NewbieQ - Which platform is most effective?

[Qinsp]

Greetings,

I just started playing with F@H. I noticed it on my PS3 when I was trying to find a config setting.

I have many Windows machines (3.11-Win7 pro), an Apple MBP with Snow Leopard, a few 32-bit Linux machines, and a couple of PS3's (and a couple CP/M machines + a couple DOS machines still running daily Can't seem to throw things out that still work.)

None of my suitable machines have high end graphic cards, but all are multi-core.

I started with a slim PS3, which takes about 5-6 hours to finish a WU. AOK.

Then I tried to run it with the MBP. Even though the computer is always on and plugged in, it appears to only run F@H when I'm actively using the computer. It might not finish before the deadline listed.

OK, now the question: How do I monitor how much real F@H work is getting done regardless of platform? If I want to leave 2-3 computers running, I'd like to know which ones would yield the most "bang-for-da-buck". I think I can figure out the power usage by using a UPS monitoring program (for PS3, run computer+PS3, record watts, run computer solo, subtract).

[Zagen30]

Welcome to the forum, Qinsp.

See http://foldingforum.org/viewtopic.php?f=14&t=52 for a list of 3rd party monitoring clients (plus other stuff you probably don't need right now). When pointed at the appropriate folders containing running F@h clients, they will tell you how many Points Per Day (PPD) each is earning, among other information. PPD is the metric that most people pay attention to as it roughly corresponds to how much science each client is doing.

The PS3 cannot be monitored by any of those clients, but when run 24/7 it yields around 900 PPD, no matter what model you have. I believe the slim PS3's draw around 90W while folding, so they're around 10 PPD/W. Older PS3's are worse in that respect.

I don't know much abouts Macs or the Mac client, but in case it helps someone else diagnose the problem, what's happening to the Mac client when you're not actively using your MBP?

[Qinsp]

Thanks! Wow, a lot of monitoring stuff out there.

The MacBookPro is a low-power-consumption notebook, and I'm not much of Mac-o-phile (Macaroni?). The F@H app says 190+% CPU when I look at it, so it's working when watched but progress stops when I leave the room (kind of like having an employee...)

I imagine it's power setting, which means I'll probably have to read the instructions for the MBP ... someday. I just use it for web surfing.

If a slim PS3 only sucks down 90w, then I'll leave it on. Our "old" PS3 does a great space-heater impersonation. In a small room it will raise the room temps a good 10-20° F. Playing Rockband with it makes you sweat like real stage lights, further enhancing the realism.

I'll play with those utilities and figure out which computers are best for 24/7 use.

Thanks again.

[Qinsp]

DOH!!

When I left for work, the MBP was 54% complete on the WU. I left Google Chrome open (need to test this) and now it's 76% 8 hours later.

So it's working as intended now.

Google Chrome is an interesting case. It does infrequent random crashes if left open on an XP quad core machine that has 2 GPU's, and on some Linux-based routers and devices, it will not process pages correctly. I like it, but it's quirky.

[uncle fuzzy]

The top point-getter/science-doer is the 8-core+ -bigadv SMP. Next are the high-end nVidia gpus. Now we get into your hardware- multi-core cpus. SMP on a quad, running 24/7, will do 3K PPD+ (points per day). On duals it will do 800-3000. The more cores, and the faster the clock, the higher the points. My [email protected] does just over 4K PPD. The [email protected] does up to 6500 PPD.

You'll probably want to use Window XP and up, quads if you got 'em, duals if you don't. Without newer video cards, running the SMP will be your best bet, if you run 24/7. If you don't want to leave them running, you can install one CPU client for each physical cpu core in your machine. The deadlines on the CPU WUs are long enough you can run just a few hours a day, and still finish them in time.

If those linux boxes are multicore, they can also run the SMP client.

[Zagen30]

If those linux boxes are multicore, they can also run the SMP client.

If TC installs a 64-bit version. SMP doesn't work on 32-bit Linux.

[Qinsp]

Well, preliminary results indicate PS3 is a "best-fit" for me.

The WinXP simple GUI client crashed at 20 points. The Apple MBP is frozen at 100%. The PS3 is just chugging along.

My son's Vista computer "ate" a number of WU's it appears, as he plays high-CPU/GPU games on it.

My goal is "set-n-forget". I haven't tried a Linux box yet though, nor do I use 64-bit Linux due to certain compatibility requirements. I don't run 64-bit Windows either. It is not compatible/stable with a huge array of hardware and software I use for engineering purposes. I really WISH 64-bit systems were usable for my apps/hw, but they aren't yet.

[uncle fuzzy]

Set and forget is the unicore CPU client, either Systray version set to start with Windows, or the console version running as a service.

[PantherX]

In addition to what uncle fuzzy said, F@h Clients for Windows can work on both 32 and 64 bit so that isn't an issue for your. AFAIK, a Linux 32 bit only uses Classic Console so again that fits your requirement. However, I am interested in:

...The WinXP simple GUI client crashed at 20 points...

Can you please elaborate this statement?

...The Apple MBP is frozen at 100%...

It could be possible that it is uploading the wuresult and the time taken to upload depends on your internet connection.

...My son's Vista computer "ate" a number of WU's it appears, as he plays high-CPU/GPU games on it...

If you mean that you have folded and returned a number of WUs then it is fine. If those WUs didn't fold, then can you provide us additional information so we can help you (Details)

[Qinsp]

In addition to what uncle fuzzy said, F@h Clients for Windows can work on both 32 and 64 bit so that isn't an issue for your. AFAIK, a Linux 32 bit only uses Classic Console so again that fits your requirement. However, I am interested in:

...The WinXP simple GUI client crashed at 20 points...

Can you please elaborate this statement?

...The Apple MBP is frozen at 100%...

It could be possible that it is uploading the wuresult and the time taken to upload depends on your internet connection.

...My son's Vista computer "ate" a number of WU's it appears, as he plays high-CPU/GPU games on it...

If you mean that you have folded and returned a number of WUs then it is fine. If those WUs didn't fold, then can you provide us additional information so we can help you (Details)

Thanks!

I don't know how to check the logfile on the Apple MBP, but I did look at the WinXP and Vista machines.

Recurring theme:

* It can take over 4.25 hours to transmit results successfully. We have a stable internet connection with 12msec latency and 40+ mbps down, and 3+ mbps up. It's rejecting at the F@H end.

* It also says something like: FAH_corea4.exe not found or corrupt, then downloads it again.

* It will drop in the middle of job, I don't remember the syntax of the error. The computers aren't crashing.

But in the long run, the goal is to contribute without interfering or wasting power unnecessarily. I'll let them run as-is until the end of the week, and the ones that aren't working well I'll drop, then test other units. The PS3 is currently doing more work than the three x86-CPU machines combined.

[Baowoulf]

What's the server ip of the server that's rejecting your completed WU? The server status page will show you if a server is currently down or not accepting finished WU's. The link to the server status page is at the top of the page in the forums.

Also are you shutting down F@H with it being paused? Sometimes that can lead to problems. Always unpause F@H before closing it.

[Qinsp]

Thanks,

No, we are not turning off the app or the computers.

Team data is #195247
McRat is the Apple MBP (although first WU was a PS3)
McRatPS3a is a Slim PS3
McRatAMD2 is a 3.4ghz AMD XP
Tom is a Core 2 Duo 2ghz Vista 64.

Only the PS3 is working steady.

[PantherX]

Here is what I would suggest:
McRat -> No experience with Mac.
McRatAMD2 -> Single instance of Classic Client
Tom -> Single instance of Classic Client

Now I am pretty sure that the single instance will work without any issues and is a set-and-forget Client.
So can you please tell us what F@h Clients you have on your systems?

[Qinsp]

Both Windows machines are Systray GUI client.
Apple is OS X Intel.

Another racer (John_M) joined, all I know is he has an Alienware box, his is running sweet. He's in Illinois.

Looks like racer MMLMM is also experiencing dropped WU's, he's in AZ.

Crash error on WinXP (McRatAMD2) =

Code: Select all

[02:24:26] *********************** Log Started 18/Oct/2010 02:24:26 ***********************
[02:24:26] ************************** ProtoMol Folding@Home Core **************************
[02:24:26]   Version: 25
[02:24:26]      Type: 180
[02:24:26]      Core: ProtoMol
[02:24:26]   Website: http://folding.stanford.edu/
[02:24:26] Copyright: (c) 2009 Stanford University
[02:24:26]    Author: Joseph Coffland <[email protected]>
[02:24:26]      Args: -dir work/ -suffix 01 -checkpoint 15 -lifeline 3456 -version 623
[02:24:26] ************************************ Build *************************************
[02:24:26]      Date: May 18 2010
[02:24:26]      Time: 23:43:52
[02:24:26]  Revision: 1819
[02:24:26]  Compiler: Intel(R) C++ MSVC 1500 mode 1110
[02:24:26]   Options: /TP /nologo /EHsc /wd4297 /wd4103 /wd1786 /arch:IA32 /Ox
[02:24:26]            /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qrestrict /MT
[02:24:26]   Defines: _CRT_SECURE_NO_WARNINGS NDEBUG HAVE_GEEKINFO BOOST_ALL_NO_LIB
[02:24:26]            XML_STATIC HAVE_EXPAT HAVE_OPENSSL HAVE_LIBFAH HAVE_SIMTK_LAPACK
[02:24:26]  Platform: Windows XP
[02:24:26]      Bits: 32
[02:24:26]      Mode: Release
[02:24:26] ************************************ System ************************************
[02:24:26]        OS: Microsoft Windows XP Home Edition
[02:24:26]       CPU: AMD Phenom(tm) II X2 550 Processor
[02:24:26]    CPU ID: AuthenticAMD Family 16 Model 4 Stepping 3
[02:24:26]      CPUs: 2 Logical, 2 Physical
[02:24:26]    Memory: 3.00 GB
[02:24:26]   Threads: Windows
[02:24:26] ********************************************************************************
[02:24:26] Project: 10012 (Run 2536, Clone 0, Gen 46)
[02:24:26] Unit: 0x000000680001329c4bafd485000013cc
[02:24:26] User: 0x00000000000000000000000000000000
[02:24:26] Machine: 1
[02:24:26] Reading tar file par_all27_prot_lipid.inp
[02:24:26] Reading tar file scpismQuartic.inp
[02:24:26] Reading tar file ww.pdb
[02:24:26] Reading tar file ww.psf
[02:24:26] Reading tar file checkpt
[02:24:26] Reading tar file ww.5258.pos
[02:24:27] Reading tar file ww.5258.vel
[02:24:27] Reading tar file protomol.conf
[02:24:27] Reading tar file core.xml
[02:24:27] Digital signatures verified
[02:24:27] Completed 0 out of 499375 steps (0%)
[02:28:47] Opening C:\Documents and Settings\Owner\Application Data\Folding@home-x86\MyFolding.html...
[02:32:29] Completed 5000 out of 499375 steps (1%)
[02:40:27] Completed 10000 out of 499375 steps (2%)
[02:48:21] Completed 15000 out of 499375 steps (3%)
[02:56:40] Completed 20000 out of 499375 steps (4%)
[03:04:58] Completed 25000 out of 499375 steps (5%)
[03:13:04] Completed 30000 out of 499375 steps (6%)
[03:21:25] Completed 35000 out of 499375 steps (7%)
[03:29:33] Completed 40000 out of 499375 steps (8%)
[03:37:46] Completed 45000 out of 499375 steps (9%)
[03:45:55] Completed 50000 out of 499375 steps (10%)
[03:54:16] Completed 55000 out of 499375 steps (11%)
[04:02:36] Completed 60000 out of 499375 steps (12%)
[04:11:11] Completed 65000 out of 499375 steps (13%)
[04:19:22] Completed 70000 out of 499375 steps (14%)
[04:27:44] Completed 74900 out of 499375 steps (14%)
[04:36:14] Completed 79900 out of 499375 steps (16%)
[04:36:28][b] ERROR: ProtoMol ERROR: ABORT[/b]
[04:36:28] Saving result file logfile_01.txt
[04:36:28] Saving result file checkpt
[04:36:28] Saving result file checkpt.crc
[04:36:28] Saving result file log.txt
[04:36:29] Saving result file protomol.conf
[04:36:29] Saving result file ww.5266.pos
[04:36:29] Saving result file ww.5266.vel
[04:36:29] Saving result file ww.dcd
[04:36:29] Folding@home Core Shutdown: BAD_WORK_UNIT
[04:36:32] CoreStatus = 72 (114)
[04:36:32] Sending work to server
[04:36:32] Project: 10012 (Run 2536, Clone 0, Gen 46)
[04:36:32] + Attempting to send results [October 18 04:36:32 UTC]
[04:36:42] + Results successfully sent

[Qinsp]

Here is what I would suggest:
McRat -> No experience with Mac.
McRatAMD2 -> Single instance of Classic Client
Tom -> Single instance of Classic Client

Now I am pretty sure that the single instance will work without any issues and is a set-and-forget Client.
So can you please tell us what F@h Clients you have on your systems?

Thanks!

All are Systray for Windows from the Download page.

Is ProtoMol Error a problem with the F@H executable not being able to process a data packet it's given? If so, no biggie, and might explain the "eaten" packages.

Apple MBP is probably not going to be effective. Why it runs fine one day, then halts for extended periods is unknown, nor would time be well spent diagnosing it when I have other machines that I can try.

If I understand the project correctly, it's both massively parallel and sequential at the same time. Seems the goal is to process each job ASAP to empower the sequential aspect, while having huge amounts of processors engaged in the same mission. So if I understand it right, it's better for me to have "fast" machines than "wide" machines. Several machines that halt for extended periods are worse than having fewer processors on line?

Pat