

Am I going to be listed twice?

An inquiry was posted at the Stanford Folding forums.

From the response, it seems as if you will continue to have have two entries in the Stanford Stats.

-------------------------------------------------------------------------
Selected stats from Extreme Overclocking:

Team Rank - 183 (new record)
Current active contributors - 81
Points per Day (Avg. ppd) - 53,470 (new high)

Other F@H related topics at the Longevity Meme Team Forum

[dnamechanic]

Welcome PS3aiguy,

You are the 290th contributor to sign on with The Longevity Meme F@H team.

I did not see your username in the F@H Prize Registration Thread, please correct me if I missed it.

Registration is a requirement for Prize participation.

Thank you.

-------------------------------------------------------------------------

Selected stats from Extreme Overclocking:

Team Rank - 183
Current active contributors - 82
Points per Day (Avg. ppd) - 51,768

Other F@H related topics at the Longevity Meme Team Forum

[mitkat]

Thanks dna,

I appreciate you checking it out for me. At least I'll be pumping out more #'s this time!

Am I going to be listed twice?

An inquiry was posted at the Stanford Folding forums.

From the response, it seems as if you will continue to have have two entries in the Stanford Stats.

-------------------------------------------------------------------------
Selected stats from Extreme Overclocking:

Team Rank - 183 (new record)
Current active contributors - 81
Points per Day (Avg. ppd) - 53,470 (new high)

Other F@H related topics at the Longevity Meme Team Forum

[dnamechanic]

Welcome Abolitionist,

You are the 291st contributor to sign on with The Longevity Meme F@H team.

I did not see your username in the F@H Prize Registration Thread, please correct me if I missed it.

Registration is a requirement for Prize participation.

Thank you.

-------------------------------------------------------------------------

Selected stats from Extreme Overclocking:

Team Rank - 183
Current active contributors - 82
Points per Day (Avg. ppd) - 51,864

Other F@H related topics at the Longevity Meme Team Forum

[dnamechanic]

[sbehrens]

Just signed up with a PC I will fully dedicate for folding. Dual Xeon Quad Core 3.0Ghz. WOO!

[dnamechanic]

Just signed up with a PC I will fully dedicate for folding. Dual Xeon Quad Core 3.0Ghz. WOO!

Thanks sbehrens, sounds very good. Looking forward to seeing your username here and here.

Please consider registering for the F@H Prize in the F@H Prize Registration Thread.

Registration is a requirement for Prize participation.

[dnamechanic]

Thanks for registering for the F@H Prize, quintin3265.

You are a long-term contributor to the TLM F@H team, for the record:

Your initial score for the F@H Prize is 161,468 points.

Many thanks to you for your efforts.

[dnamechanic]

...I will fully dedicate for folding. Dual Xeon Quad Core 3.0Ghz. WOO!

You are probably already aware of this, if you have the inclination, those 8 cores will likely produce the most points folding SMP work units. The Linux SMP client seems to be stable and well developed (also slightly more efficient than Windows SMP). The Windows SMP client may require a bit more effort to get running well, but many contributors use it and generate lots of points.

A little late in the game, but I'll put in 8 cores. user: sbehrens

Your contributions will undoubtedly help the team and the cause. Also, a new competition for the F@H Prize will begin again in about a month (around July 1st). Maybe you will have time to get your dual-quads fully tweaked!

[sbehrens]

...I will fully dedicate for folding. Dual Xeon Quad Core 3.0Ghz. WOO!

You are probably already aware of this, if you have the inclination, those 8 cores will likely produce the most points folding SMP work units. The Linux SMP client seems to be stable and well developed (also slightly more efficient than Windows SMP). The Windows SMP client may require a bit more effort to get running well, but many contributors use it and generate lots of points.

A little late in the game, but I'll put in 8 cores. user: sbehrens

Your contributions will undoubtedly help the team and the cause. Also, a new competition for the F@H Prize will begin again in about a month (around July 1st). Maybe you will have time to get your dual-quads fully tweaked!

Yeah afte a bit of tweaking I am not quite to 7 cores. I'm currently using the Windows SMP beta client with a bit of config to try to get it to use 100% of 7 cores. Right now I am using 100% of 4 cores, so any suggestions are welcome. I could load centOS on the box and go linux but I wanted to get this thing up and running with Windows if possible.

Let me know if you have found anything special for getting it to use all of the cores.

-Scott

[sbehrens]

Oh I should mention I am using client 5.9.1. I tried the Beta 6 client but it only ran on 1 core. Maybe I missed something....

[dnamechanic]

... Right now I am using 100% of 4 cores, so any suggestions are welcome.

As far as I know, the current release of clients are only capable of loading four cores.

You could run two instances of the SMP client, four cores per client (analogous to this discussion).

- Place folding clients in clients separate folders
- Start clients with -local flag
- Assign machineid's differently in client.cfg file, such as machineid=1, machineid=2 (client.cfg setup menu appears when client is first activated)

A thread, here, at Stanford Folding Forums sounds like your situation.

I am using client 5.9.1. I tried the Beta 6 client

Looks like there are two Windows clients available:

- 5.91 beta6
- 5.92 beta (does not currently support 64-bit Windows)

I could load centOS on the box and go linux but I wanted to get this thing up and running with Windows if possible.

You may want to use Linux later to max out points-per-day (PPD).

There are a few other tricks you may want to try later to maximize PPD.

I think your approach with Windows client is good for now.

[naapi]

... Right now I am using 100% of 4 cores, so any suggestions are welcome.

As far as I know, the current release of clients are only capable of loading four cores.

The newest Linux SMP Beta client allows you to use more than 4 cores - you can set the number of cores in config. Not sure about Windows, though.

You could run two instances of the SMP client, four cores per client (analogous to this discussion).

Actually, it seems that 2x SMP uses only 60-70% of 2xQuad core. See here:
http://foldingforum....php?f=12&t=2671

This would be analogous to the fact that 1xSMP uses only about 70% of a Quad core. Going with 2xSMP takes you to 90% of a Quad. If you set affinity of 1st SMP instance to CPU cores 0 and 1 and second SMP instance to CPU cores 2 and 3 you may get additional 1 or 2% of performance.
Adding 3rd SMP instance does not improve CPU utilisation beyond 90-92%. 3 instances give about the same PPD as 2 instances or less, if you get a memory hungry SMP project (such as 3065). All these CPU utilisations are under Linux (Ubuntu 7.10)

Now, with 8 cores (2xQuad) things are likely even more complicated, but I would check PPD performance with the the following combinations:
1) 1xSMP
2) 2xSMP
3) 3xSMP
4) 4XSMP
5) 2XSMP with first instance bound to CPU cores 0,1,2,3 and second instance with CPU cores 4,5,6,7.
6) 4x SMP with respective SMP instances tied to CPU cores 0&1 then 2&3 then 4&5 then 6&7

Now, I do not know which one yields the best results, but my guess would be combinations 4) or 6). This is very new and very experimental - I have not seen any benchmarks concerning 2xQuad (Octacore) in neither foldingforum.org nor extremeoverclocking.com forum (these two forums are the best IMHO). Any results of such experiments would be of importance to the folding community and would be very welcome to be posted in one of these forums.

Of course the remaining comments of dnamechanic concerning different directories for different FAH clients and machine ID numbers apply on top.

Sbehrens - if you have any more questions - I will be glad to help. 2xQuad could give you as much as 6000-6500 PPD, if used 24/7 and efficiently set up. For the benefit of mankind, our team and yourself, I hope you will succeed

Happy folding,

Maciek.

[dnamechanic]

For the benefit of mankind, our team and yourself, I hope you will succeed

Thanks naapi, for your informative comments.

Here is an early article that I appreciated:

Folding on the Intel V8 - Eight Cores of F@H

The performance numbers shown in the article are good relative indicators. Based on experience with single-quad systems, one can likely obtain higher PPD than indicated, in some cases considerably higher, folding the currently available work units.

...which one yields the best results, but my guess would be combinations 4) or 6).

I agree.

Happy folding, Maciek

Yes, indeed!

[sbehrens]

... Right now I am using 100% of 4 cores, so any suggestions are welcome.

As far as I know, the current release of clients are only capable of loading four cores.

The newest Linux SMP Beta client allows you to use more than 4 cores - you can set the number of cores in config. Not sure about Windows, though.

You could run two instances of the SMP client, four cores per client (analogous to this discussion).

Actually, it seems that 2x SMP uses only 60-70% of 2xQuad core. See here:
http://foldingforum....php?f=12&t=2671

This would be analogous to the fact that 1xSMP uses only about 70% of a Quad core. Going with 2xSMP takes you to 90% of a Quad. If you set affinity of 1st SMP instance to CPU cores 0 and 1 and second SMP instance to CPU cores 2 and 3 you may get additional 1 or 2% of performance.
Adding 3rd SMP instance does not improve CPU utilisation beyond 90-92%. 3 instances give about the same PPD as 2 instances or less, if you get a memory hungry SMP project (such as 3065). All these CPU utilisations are under Linux (Ubuntu 7.10)

Now, with 8 cores (2xQuad) things are likely even more complicated, but I would check PPD performance with the the following combinations:
1) 1xSMP
2) 2xSMP
3) 3xSMP
4) 4XSMP
5) 2XSMP with first instance bound to CPU cores 0,1,2,3 and second instance with CPU cores 4,5,6,7.
6) 4x SMP with respective SMP instances tied to CPU cores 0&1 then 2&3 then 4&5 then 6&7

Now, I do not know which one yields the best results, but my guess would be combinations 4) or 6). This is very new and very experimental - I have not seen any benchmarks concerning 2xQuad (Octacore) in neither foldingforum.org nor extremeoverclocking.com forum (these two forums are the best IMHO). Any results of such experiments would be of importance to the folding community and would be very welcome to be posted in one of these forums.

Of course the remaining comments of dnamechanic concerning different directories for different FAH clients and machine ID numbers apply on top.

Sbehrens - if you have any more questions - I will be glad to help. 2xQuad could give you as much as 6000-6500 PPD, if used 24/7 and efficiently set up. For the benefit of mankind, our team and yourself, I hope you will succeed

Happy folding,

Maciek.

Thanks Macjek. My goal is to take this as efficent as possible. I also started up the smp on an ultra fast dual core box and. In the future, I will potentially be able to get the client on a few more dual cores and potentially 1 more 2xquad core. I work as a security consultant and have accumulated a surplus of machines, which I can donate time to folding.

My current configuration is using 2xSMP (2 directories, local, different machine ID's. I have reached 4500PPD (using fahmon), which by your stats is less than 75% optimal. There is no affinity between which SMP clients run on which cores and I am not sure where to set affinity in the configuration files. In addition, I am more than willing to try different configurations as well as document the efficiency of there configurations. I will post some statistics of my 2xSMP results tomorrow morning when I am back at my office.

Macjek and Dnamechanic any other suggestions are welcome!

-Scott

[sbehrens]

Found the affinity code syntax, I'll make adjustments tomorrow.

-Scott

[dnamechanic]

This post is a continuation of a discussion started here.

... I'll continue running F@H although I'm not a "strong" contrubutor

Thanks Shonghow, in a relatively short time you have moved up, solidly into the top half of the folding team contributors.

...it doesn't always continue working with the same work unit when restarting the computer.This has happened at least 3 times for me and every time when let's say I've been running the client for several days with the same work unit being nearly finished and restart the computer,Well then the nearly finished work unit is gone and it suddenly starts a new one and I get no points.

This can be frustrating. It is fairly unusual for WindowsXP to corrupt the files when restarting. If folding with Windows98 or Windows ME it can happen almost every time. The best thing to do, regardless of the type of operating system, is to stop the client with the client control before restarting the computer. For the windows Graphical client, right-click and select 'Quit' in the menu. For the Windows console client, stop it with 'Control-C'. Granted, this may not be possible if some other application has 'frozen' the computer.

Please mention if this continues to be a problem.

Edited by dnamechanic, 22 May 2008 - 12:35 AM.

[sbehrens]

Quick question. I just saw some of my points post to the team homepage. It says on my individual profile that I have one active processor, however I am using 8 cores on one machine and 2 on another. Maybe I'm missing something....Also what is the frequency that the folding@home page updates with statistics? My average PPD is suggested by fahmon to be 4500 but I only see 1700. Thanks in advance!

_Scott_

[dnamechanic]

Quick question. I just saw some of my points post to the team homepage. It says on my individual profile that I have one active processor, however I am using 8 cores on one machine and 2 on another.

Yep you're in.

Welcome and thanks.

Stanford stats reports:

Rank, User, Score, WU

199 sbehrens 1732 1

Stanford reports receiving one completed work unit. The report of one active processor is consistent with the single work unit reported complete. Stanford doesn't count cores unless they happen to be running a client per core, as in the regular work units. They count clients, as determined by machineid (clint.cfg) and UserID (Win registry). So far, only four at most, of your cores (representing one client) have submitted completed work units.

Maybe I'm missing something....Also what is the frequency that the folding@home page updates with statistics? My average PPD is suggested by fahmon to be 4500 but I only see 1700. Thanks in advance!

They currently update every two hours, starting just after the even hour number (here in CDT zone), 12, 2, 4, etc. (convert to your zone). The updates usually take 40-45 minutes, during which the site has a page up that says the stats are inaccessable.

FahMon calculates PPD bases on recent frames; can be set to single frame, or three frames, or complete work unit. Setting is located in FAHMon>Preferences> Monitor.

FahMon calculates PPD regardless if a work unit is completed and submitted, or not. Stanford lists only completed work units.

[dnamechanic]

Welcome sbehrens,

You are the 292nd contributor to sign on with The Longevity Meme F@H team.

Thank you
-------------------------------------------------------------------------

Selected stats from Extreme Overclocking:

Team Rank - 183
Current active contributors - 81
Points per Day (Avg. ppd) - 52,048

Other F@H related topics at the Longevity Meme Team Forum

[dnamechanic]

Thanks for registering for the F@H Prize, CosmicV.

You have been a contributor for a while. For the F@H Prize record:

Your initial score for the F@H Prize is 29,469 points.
Many thanks to you for your efforts.
-------------------------------------------------------------------------

Selected stats from Extreme Overclocking:

Team Rank - 180 (new record)
Current active contributors - 81
Points per Day (Avg. ppd) - 54,671 (new high)

Other F@H related topics at the Longevity Meme Team Forum

[dnamechanic]

My average PPD is suggested by fahmon to be 4500 but I only see 1700. _Scott_

According to EOCs stats, you are crunching through the WUs

Since you are probably running SMP clients with WinXP, you or others, might be interested in the SMP Affinity Changer.

The Affinity Changer was developed by Rilian, of a Ukraine folding team. It is a utility that automatically sets affinity for the folding cores. The original version examined RAM allocation and assigned cores accordingly every 10 minutes. It is a small application not requiring much RAM itself and very little CPU usage. (overhead). I have used earlier versions of Affinity Changer and found they did increase PPD compared to either not setting affinity manually, or not setting it optimally. At minimum, the Affinity Changer saved the effort of finding optimal settings.

Only recently has a version become available for dual-quads, discussed here at Folding Forums.

[naapi]

My average PPD is suggested by fahmon to be 4500 but I only see 1700.

_Scott_

Scott, I see you are currently approaching 4000 PPD. This is somewhat lower than I would normally expect (6000 PPD). However, I see that this is coming from unususlly high number of 2665 WUs being fed to your double quad. For the last 6 days I see your scores as: 3840, 3840, 3680, 3680, 3840 and 5440 which I read as 2x1920, 2x1920, 1920+1760, 1920+1760, 2x1920 and 2x1760+1920. This gives us 9x1920 and just 4x1760. What's important is that 2665 are yielding much less PPD than 2605/2653, 3062/3064 or even 3065 (2144 pointer). My laptop is being fed 2665 for couple of days now and I only get ~576 PPD, vs 845 PPD I used to get on 2653 WUs. This is a drop of some 32%, unfortunately. I'm afraid there's nothing we can do, as number of points awarded for this project has already been raised by 50% after complaints from community.
This is just to let you know, that it looks to me that your rig seems to be set up properly, you just seem to have bad luck with regards to your WUs.

Regards,

Maciek.

[VictorBjoerk]

Just wanted to mention that my client stopped working when I had finished about 4200 of 5000 work units and then started from zero again picking up a new work unit from zero.I had been running it for 6 days so it was a little irritating.I got no points for it.

What should I do to prevent this from happening?It happens when I turn off my computer but not always.

[dnamechanic]

...my client stopped working when I had finished about 4200 of 5000 work units and then started from zero again picking up a new work unit from zero.I had been running it for 6 days so it was a little irritating.

Sorry you lost your work, Shonghow.

Sounds like your work unit was 84% complete, 4200/5000 frames or steps of the unit.

Is your operating system WinXP? or some other version?

When something like this happens it is often the result of corrupted files in the F@H folder. An OS can corrupt the files if shut down at inopportune time in the folding process.

I got no points for it.

Stanford awards points for partiallly completed work units, whenever enough of the folding records survive to determine what work unit, and point value, and how long the work unit was worked. If you receive no points at all then the work files must have been obliterated.

What should I do to prevent this from happening?It happens when I turn off my computer but not always.

Did you stop the client with the controls available within the client, before turning off the computer? Which client are you using, Graphical Windows, version 5.03?

Normally, stopping and starting the F@H client causes no problems if it is stopped within the client itself. WinXP seems pretty good at closing applications during shut down without corrupting files, but Win98 and WinME could regularily corrupt files by shutting down the OS without first stopping the client software. In any event it is best to stop the client with the client command "Quit" before turning off a computer.

In the folding folder is a file called FAHlog.txt, this file contains records of events that occur during the folding process. FAHlog.txt can be valuable for determining what happens when problems do occur.

Here is an excerpt from FAHlog.txt:

[00:11:35] Preparing to commence simulation
[00:11:35] - Looking at optimizations...
[00:11:35] - Created dyn
[00:11:35] - Files status OK
[00:11:35] - Expanded 18105 -> 126519 (decompressed 698.8 percent)
[00:11:35]
[00:11:35] Project: 4545 (Run 66, Clone 92, Gen 0)
[00:11:35]
[00:11:35] Assembly optimizations on if available.
[00:11:35] Entering M.D.
[00:11:41] Protein: p4545_T0396_2-16_minout
[00:11:41]
[00:11:41] Completed 0 out of 2500000 steps (0)
[00:18:30] Writing local files
[00:18:30] Completed 25000 out of 2500000 steps (1)

You may wish to start examining FAHlog.txt on your computer, looking for anomalies. If you have another loss of work unit, try to find the area in the FAHlog.txt where it indicates a problem and post it here. The FAHlog.txt can contain records from previous work units. You may find in the current log where the client was working step 83 or 84 of previous work unit, then examine log contents afterward. This may provide clues as to what happened.

[dnamechanic]

...you just seem to have bad luck with regards to your WUs. Maciek.

Yes, this happens.

It appears that Stanford does some assignment according to hardware (2 core vs. 4 core, WinXP vs. Linux).

Seems that systems configured as dual-core running the Linux client currently get some good work units (2605/2653) for PPD. Of course, things can change.

Scott, earlier you mentioned going to Linux. If you decide to experiment, a fairly good option is to use VMware to run Linux virtual machines under Windows. The free version of VMware server limits the number of cores for each virtual machine to two (this happens to be convenient for folding machines Run two virtual machines with two cores each for each quad. In your case, guess it would be 4 virtual machines of 2 cores each.

[dnamechanic]

Thanks for registering for the F@H Prize, frenchhorn1.

You have been a while. For the F@H Prize record:

Your initial score for the F@H Prize is 11,017 points.

Many thanks to you for your efforts.

[dnamechanic]

Welcome toxik,

You are the 293rd contributor to sign on with The Longevity Meme F@H team.

I did not see your username in the F@H Prize Registration Thread, please correct me if I missed it.

Registration is a requirement for Prize participation.

Thank you.

-------------------------------------------------------------------------

Selected stats from Extreme Overclocking:

Team Rank - 176 (new record)
Current active contributors - 78
Points per Day (Avg. ppd) - 47,857

Other F@H related topics at the Longevity Meme Team Forum

[Live Forever]

Up to rank 176, woo woo! We are going to cross that 100 barrier some day, I feel confident.