Project 11920 (9752) has a short deadline

[CandaceMcRae]

I know I have been told before that some projects with short deadlines were tested on a different type of computer and had no issues with the timing. I last had issues with a project a while back that I kept getting assigned and unless I left my computer up 24/7 when I got that one I was never able to finish it and all that time was wasted in any attempt. I think I am getting that project again in another form. I hope I sent the correct info down below.

I just think that all the hours running this project to have it all rejected because it is not done by deadline is a horrible waste when it is only due to the fact that this gives you like 24 hours to complete a 15 hour folding project. I cant be the only one bothered by the waste of computing for no reason. Even if it was extended for another 24 hours it would allow anyone who has to shut down the computer to sleep to finish it with in reason. I know my computer might be running it slower than the test computer.

I just think someone needs to look at how many times this project was started with the average computer to not finished (wasted computing time) due to this deadline.

Project 11920

This is a copy project 9752 running on our new Gromacs 0xa7 for testing purposes.

Sodium channels initiate signaling in neurons and other cells which go on to become sensations of pleasure and pain, as well as thoughts and feelings. Sodium channel malfunction has been linked with cardiac arrhythmia, epilepsy, and neuropathic pain. These projects simulate a whole voltage-gated sodium channel in a bi-lipid membrane based on crystal structures deduced only recently (2013). The channel is made up of four homologous domain which form a pore through the membrane. ECT....

Assigned 2016-10-26T05:55:46Z
Timeout 2016-10-26T17:55:46Z
Expiration 2016-10-27T05:55:46Z

Code: Select all

*********************** Log Started 2016-10-27T01:22:01Z ***********************
01:22:01:************************* Folding@home Client *************************
01:22:01:      Website: http://folding.stanford.edu/
01:22:01:    Copyright: (c) 2009-2014 Stanford University
01:22:01:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
01:22:01:         Args: 
01:22:01:       Config: C:/Users/aarons/AppData/Roaming/FAHClient/config.xml
01:22:01:******************************** Build ********************************
01:22:01:      Version: 7.4.4
01:22:01:         Date: Mar 4 2014
01:22:01:         Time: 20:26:54
01:22:01:      SVN Rev: 4130
01:22:01:       Branch: fah/trunk/client
01:22:01:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
01:22:01:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
01:22:01:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
01:22:01:     Platform: win32 XP
01:22:01:         Bits: 32
01:22:01:         Mode: Release
01:22:01:******************************* System ********************************
01:22:01:          CPU: AMD A8-4500M APU with Radeon(tm) HD Graphics
01:22:01:       CPU ID: AuthenticAMD Family 21 Model 16 Stepping 1
01:22:01:         CPUs: 4
01:22:01:       Memory: 5.47GiB
01:22:01:  Free Memory: 4.09GiB
01:22:01:      Threads: WINDOWS_THREADS
01:22:01:   OS Version: 6.2
01:22:01:  Has Battery: true
01:22:01:   On Battery: false
01:22:01:   UTC Offset: -4
01:22:01:          PID: 7192
01:22:01:          CWD: C:/Users/aarons/AppData/Roaming/FAHClient
01:22:01:           OS: Windows 8
01:22:01:      OS Arch: AMD64
01:22:01:         GPUs: 1
01:22:01:        GPU 0: ATI:5 Trinity [Radeon HD 7640G]
01:22:01:         CUDA: Not detected
01:22:01:Win32 Service: false
01:22:01:***********************************************************************
01:22:01:<config>
01:22:01:  <!-- Network -->
01:22:01:  <proxy v=':8080'/>
01:22:01:
01:22:01:  <!-- Slot Control -->
01:22:01:  <pause-on-battery v='false'/>
01:22:01:  <power v='FULL'/>
01:22:01:
01:22:01:  <!-- User Information -->
01:22:01:  <passkey v='********************************'/>
01:22:01:  <user v='CandaceMcRae'/>
01:22:01:
01:22:01:  <!-- Folding Slots -->
01:22:01:  <slot id='0' type='CPU'>
01:22:01:    <cpus v='3'/>
01:22:01:  </slot>
01:22:01:</config>
01:22:01:Trying to access database...
01:22:01:Successfully acquired database lock
01:22:01:Enabled folding slot 00: READY cpu:3
01:22:01:WU00:FS00:Starting
01:22:01:WU00:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/aarons/AppData/Roaming/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/AVX/Core_a7.fah/FahCore_a7.exe -dir 00 -suffix 01 -version 704 -lifeline 7192 -checkpoint 15 -np 3
01:22:01:WU00:FS00:Started FahCore on PID 11580
01:22:01:WU00:FS00:Core PID:3864
01:22:01:WU00:FS00:FahCore 0xa7 started
01:22:05:WU00:FS00:0xa7:*********************** Log Started 2016-10-27T01:22:05Z ***********************
01:22:05:WU00:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
01:22:05:WU00:FS00:0xa7:       Type: 0xa7
01:22:05:WU00:FS00:0xa7:       Core: Gromacs
01:22:05:WU00:FS00:0xa7:    Website: http://folding.stanford.edu/
01:22:05:WU00:FS00:0xa7:  Copyright: (c) 2009-2016 Stanford University
01:22:05:WU00:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
01:22:05:WU00:FS00:0xa7:       Args: -dir 00 -suffix 01 -version 704 -lifeline 11580 -checkpoint 15 -np
01:22:05:WU00:FS00:0xa7:             3
01:22:05:WU00:FS00:0xa7:     Config: <none>
01:22:05:WU00:FS00:0xa7:************************************ Build *************************************
01:22:05:WU00:FS00:0xa7:    Version: 0.0.11
01:22:05:WU00:FS00:0xa7:       Date: Sep 21 2016
01:22:05:WU00:FS00:0xa7:       Time: 01:43:48
01:22:05:WU00:FS00:0xa7: Repository: Git
01:22:05:WU00:FS00:0xa7:   Revision: 957bd90e68d95ddcf1594dc15ff6c64cc4555146
01:22:05:WU00:FS00:0xa7:     Branch: master
01:22:05:WU00:FS00:0xa7:   Compiler: GNU 4.2.1 Compatible Clang 3.9.0 (trunk 274080)
01:22:05:WU00:FS00:0xa7:    Options: -std=gnu++98 -O3 -funroll-loops -ffast-math -mfpmath=sse
01:22:05:WU00:FS00:0xa7:             -fno-unsafe-math-optimizations -msse2 -I/mingw64/include
01:22:05:WU00:FS00:0xa7:             -Wno-inconsistent-dllimport -Wno-parentheses-equality
01:22:05:WU00:FS00:0xa7:             -Wno-deprecated-register -Wno-unused-local-typedef
01:22:05:WU00:FS00:0xa7:   Platform: linux2 4.6.0-1-amd64
01:22:05:WU00:FS00:0xa7:       Bits: 64
01:22:05:WU00:FS00:0xa7:       Mode: Release
01:22:06:WU00:FS00:0xa7:       SIMD: avx_256
01:22:06:WU00:FS00:0xa7:************************************ System ************************************
01:22:06:WU00:FS00:0xa7:        CPU: AMD A8-4500M APU with Radeon(tm) HD Graphics
01:22:06:WU00:FS00:0xa7:     CPU ID: AuthenticAMD Family 21 Model 16 Stepping 1
01:22:06:WU00:FS00:0xa7:       CPUs: 4
01:22:06:WU00:FS00:0xa7:     Memory: 5.47GiB
01:22:06:WU00:FS00:0xa7:Free Memory: 4.06GiB
01:22:06:WU00:FS00:0xa7:    Threads: WINDOWS_THREADS
01:22:06:WU00:FS00:0xa7: OS Version: 6.2
01:22:06:WU00:FS00:0xa7:Has Battery: true
01:22:06:WU00:FS00:0xa7: On Battery: false
01:22:06:WU00:FS00:0xa7: UTC Offset: -4
01:22:06:WU00:FS00:0xa7:        PID: 3864
01:22:06:WU00:FS00:0xa7:        CWD: C:\Users\aarons\AppData\Roaming\FAHClient\work
01:22:06:WU00:FS00:0xa7:         OS: Windows 8
01:22:06:WU00:FS00:0xa7:    OS Arch: AMD64
01:22:06:WU00:FS00:0xa7:********************************************************************************
01:22:06:WU00:FS00:0xa7:Project: 11920 (Run 2482, Clone 0, Gen 47)
01:22:06:WU00:FS00:0xa7:Unit: 0x00000047ab40412957bff2a8e2a09ef5
01:22:06:WU00:FS00:0xa7:Digital signatures verified
01:22:06:WU00:FS00:0xa7:Calling: mdrun -s frame47.tpr -o frame47.trr -cpi state.cpt -cpt 15 -nt 3
01:22:22:WU00:FS00:0xa7:Steps: first=3760000 total=80000
01:22:42:WU00:FS00:0xa7:Completed 32701 out of 80000 steps (40%)
01:23:31:WU00:FS00:0xa7:Completed 32800 out of 80000 steps (41%)
01:24:36:35:127.0.0.1:New Web connection
01:31:15:WU00:FS00:0xa7:Completed 33600 out of 80000 steps (42%)

Mod edit: added Code tags to log listing

[CandaceMcRae]

Oh and I have gotten and completed the project recently but only because I didnt shut down the computer that night when I saw the deadline. However now I am started to get assigned this multiple times again. I know I can lower my CPU down to avoid it but why should I have to when it is not my fault it has a short deadline.

[toTOW]

Did you fold p9752 before ? The new project should fold as fast or faster than the old one because of the optimization included in the new core ...

[CandaceMcRae]

Yes I have folded P9752 before and that was the only project prior to this that I have not been able to complete for the same exact reason. That project also would have the folding deadline expire prior to being able to finish folding it. I kept getting that one assigned also multiple times. I was just looking at the fact that I just logged on for the day and looked at my current assignment and it is this project and it has 7 hours to do 13 hours of work. I was assigned at 2016-10-31T23:40:56Z shut down the computer at 3/4am I think to sleep and just logged back on here at 12noon. This project will timeout at 2016-11-01T11:40:56Z and expire at 2016-11-01T23:40:56Z. I will most likely not shut down till about midnight. So unless I happen to get this project first thing when I log on for the day then I can never finish folding it before it expires because of the extremely short deadline. I have only completed this project once because I didnt shut down the computer while I slept (not doing on a regular basis).

[CandaceMcRae]

Ok correction. I forgot to give the system a little bit of time before looking. I have 7 hours to do 4 hours of work. Which is doable but the project deadline is so tight that if it was even extended to 48 hours from the current 24 hours it would prevent most uncompleted jobs.

[Joe_H]

I have folded WU's from Project 9752 on a MacBook with a 2.4 GHz Core 2 Duo as part of the original beta test. They completed just barely within the preferred timeout. Then the project manager restricted the assignment to 3 cores or more.

11920 is a direct copy of that project for testing of the A7 core, my understanding is that no re-benchmarking is planned for this. So I don't think the deadlines will be modified.

One thing that may help CPU processing, the client setup on your machine detected the integrated GPU in your APU and reserved one core for that. So though you do not have a GPU slot configured, Full is only using 3 of the 4 cores available. You might see an improvement in folding time by changing the "-1" in the CPU folding slot configuration to 4.

[CandaceMcRae]

Edited - the Code tag in the log listing - to the correct project number in ( )

I will see if that improves anything. I just worry about others who also get these projects and are just average folders. Sorry I just hate to see the precious time folding lost because they cant complete it.

2nd Edit... ok I would make that change if I knew how to. I know I am suppose to go to the advanced control then configure. I am assuming it is slots. I am not computer savvy but not totally inept. ID 0 Type cpu ... is what it is currently. How and what does that get changed to because you were talking about a -1 and I know it is set for 3 because I am running this project and I know I have 4 cpu's total. So bear with me on not knowing what to input/add/remove/edit in the slots tab. Or am I totally in the wrong place completely?


I know I stopped the GPU from folding as I assume that was what was causing my issues doing other tasks as the folding was running in the background back a year ago or more when I finally got the GPU to even start folding properly instead of it failing. I ask a lot of the computer sometimes with a ton of tabs open, playing videos and games at the same time so I stress the GPU out. I have not had any errors in the last couple of months but I was causing the Graphics to crash even after I eliminated the folding on it. I might have gotten a update that fixed my issue or it just stopped being a weakling and got over its issue. So I am staying away from folding on the GPU altogether

[Joe_H]

First, how to modify the CPU slot settings. Using FAHControl open the Configure window and select Slots. Click on the CPU slot to highlight is and then click on Edit at the bottom of the window. The Configure Folding Slot window will open, settings for a CPU slot are at the top. The default setting for number of CPU threads is a "-1" which leaves control to the client and power slider position. You change that to 4 and the client will use all CPU threads available on your CPU.

Quick guide to power slider position translated onto core usage - Full uses all cores available after whatever is reserved for the GPU. Medium uses one less core, and Light uses half the available cores. Your APU has 4 cores, so with one reserved for the GPU Full currently uses just 3.

As for GPU folding on the integrated GPU that is part of an APU, reports have been mixed. Some GPU sections do not even have the processing power to complete WU's within deadlines; and for those that do, success has been hit or miss depending on which driver updates is being used. Even then since the heat generated by the GPU section falls within the TDP set for the APU, it can reduce the CPU's speed to stay within cooling limits.

[bruce]

Your HD 7640G has 256 shaders running at 685 GHz. That produces a minimal amount of folding power compared to most (all?) discrete GPUs. Personally, I have a HD 7560D which also has 256 shaders but at 760 GHz (i.e. very similar). It, too, is too slow to fold any of the FAH projects being currently assigned.

I understand the need to support fast GPUs and to complete projects promptly but supporting the extremely wide range of GPU speeds that are available on the market is next to impossible. Either FAH needs to establish a unique class of projects with longer deadlines that can use the more limited resoruces of such iGPUs or it needs to drop support for them entirely.

What I'm not sure of is whether a CPU slot is more productive when the GPU is folding (due to shared resources) or if it's still equally productive. My goal is to maximize the resources devoted to FAH and it's probably best achieved by devoting them all to the CPU slot.

[Aesprit]

Given that this is stated to be for testing, is it possible that they intentionally set a short deadline to assess how many computers would be able to meet it? After all, 9752, the project that actually got the data, has already been completed. They don't really need people to complete the Work Unit in order to be testing (unless I'm mistaken about the manner or purpose of this particular testing.)

[Joe_H]

There are two aspects to this test. One is that the folding core is new and based on a later Gromacs version. So it is desirable to complete the work to validate that the new A7 core gives the same results as the previous processing using the older A4 core.

Second is to see what performance increase comes from using this new core. That in itself has two components as Core_A7 comes in two varieties, one using SSE2 processing on older CPU's and the other using AVX on newer CPU's which include that instruction set. Having the same deadlines and base points as the original project used after being benchmarked is part of that.

[Aesprit]

So would it be correct to say that it's helpful for the testing for the Work Unit to be completed, but that even when the WU fails to meet the deadline, that's a data point that's useful for testing, at least for the performance increase aspect?

[bruce]

Fahcore_a7 is supported by either SSE or AVX. If your hardware supports AVX it will certainly run faster than if it only supports SSE2.

FahCore_a7 also replaces an older version of GROMACS found in FAHCOre_a4 with a newer version> I'd expect the performance of the newer version to be equal to or faster than FahCore_a4.

Thus completion times for the same WU should be
FahCore_a4(sse)>=FahCore_a7(sse)> FAHCore_a7(avx)

Project 9752 has deadlines of 0.5 and 1.0 days. Project 11920 has the same deadlines. Thus, if your hardware could meet the deadlines on p9752, it should be able to meet or exceed them on p11920.

The only other think I can think of is maybe you never received a WU from p9752.

The following is pure speculation, not based on any actual knowledge of these projects:
Let's suppose you're running on a system with N dedicated CPUs. Let's suppose that the server's assignment rules for p9752 had a minimum requirement of N+1 CPUs and maybe that restriction wasn't migrated to p11920. That would explain why nobody complained about the deadlines on the old project -- in that all WUs were restricted to more powerful systems.