New work units have tiny PPD yields?

Moderators: Site Moderators, FAHC Science Team

New work units have tiny PPD yields?

Postby peter.dolkens » Tue Oct 09, 2012 8:57 am

My machine normally churns out between 60-70k PPD according to the FAHControl estimates (Which aren't far off my actual yields)

Recently, however, I've noticed that my smp core has dropped from 20k PPD down to just 4k PPD.

For a while I was getting assigned units that would take my machine all day and yield 20k points at once, now I'm getting things that take half the time, and yield a quarter of the points :(

What gives?

Current project: 10090 (156, 0, 51)
FahCore 0xa4

CPU: 3930k @ 4GHz
RAM: 64GB 1600MHz Corsair
GPU: Dual Asus nVidia 680
peter.dolkens
 
Posts: 4
Joined: Tue Oct 09, 2012 8:52 am

Re: New work units have tiny PPD yields?

Postby P5-133XL » Tue Oct 09, 2012 9:15 am

Please supply us with a log.
Image
P5-133XL
 
Posts: 2948
Joined: Sun Dec 02, 2007 5:36 am
Location: Salem. OR USA

Re: New work units have tiny PPD yields?

Postby bruce » Tue Oct 09, 2012 8:23 pm

Welcome to foldingforum.org, peter.dolkens.

In addition to the log, I have two questions:
> Did you recently add another application that runs continuously? FAH is designed to run at an extremely low priority and yields resources to anything else that might be running. Checking Task Monitor might provide useful information.
> How are you determining PPD? The numbers computed by the new client still have some inaccuracies. The actual numbers reported by other methods are all subject to a rather complex system where the bonus points might be impacted if your passkey has accumulated too many WUs that missed the Timeout requirements (Preferred Deadline).
bruce
 
Posts: 20140
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

Re: New work units have tiny PPD yields?

Postby 7im » Tue Oct 09, 2012 8:28 pm

You mentioned the new WUs and their PPD. What were the old WU Project Numbers at that higher PPD?
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 10189
Joined: Thu Nov 29, 2007 5:30 pm
Location: Arizona

Re: New work units have tiny PPD yields?

Postby peter.dolkens » Wed Oct 10, 2012 3:20 am

http://blog.ddrit.com/fold.txt for the log.

Main thing is not necessarily the yield per WU, but the yield per day.

As for new apps that run continuously, no nothing that should have that kind of impact, My machine has more than enough resources to handle what I do on a day to day basis without affecting folding too much (When I'm not folding, I'm transcoding video, which runs at a higher priority than folding, at 100%, and doesn't affect any of my Builds / Coding / work related tasks)

I'm determining PPD based on the estimates in FAHControl, which so far have seemed fairly accurate.

Some of the higher yielding projects:

Project: 8066 (Run 181, Clone 1, Gen 28)
Project: 8067 (Run 1104, Clone 0, Gen 28)
Project: 7005 (Run 0, Clone 329, Gen 31)

Even the smaller projects that were yielding 700 points per work unit, were completing much faster than this particular unit.

Today that unit is completed, and my estimates currently reading 21213.38 PPD, so it's back to the better WUs, I was just wondering if that particular unit was broke?

http://folding.extremeoverclocking.com/team_summary.php?s=&t=221930
peter.dolkens
 
Posts: 4
Joined: Tue Oct 09, 2012 8:52 am

Re: New work units have tiny PPD yields?

Postby heikosch » Wed Oct 10, 2012 10:35 pm

@peter.dolkens
Other processes with high CPU load often result in low PPD for SMP clients.

PS: Check whether your GPU clients run correctly. Some people (me too) reported that their GTX680s tried to run non beta projects => not working, endless loops with high CPU load which disturbs the SMP client.

Heiko
Image Image
Image
heikosch
 
Posts: 110
Joined: Thu Apr 30, 2009 8:31 pm
Location: Essen, Germany

Re: New work units have tiny PPD yields?

Postby peter.dolkens » Thu Oct 11, 2012 1:33 am

@heikosch: GPUs are running fine, FAHControl is reporting them doing lots of 5k point work units.

Also, don't get me wrong, my usage of the machine hasn't changed, and I turn folding off when I am transcoding, I was just using it as a comparison.

In order for me to actually make a dint on folding, I'd have to be trying REALLY hard, especially as I have 12 threads, and my development environment only uses 1 of those by default, leaving 11 more for folding.
peter.dolkens
 
Posts: 4
Joined: Tue Oct 09, 2012 8:52 am

Re: New work units have tiny PPD yields?

Postby P5-133XL » Thu Oct 11, 2012 3:30 am

If your dev work uses just a single core try SMP:10 (SMP:11 is likely to cause problems with a few projects). SMP cores are highly synchronized and even if one core is significantly being used (10%) by some other process that can cause the other folding threads to sit in a waiting mode for the suspended thread to return causing very significant PPD drops. It is likely that since we are talking hyper-threaded cores that dropping the cores from 12 to 10 will not affect the SMP productivity very much at all even when it is not being used by some other process and help very significantly when there is CPU usage. So I would recommend trying it and see if that helps.
P5-133XL
 
Posts: 2948
Joined: Sun Dec 02, 2007 5:36 am
Location: Salem. OR USA

Re: New work units have tiny PPD yields?

Postby patonb » Thu Oct 11, 2012 3:28 pm

Actually, I'm seeing a huge cut in WU# 7085 (0 143 3)

Machine usually makes 14K ppd and that unit is squeeking out 9284ppd. Thats the lowest I've ever seen on it.

Its my i7 950 running at 3.06Ghz doing nothing but folding.
WooHoo = L5639 @ 3.3Ghz Evga SR-2 6x2gb Corsair XMS3 CM 212+ Corsair 1050hx Blackhawk Ultra EVGA 560ti

Foldie = i7 950@ 4.0Ghz x58a-ud3r 216-216 @ 850/2000 3x2gb OCZ Gold NH-u12 Heatsink Corsair hx520 Antec 900
patonb
 
Posts: 348
Joined: Thu Oct 23, 2008 3:42 am

Re: New work units have tiny PPD yields?

Postby 7im » Thu Oct 11, 2012 5:58 pm

There are a4 work units which are pure SMP, and follow the standard SMP benchmark, same deadline, and hence the same bonus. These are for SMP 2 and up.

There are also what has been called "hybrid" a4 work units which are for SMP 1 (UNIprocessor) and up. These are benchmarked exactly the same as the above. So the base points are exactly the same as above. However, these are designed for both UNI processor slots and SMP slots on lower end systems. These projects have been given longer deadlines. And the longer deadline changes the points curve on the bonus calculation. The curve gets stretched out...

So System X may score exactly the same daily PPD on both types of a4 work units.
System Y will score higher on the pure SMP a4s, and score less on the hybrid a4s.

No, you can't control the WU selection so that you get only SMP a4s. As always, fold what you get.
Yes, getting hybrid a4 work units gets you more PPD than letting your client sit idle waiting for SMP 4s.

THESE WERE NOT BENCHMARKED INCORRECTLY, SO THERE IS NOTHING WRONG WITH THE SERVERS, THE CLIENT, THE BONUS, THE BENCHMARK, AND THERE IS NOTHING WRONG WITH YOUR SYSTEM.

@peter.dolkens - The work server sending out project 10090 has a SMP Minimum setting of 1. So this is mostly like a hybrid a4 work units, which explains the lower average PPD.

@patonb - Project 7085 is also coming from a server with SMP Minimum of 1 core, so this is also a hybrid a4 WU.
User avatar
7im
 
Posts: 10189
Joined: Thu Nov 29, 2007 5:30 pm
Location: Arizona

Re: New work units have tiny PPD yields?

Postby bruce » Thu Oct 11, 2012 7:48 pm

patonb wrote:Actually, I'm seeing a huge cut in WU# 7085 (0 143 3)

Machine usually makes 14K ppd and that unit is squeeking out 9284ppd. Thats the lowest I've ever seen on it.

Its my i7 950 running at 3.06Ghz doing nothing but folding.


A report of a low PPD cannot be considered a bug report without an associated log. You have not shown the actual frame times. You have not documented whether it's a problem with unexpectedly long TPF or if it's a problem with the PPD that you are reporting was calculated.
bruce
 
Posts: 20140
Joined: Thu Nov 29, 2007 11:13 pm
Location: So. Cal.

Re: New work units have tiny PPD yields?

Postby patonb » Fri Oct 12, 2012 2:36 am

I never claimed it to be a bug, I'm adding to the op's information that the newer units are dropping in points. However 7im pointed out that it is possible it's due to the fact its a unicore/smp unit.

My tpf is 11min 16sec, and I totally hate v7 logs

Code: Select all
*********************** Log Started 2012-10-11T21:07:04Z ***********************
21:07:04:************************* Folding@home Client *************************
21:07:04:      Website: http://folding.stanford.edu/
21:07:04:    Copyright: (c) 2009-2012 Stanford University
21:07:04:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
21:07:04:         Args: --lifeline 1764 --command-port=36330
21:07:04:       Config: C:/ProgramData/FAHClient/config.xml
21:07:04:******************************** Build ********************************
21:07:04:      Version: 7.1.52
21:07:04:         Date: Mar 20 2012
21:07:04:         Time: 19:37:42
21:07:04:      SVN Rev: 3515
21:07:04:       Branch: fah/trunk/client
21:07:04:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
21:07:04:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
21:07:04:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT
21:07:04:     Platform: win32 XP
21:07:04:         Bits: 32
21:07:04:         Mode: Release
21:07:04:******************************* System ********************************
21:07:04:          CPU: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz
21:07:04:       CPU ID: GenuineIntel Family 6 Model 26 Stepping 5
21:07:04:         CPUs: 8
21:07:04:       Memory: 6.00GiB
21:07:04:  Free Memory: 5.33GiB
21:07:04:      Threads: WINDOWS_THREADS
21:07:04:   On Battery: false
21:07:04:   UTC offset: -4
21:07:04:          PID: 2528
21:07:04:          CWD: C:/ProgramData/FAHClient
21:07:04:           OS: Windows Server 2008 R2 Standard
21:07:04:      OS Arch: AMD64
21:07:04:         GPUs: 1
21:07:04:        GPU 0: UNSUPPORTED: Rage XL (Intel Corporation)
21:07:04:         CUDA: Not detected
21:07:04:Win32 Service: false
21:07:04:***********************************************************************
21:07:04:<config>
21:07:04:  <!-- Network -->
21:07:04:  <proxy v=':8080'/>
21:07:04:
21:07:04:  <!-- Remote Command Server -->
21:07:04:  <command-allow v='127.0.0.1 192.168.1.0-192.168.1.255'/>
21:07:04:  <command-allow-no-pass v='127.0.0.1 192.168.1.0-192.168.1.255'/>
21:07:04:
21:07:04:  <!-- User Information -->
21:07:04:  <passkey v='********************************'/>
21:07:04:  <team v='12864'/>
21:07:04:  <user v='patonb'/>
21:07:04:
21:07:04:  <!-- Folding Slots -->
21:07:04:  <slot id='0' type='SMP'/>
21:07:04:</config>
21:07:04:Trying to access database...
21:07:08:Successfully acquired database lock
21:07:08:Enabled folding slot 00: READY smp:8
21:07:09:WU00:FS00:Starting
21:07:09:WU00:FS00:Running FahCore: \"C:\\Program Files (x86)\\FAHClient/FAHCoreWrapper.exe\" C:/ProgramData/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/Core_a4.fah/FahCore_a4.exe -dir 00 -suffix 01 -version 701 -lifeline 2528 -checkpoint 15 -np 8
21:07:09:WU00:FS00:Started FahCore on PID 2624
21:07:11:Server connection id=1 on 0.0.0.0:36330 from 127.0.0.1
21:07:12:WU00:FS00:Core PID:2640
21:07:12:WU00:FS00:FahCore 0xa4 started
21:07:13:WU00:FS00:0xa4:
21:07:13:WU00:FS00:0xa4:*------------------------------*
21:07:13:WU00:FS00:0xa4:Folding@Home Gromacs GB Core
21:07:13:WU00:FS00:0xa4:Version 2.27 (Dec. 15, 2010)
21:07:13:WU00:FS00:0xa4:
21:07:13:WU00:FS00:0xa4:Preparing to commence simulation
21:07:13:WU00:FS00:0xa4:- Ensuring status. Please wait.
21:07:22:WU00:FS00:0xa4:- Looking at optimizations...
21:07:22:WU00:FS00:0xa4:- Working with standard loops on this execution.
21:07:22:WU00:FS00:0xa4:- Previous termination of core was improper.
21:07:22:WU00:FS00:0xa4:- Going to use standard loops.
21:07:22:WU00:FS00:0xa4:- Files status OK
21:07:23:WU00:FS00:0xa4:- Expanded 109030 -> 443200 (decompressed 406.4 percent)
21:07:23:WU00:FS00:0xa4:Called DecompressByteArray: compressed_data_size=109030 data_size=443200, decompressed_data_size=443200 diff=0
21:07:23:WU00:FS00:0xa4:- Digital signature verified
21:07:23:WU00:FS00:0xa4:
21:07:23:WU00:FS00:0xa4:Project: 7085 (Run 0, Clone 143, Gen 3)
21:07:23:WU00:FS00:0xa4:
21:07:23:WU00:FS00:0xa4:Entering M.D.
21:07:29:WU00:FS00:0xa4:Using Gromacs checkpoints
21:07:29:WU00:FS00:0xa4:Mapping NT from 8 to 8
21:07:32:WU00:FS00:0xa4:Resuming from checkpoint
21:07:32:WU00:FS00:0xa4:Verified 00/wudata_01.log
21:07:32:WU00:FS00:0xa4:Verified 00/wudata_01.trr
21:07:32:WU00:FS00:0xa4:Verified 00/wudata_01.xtc
21:07:32:WU00:FS00:0xa4:Verified 00/wudata_01.edr
21:07:33:WU00:FS00:0xa4:Completed 4000001 out of 10000000 steps  (40%)
21:18:48:WU00:FS00:0xa4:Completed 4100000 out of 10000000 steps  (41%)
21:30:01:WU00:FS00:0xa4:Completed 4200000 out of 10000000 steps  (42%)
21:41:15:WU00:FS00:0xa4:Completed 4300000 out of 10000000 steps  (43%)
21:52:29:WU00:FS00:0xa4:Completed 4400000 out of 10000000 steps  (44%)
22:03:43:WU00:FS00:0xa4:Completed 4500000 out of 10000000 steps  (45%)
22:14:57:WU00:FS00:0xa4:Completed 4600000 out of 10000000 steps  (46%)
22:25:36:Server connection id=2 on 0.0.0.0:36330 from 192.168.1.101
22:26:11:WU00:FS00:0xa4:Completed 4700000 out of 10000000 steps  (47%)
22:37:25:WU00:FS00:0xa4:Completed 4800000 out of 10000000 steps  (48%)
22:48:38:WU00:FS00:0xa4:Completed 4900000 out of 10000000 steps  (49%)
22:59:52:WU00:FS00:0xa4:Completed 5000000 out of 10000000 steps  (50%)
23:11:06:WU00:FS00:0xa4:Completed 5100000 out of 10000000 steps  (51%)
23:22:20:WU00:FS00:0xa4:Completed 5200000 out of 10000000 steps  (52%)
23:33:35:WU00:FS00:0xa4:Completed 5300000 out of 10000000 steps  (53%)
23:44:49:WU00:FS00:0xa4:Completed 5400000 out of 10000000 steps  (54%)
23:56:04:WU00:FS00:0xa4:Completed 5500000 out of 10000000 steps  (55%)
00:07:18:WU00:FS00:0xa4:Completed 5600000 out of 10000000 steps  (56%)
00:18:33:WU00:FS00:0xa4:Completed 5700000 out of 10000000 steps  (57%)
00:29:47:WU00:FS00:0xa4:Completed 5800000 out of 10000000 steps  (58%)
00:41:01:WU00:FS00:0xa4:Completed 5900000 out of 10000000 steps  (59%)
00:52:15:WU00:FS00:0xa4:Completed 6000000 out of 10000000 steps  (60%)
01:03:31:WU00:FS00:0xa4:Completed 6100000 out of 10000000 steps  (61%)
01:14:47:WU00:FS00:0xa4:Completed 6200000 out of 10000000 steps  (62%)
01:26:03:WU00:FS00:0xa4:Completed 6300000 out of 10000000 steps  (63%)
patonb
 
Posts: 348
Joined: Thu Oct 23, 2008 3:42 am

Re: New work units have tiny PPD yields?

Postby peter.dolkens » Tue Oct 16, 2012 12:39 am

7im wrote:There are a4 work units which are pure SMP, and follow the standard SMP benchmark, same deadline, and hence the same bonus. These are for SMP 2 and up.

There are also what has been called "hybrid" a4 work units which are for SMP 1 (UNIprocessor) and up. These are benchmarked exactly the same as the above. So the base points are exactly the same as above. However, these are designed for both UNI processor slots and SMP slots on lower end systems. These projects have been given longer deadlines. And the longer deadline changes the points curve on the bonus calculation. The curve gets stretched out...


Thanks 7im, I guess I'll just assume that this accounts for the difference I was seeing (Though in reverse)

The PPD of the "odd" work unit was closer to benchmarks, but the PPD of the larger work units was much higher, I'm assuming because my machine was able to complete them so much faster (putting me higher on that "curve")

It doesn't seem I've been getting too many more like that, and I'm fairly new to this whole folding thing, so thankyou for your help.
peter.dolkens
 
Posts: 4
Joined: Tue Oct 09, 2012 8:52 am


Return to V7.1.52 Windows/Linux

Who is online

Users browsing this forum: No registered users and 1 guest

cron