A lot of problems 5870 crossfire. Help would be appreciated

[cud0s]

Hello i will keep this short because i really neet to go to sleep I have got amd 5870 crossfire and amd 8320fx however i get low gpu utilisation using 16 core and very lov ppd -5000 for each gpu. 17 core improves gpu usage, but now it shows ppd unknown, eta unknown and progress stuck at 0%. I also have very low ppd when using cpu. This is verbose 5 log, one gpu using 16 one 17 core:

Code: Select all

*********************** Log Started 2013-08-13T22:17:32Z ***********************
22:17:32:************************* Folding@home Client *************************
22:17:32:      Website: http://folding.stanford.edu/
22:17:32:    Copyright: (c) 2009-2013 Stanford University
22:17:32:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
22:17:32:         Args: 
22:17:32:       Config: C:/ProgramData/FAHClient/config.xml
22:17:32:******************************** Build ********************************
22:17:32:      Version: 7.3.6
22:17:32:         Date: Feb 18 2013
22:17:32:         Time: 15:25:17
22:17:32:      SVN Rev: 3923
22:17:32:       Branch: fah/trunk/client
22:17:32:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
22:17:32:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
22:17:32:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
22:17:32:     Platform: win32 XP
22:17:32:         Bits: 32
22:17:32:         Mode: Release
22:17:32:******************************* System ********************************
22:17:32:          CPU: AMD FX(tm)-8320 Eight-Core Processor
22:17:32:       CPU ID: AuthenticAMD Family 21 Model 2 Stepping 0
22:17:32:         CPUs: 8
22:17:32:       Memory: 7.96GiB
22:17:32:  Free Memory: 4.83GiB
22:17:32:      Threads: WINDOWS_THREADS
22:17:32:  Has Battery: false
22:17:32:   On Battery: false
22:17:32:   UTC offset: 3
22:17:32:          PID: 4800
22:17:32:          CWD: C:/ProgramData/FAHClient
22:17:32:           OS: Windows 8 Pro
22:17:32:      OS Arch: AMD64
22:17:32:         GPUs: 2
22:17:32:        GPU 0: ATI:4 Radeon HD 5870 (Cypress)
22:17:32:        GPU 1: ATI:4 Radeon HD 5870 (Cypress)
22:17:32:         CUDA: Not detected
22:17:32:Win32 Service: false
22:17:32:***********************************************************************
22:17:32:<config>
22:17:32:  <service-description v='Folding@home Client'/>
22:17:32:  <service-restart v='true'/>
22:17:32:  <service-restart-delay v='5000'/>
22:17:32:
22:17:32:  <!-- Client Control -->
22:17:32:  <client-threads v='4'/>
22:17:32:  <cycle-rate v='4'/>
22:17:32:  <cycles v='-1'/>
22:17:32:  <data-directory v='.'/>
22:17:32:  <disable-sleep-when-active v='true'/>
22:17:32:  <exec-directory v='C:\Program Files (x86)\FAHClient'/>
22:17:32:  <exit-when-done v='false'/>
22:17:32:  <fold-anon v='false'/>
22:17:32:  <open-web-control v='false'/>
22:17:32:
22:17:32:  <!-- Configuration -->
22:17:32:  <config-rotate v='true'/>
22:17:32:  <config-rotate-dir v='configs'/>
22:17:32:  <config-rotate-max v='16'/>
22:17:32:
22:17:32:  <!-- Debugging -->
22:17:32:  <assignment-servers>
22:17:32:    assign3.stanford.edu:8080 assign4.stanford.edu:80
22:17:32:  </assignment-servers>
22:17:32:  <capture-directory v='capture'/>
22:17:32:  <capture-on-error v='false'/>
22:17:32:  <capture-packets v='false'/>
22:17:32:  <capture-requests v='false'/>
22:17:32:  <capture-responses v='false'/>
22:17:32:  <capture-sockets v='false'/>
22:17:32:  <debug-sockets v='false'/>
22:17:32:  <exception-locations v='true'/>
22:17:32:  <gpu-assignment-servers>
22:17:32:    assign-GPU.stanford.edu:80 assign-GPU.stanford.edu:8080
22:17:32:  </gpu-assignment-servers>
22:17:32:  <stack-traces v='false'/>
22:17:32:
22:17:32:  <!-- Error Handling -->
22:17:32:  <max-slot-errors v='5'/>
22:17:32:  <max-unit-errors v='5'/>
22:17:32:
22:17:32:  <!-- Folding Core -->
22:17:32:  <checkpoint v='15'/>
22:17:32:  <core-dir v='cores'/>
22:17:32:  <core-priority v='idle'/>
22:17:32:  <cpu-affinity v='false'/>
22:17:32:  <cpu-usage v='100'/>
22:17:32:  <gpu-usage v='100'/>
22:17:32:  <no-assembly v='false'/>
22:17:32:
22:17:32:  <!-- Folding Slot Configuration -->
22:17:32:  <cause v='ANY'/>
22:17:32:  <client-subtype v='STDCLI'/>
22:17:32:  <client-type v='normal'/>
22:17:32:  <cpu-species v='X86_AMD'/>
22:17:32:  <cpu-type v='AMD64'/>
22:17:32:  <cpus v='-1'/>
22:17:32:  <cuda-index v='0'/>
22:17:32:  <gpu v='true'/>
22:17:32:  <max-packet-size v='normal'/>
22:17:32:  <opencl-index v='0'/>
22:17:32:  <os-species v='UNKNOWN'/>
22:17:32:  <os-type v='WIN32'/>
22:17:32:  <power v='full'/>
22:17:32:  <project-key v='0'/>
22:17:32:  <smp v='true'/>
22:17:32:
22:17:32:  <!-- HTTP Server -->
22:17:32:  <allow v='127.0.0.1'/>
22:17:32:  <connection-timeout v='60'/>
22:17:32:  <deny v='0/0'/>
22:17:32:  <http-addresses v='0:7396'/>
22:17:32:  <https-addresses v=''/>
22:17:32:  <max-connect-time v='900'/>
22:17:32:  <max-connections v='800'/>
22:17:32:  <max-request-length v='52428800'/>
22:17:32:  <min-connect-time v='300'/>
22:17:32:  <threads v='8'/>
22:17:32:
22:17:32:  <!-- Logging -->
22:17:32:  <log v='log.txt'/>
22:17:32:  <log-color v='false'/>
22:17:32:  <log-crlf v='true'/>
22:17:32:  <log-date v='false'/>
22:17:32:  <log-date-periodically v='21600'/>
22:17:32:  <log-debug v='true'/>
22:17:32:  <log-domain v='false'/>
22:17:32:  <log-header v='true'/>
22:17:32:  <log-level v='true'/>
22:17:32:  <log-no-info-header v='true'/>
22:17:32:  <log-redirect v='false'/>
22:17:32:  <log-rotate v='true'/>
22:17:32:  <log-rotate-dir v='logs'/>
22:17:32:  <log-rotate-max v='16'/>
22:17:32:  <log-short-level v='false'/>
22:17:32:  <log-simple-domains v='true'/>
22:17:32:  <log-thread-id v='false'/>
22:17:32:  <log-thread-prefix v='true'/>
22:17:32:  <log-time v='true'/>
22:17:32:  <log-to-screen v='true'/>
22:17:32:  <log-truncate v='false'/>
22:17:32:  <verbosity v='5'/>
22:17:32:
22:17:32:  <!-- Network -->
22:17:32:  <proxy v=':8080'/>
22:17:32:  <proxy-enable v='false'/>
22:17:32:  <proxy-pass v=''/>
22:17:32:  <proxy-user v=''/>
22:17:32:
22:17:32:  <!-- Process Control -->
22:17:32:  <child v='false'/>
22:17:32:  <daemon v='false'/>
22:17:32:  <pid v='false'/>
22:17:32:  <pid-file v='Folding@home Client.pid'/>
22:17:32:  <respawn v='false'/>
22:17:32:  <service v='false'/>
22:17:32:
22:17:32:  <!-- Remote Command Server -->
22:17:32:  <command-address v='0.0.0.0'/>
22:17:32:  <command-allow-no-pass v='127.0.0.1'/>
22:17:32:  <command-deny-no-pass v='0/0'/>
22:17:32:  <command-port v='36330'/>
22:17:32:
22:17:32:  <!-- Slot Control -->
22:17:32:  <idle v='false'/>
22:17:32:  <max-shutdown-wait v='60'/>
22:17:32:  <pause-on-battery v='true'/>
22:17:32:  <pause-on-start v='false'/>
22:17:32:
22:17:32:  <!-- User Information -->
22:17:32:  <machine-id v='0'/>
22:17:32:  <passkey v='********************************'/>
22:17:32:  <team v='0'/>
22:17:32:  <user v='cud0s'/>
22:17:32:
22:17:32:  <!-- Web Server -->
22:17:32:  <session-timeout v='3600'/>
22:17:32:  <web-allow v='127.0.0.1'/>
22:17:32:  <web-deny v='0/0'/>
22:17:32:
22:17:32:  <!-- Work Unit Control -->
22:17:32:  <dump-after-deadline v='true'/>
22:17:32:  <max-queue v='16'/>
22:17:32:  <max-units v='0'/>
22:17:32:  <next-unit-percentage v='99'/>
22:17:32:
22:17:32:  <!-- Folding Slots -->
22:17:32:  <slot id='0' type='GPU'/>
22:17:32:  <slot id='1' type='GPU'>
22:17:32:    <client-type v='advanced'/>
22:17:32:  </slot>
22:17:32:</config>
22:17:32:Trying to access database...
22:17:32:Successfully acquired database lock
22:17:32:Enabled folding slot 00: READY gpu:0:Radeon HD 5870 (Cypress)
22:17:32:Enabled folding slot 01: READY gpu:1:Radeon HD 5870 (Cypress)
22:17:32:Started thread 6 on PID 4800
22:17:32:Started thread 7 on PID 4800
22:17:32:Started thread 5 on PID 4800
22:17:32:Started thread 4 on PID 4800
22:17:32:WU01:FS01:Starting
22:17:32:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/ATI/R600/Core_17.fah/FahCore_17.exe -dir 01 -suffix 01 -version 703 -lifeline 4800 -checkpoint 15 -gpu 1 -gpu-vendor ati
22:17:32:WU01:FS01:Started FahCore on PID 4864
22:17:32:Started thread 8 on PID 4800
22:17:34:WU01:FS01:Core PID:4932
22:17:34:WU01:FS01:FahCore 0x17 started
22:17:35:WU00:FS00:Starting
22:17:35:WU00:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/ATI/R600/Core_16.fah/FahCore_16.exe -dir 00 -suffix 01 -version 703 -lifeline 4800 -checkpoint 15 -gpu 0 -gpu-vendor ati
22:17:35:WU00:FS00:Started FahCore on PID 4952
22:17:35:Started thread 9 on PID 4800
22:17:39:WU00:FS00:Core PID:5116
22:17:39:WU00:FS00:FahCore 0x16 started
22:17:39:WU01:FS01:0x17:*********************** Log Started 2013-08-13T22:17:39Z ***********************
22:17:39:WU01:FS01:0x17:Project: 8900 (Run 468, Clone 0, Gen 104)
22:17:39:WU01:FS01:0x17:Unit: 0x00000087028c126651a68ebdadbd1c48
22:17:39:WU01:FS01:0x17:CPU: 0x00000000000000000000000000000000
22:17:39:WU01:FS01:0x17:Machine: 1
22:17:39:WU01:FS01:0x17:Digital signatures verified
22:17:40:WU00:FS00:0x16:
22:17:40:WU00:FS00:0x16:*------------------------------*
22:17:40:WU00:FS00:0x16:Folding@Home GPU Core
22:17:40:WU00:FS00:0x16:Version 2.11 (Thu Dec 9 15:00:14 PST 2010)
22:17:40:WU00:FS00:0x16:
22:17:40:WU00:FS00:0x16:Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 15.00.30729.01 for 80x86 
22:17:40:WU00:FS00:0x16:Build host: user-f6d030f24f
22:17:40:WU00:FS00:0x16:Board Type: AMD/OpenCL
22:17:40:WU00:FS00:0x16:Core      : x=16
22:17:40:WU00:FS00:0x16: Window's signal control handler registered.
22:17:40:WU00:FS00:0x16:Preparing to commence simulation
22:17:40:WU00:FS00:0x16:- Looking at optimizations...
22:17:40:WU00:FS00:0x16:- Files status OK
22:17:40:WU00:FS00:0x16:sizeof(CORE_PACKET_HDR) = 512 file=<>
22:17:40:WU00:FS00:0x16:- Expanded 45144 -> 171163 (decompressed 379.1 percent)
22:17:40:WU00:FS00:0x16:Called DecompressByteArray: compressed_data_size=45144 data_size=171163, decompressed_data_size=171163 diff=0
22:17:40:WU00:FS00:0x16:- Digital signature verified
22:17:40:WU00:FS00:0x16:
22:17:40:WU00:FS00:0x16:Project: 11293 (Run 1, Clone 460, Gen 44)
22:17:40:WU00:FS00:0x16:
22:17:40:WU00:FS00:0x16:Assembly optimizations on if available.
22:17:40:WU00:FS00:0x16:Entering M.D.
22:17:42:WU00:FS00:0x16:Will resume from checkpoint file 00/wudata_01.ckp
22:17:42:WU00:FS00:0x16:Tpr hash 00/wudata_01.tpr:  1937662157 2724370671 2135592341 2099486007 3617344925
22:17:42:WU00:FS00:0x16:Working on ALZHEIMER DISEASE AMYLOID
22:17:42:WU00:FS00:0x16:Client config unavailable.
22:17:42:WU00:FS00:0x16:Starting GUI Server
22:17:43:Started thread 10 on PID 4800
22:17:45:WU00:FS00:0x16:Resuming from checkpoint
22:17:45:WU00:FS00:0x16:fcCheckPointResume: retreived and current tpr file hash:
22:17:45:WU00:FS00:0x16:   0   1937662157   1937662157
22:17:45:WU00:FS00:0x16:   1   2724370671   2724370671
22:17:45:WU00:FS00:0x16:   2   2135592341   2135592341
22:17:45:WU00:FS00:0x16:   3   2099486007   2099486007
22:17:45:WU00:FS00:0x16:   4   3617344925   3617344925
22:17:45:WU00:FS00:0x16:fcCheckPointResume: file hashes same.
22:17:45:WU00:FS00:0x16:fcCheckPointResume: state restored.
22:17:45:WU00:FS00:0x16:fcCheckPointResume: name 00/wudata_01.log Verified 00/wudata_01.log
22:17:45:WU00:FS00:0x16:fcCheckPointResume: name 00/wudata_01.trr Verified 00/wudata_01.trr
22:17:45:WU00:FS00:0x16:fcCheckPointResume: name 00/wudata_01.xtc Verified 00/wudata_01.xtc
22:17:45:WU00:FS00:0x16:fcCheckPointResume: name 00/wudata_01.edr Verified 00/wudata_01.edr
22:17:45:WU00:FS00:0x16:fcCheckPointResume: state restored 2
22:17:45:WU00:FS00:0x16:Resumed from checkpoint
22:17:45:WU00:FS00:0x16:Setting checkpoint frequency: 500000
22:17:45:WU00:FS00:0x16:Completed   1500001 out of 50000000 steps (3%).
22:20:13:WU01:FS01:0x17:Completed 0 out of 2500000 steps (0%)
22:21:59:WU00:FS00:0x16:Completed   2000000 out of 50000000 steps (4%).

As i said in the title help would be really appreciated

[Joe_H]

First, it is strongly recommended that verbosity be left at the default of 3. Higher levels rarely provide useful information, and get in the way of analyzing your log.

As for Crossfire, it is not utilized by folding and some users report that it gets in the way of successfully folding. Try with Crossfire disabled and see if that improves the situation.

[P5-133XL]

For AMD video cards running core_16 there are problems with newer drivers. The last known good driver for Core_16 is v12.8 and you need to run a driver cleaner before trying to revert from a post 12.8 to a 12.8 version (or earlier). Core_17 works much better for post 12.8 drivers. I note that you are not CPU folding but if you ever do you need to know that for AMD, Core_16 uses a full CPU core (While Core_17 does not for AMD) so you will need to manually modify the number of folding CPU cores to take that into account when CPU folding for over subscribing the CPU is very bad for PPD.

Do not worry about unknown est PPD/TPF on a core_16. That problem is a work-server issue that you have no control over. Not all the work servers are running the most recent software so they do not have the information needed to report back to the client so it can calculate that statistic.

You may wish to revert the verbosity of the client back to 3 (the default). It makes reading the config much easier for verbosity 3 only lists changes from the default while above that lists every setting and really we only need to know what changes from the defaults.

In your log, I do not see anything not working but really you didn't allow the slots enough time to see if Core_17 is progressing or not. Do note that p8900's on Core_17 are very large WU's that can take as much as 30 minutes with 0% GPU usage just to set the WU up and start running. The Core_16 slot seems to be running fine from what I can see for it did progress from 3% to 4%.

[bruce]

The Radeon HD 5xxx are pretty slow GPUs. I've seen reports that suggest that FahCore_17 works very well with high-end GPUs but with that GPU, you may do better with FahCore_16. You'll probably need to perform your own test and decide which works best for you.

[cud0s]

The Radeon HD 5xxx are pretty slow GPUs. I've seen reports that suggest that FahCore_17 works very well with high-end GPUs but with that GPU, you may do better with FahCore_16. You'll probably need to perform your own test and decide which works best for you.

5870x2 is as fast as or faster than 7970 especially where not much vram is needed. 5.44 teraflops in 5870*2 vs 3.79 teraflops in 7970. PS4 will have ~1.8 tflops. Unless you were talking specifically about folding, raw computing power of 5870*2 is quite considerably larger than 7970. So no my precious 5870s are not slow
After disabling crossfire ppd has increased as well as gpu usage on core 16 however i am still getting only ~6000 ppd per gpu. Is this normal? Also i am only getting ~900ppd from my 8320fx and i have configured to run on 6/8 cores in case gpus are using cpu up. Will try to wait longer for core 17 to start. Aldo i am using android tethering to connect to my wifi, could this affect the performance?


EDIT: Ok after more research i have found a quite old database where people with 5870 were getting 5000ppd so i guess with core 16 it is normal. Also i have waited more and core 17 indeed started progressing and now is getting ~18k ppd on each gpu. Quite an improvement, however still not impressive. Any tips how to improve it more? How the hell are guys with 7970 are getting 80k+ ppd?

[7im]

The bonus points are exponential so two low end cards do not equal one high end card. You're pretty well maxed.

[folding_hoomer]

The Radeon HD 5xxx are pretty slow GPUs. I've seen reports that suggest that FahCore_17 works very well with high-end GPUs but with that GPU, you may do better with FahCore_16. You'll probably need to perform your own test and decide which works best for you.

May be the lower-numbered HD5xxx are too slow to profit from using Core17 - the 5870 (and may be 5850/5830) is fast enough.

For comparision from my results:
HD5870@900/1250
Core16: 9,3K (WU11293) up to 12,1K PPD (some "fast" 11292)
Core17: up to 18,5K with 8900/7810/7811

So it is strongly recommended using flag: client-type=advanced and Catalyst 13.6Beta2 to get the max PPD from your GPU´s.

[Joe_H]

5870x2 is as fast as or faster than 7970 especially where not much vram is needed. 5.44 teraflops in 5870*2 vs 3.79 teraflops in 7970. PS4 will have ~1.8 tflops. Unless you were talking specifically about folding, raw computing power of 5870*2 is quite considerably larger than 7970.

As I mentioned in my first post, folding does not utilize Crossfire (or nVidia's SLI), so while for other software two of them in Crossfire are about as fast as a 7970 folding is using them independently. Also folding uses just portions of the GPU for its calculations, specifically the shaders, so the 7970 has an additional advantage there. It however was one of the high-end 5000 series cards and still is quite usable for folding Core_17 as compared to some of the lower end cards in the series.

[cud0s]

Thanks everyone for help, now FAHcontrol shows 40k-20k ppd total which it seems is normal for my hardware. Given the fact that my hardware isn't particulary well suited for folding and uses a lot of electricity do you think it would be simply better to donate the money i would spend on electricity annualy (~1000usd) ?

[7im]

That's completely up to you. Ask if your employer does matching funds for donations (assuming you're not a self employed bitcoiner). You could turn a $1000 in to a $2000 donation.

[bruce]

The bonus points are exponential so two low end cards do not equal one high end card.

This is a very important statement. Finishing one WU in X hours and then finishing a second WU in X more hours is scientifically worth a lot more than dividing the work so that you finish two WU in (X+X) hours. Unfortunately, two GPUs cannot cooperatively work on the same WU; they have to work on independent WUs. Crossfire cannot help with that issue though it can split up the time spent generating pixels for attached displays.