Low PPD with a 970

It seems that a lot of GPU problems revolve around specific versions of drivers. Though NVidia has their own support structure, you can often learn from information reported by others who fold.

Moderators: Site Moderators, FAHC Science Team

Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Low PPD with a 970

Post by Hughesy26 »

Hi All,

I brought a 970 a few months ago, specifically for folding, and was getting around 250kPPD

I recently had to move the card over to a new machine (same OS - win 7) but found PPD dropped to 125k PPD (consistently, on every WU over the last 4 days).

I'm wondering if it's the driver I'm using. I can't remember what I was using before, but now I'm on 353.62. Is there a better one for folding?

Brief extract from log below - let me know if you need more.

Thanks in advance!

Code: Select all

19:55:17:******************************** Build ********************************
19:55:17:      Version: 7.4.4
19:55:17:         Date: Mar 4 2014
19:55:17:         Time: 20:26:54
19:55:17:      SVN Rev: 4130
19:55:17:       Branch: fah/trunk/client
19:55:17:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
19:55:17:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
19:55:17:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
19:55:17:     Platform: win32 XP
19:55:17:         Bits: 32
19:55:17:         Mode: Release
19:55:17:******************************* System ********************************
19:55:17:          CPU: Intel(R) Core(TM) i5-4690K CPU @ 3.50GHz
19:55:17:       CPU ID: GenuineIntel Family 6 Model 60 Stepping 3
19:55:17:         CPUs: 4
19:55:17:       Memory: 7.77GiB
19:55:17:  Free Memory: 6.60GiB
19:55:17:      Threads: WINDOWS_THREADS
19:55:17:   OS Version: 6.1
19:55:17:  Has Battery: false
19:55:17:   On Battery: false
19:55:17:   UTC Offset: 0
19:55:17:          PID: 2452
19:55:17:          CWD: C:/Users/Ellie/AppData/Roaming/FAHClient
19:55:17:           OS: Windows 7 Home Premium
19:55:17:      OS Arch: AMD64
19:55:17:         GPUs: 1
19:55:17:        GPU 0: NVIDIA:5 GM204 [GeForce GTX 970]
19:55:17:         CUDA: 5.2
19:55:17:  CUDA Driver: 7050
19:55:17:Win32 Service: false
19:55:17:***********************************************************************
Mod edit: added Code tags to log
JimboPalmer
Posts: 2522
Joined: Mon Feb 16, 2009 4:12 am
Location: Greenwood MS USA

Re: Low PPD with a 970

Post by JimboPalmer »

The part which shows whether or not you have a passcode on the new system would help.

The part which shows how many of the 4 CPUs are folding would help (One should be reserved to feed the GTX970)
Tsar of all the Rushers
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Re: Low PPD with a 970

Post by Hughesy26 »

Yep, passcode in operation and only 3 of 4 cores folding;

19:55:17: <!-- User Information -->
19:55:17: <passkey v='********************************'/>
19:55:17: <team v='223518'/>
19:55:17: <user v='Hughesy26'/>
19:55:17:
19:55:17: <!-- Folding Slots -->
19:55:17: <slot id='0' type='CPU'/>
19:55:17: <slot id='1' type='GPU'/>
19:55:17:</config>
19:55:17:Trying to access database...
19:55:17:Successfully acquired database lock
19:55:17:Enabled folding slot 00: READY cpu:3
19:55:17:Enabled folding slot 01: READY gpu:0:GM204 [GeForce GTX 970]
toTOW
Site Moderator
Posts: 6349
Joined: Sun Dec 02, 2007 10:38 am
Location: Bordeaux, France
Contact:

Re: Low PPD with a 970

Post by toTOW »

We also need to see which GPU project is running ...
Image

Folding@Home beta tester since 2002. Folding Forum moderator since July 2008.
Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Re: Low PPD with a 970

Post by Hughesy26 »

This is the current one - 110k PPD, but I have run lots of core 18 and core 21 with similar results

14:09:14:WU00:FS01:Started FahCore on PID 8040
14:09:15:WU00:FS01:Core PID:4704
14:09:15:WU00:FS01:FahCore 0x21 started
14:09:15:WU00:FS01:0x21:*********************** Log Started 2015-11-06T14:09:15Z ***********************
14:09:15:WU00:FS01:0x21:Project: 9704 (Run 27, Clone 15, Gen 89)
14:09:15:WU00:FS01:0x21:Unit: 0x0000008bab404162553ebff90a48397f
14:09:15:WU00:FS01:0x21:CPU: 0x00000000000000000000000000000000
14:09:15:WU00:FS01:0x21:Machine: 1
14:09:15:WU00:FS01:0x21:Reading tar file core.xml
14:09:15:WU00:FS01:0x21:Reading tar file system.xml
14:09:16:WU00:FS01:0x21:Reading tar file integrator.xml
14:09:16:WU00:FS01:0x21:Reading tar file state.xml
14:09:18:WU00:FS01:0x21:Digital signatures verified
14:09:18:WU00:FS01:0x21:Folding@home GPU Core21 Folding@home Core
14:09:18:WU00:FS01:0x21:Version 0.0.12
parkut
Posts: 363
Joined: Tue Feb 12, 2008 7:33 am
Hardware configuration: Running exclusively Linux headless blades. All are dedicated crunching machines.
Location: SE Michigan, USA

Re: Low PPD with a 970

Post by parkut »

For comparison, I have a EVGA GTX 960 under Ubuntu Linux 12.04 -
I have captured two 9704s and the associated client's reported PPD at 130k -
Driver Version: 352.41 - The 9704's seem similar to other WU's PPD,
at least on my machine

Code: Select all

Sun Nov 1 14:15:04 EST 2015   9704   0x21   Progress: 17% [|_________]   ppd: 134498   67C
Sun Nov 1 15:01:01 EST 2015   9704   0x21   Progress: 29% [||________]   ppd: 134426   67C
Sun Nov 1 16:01:01 EST 2015   9704   0x21   Progress: 45% [||||______]   ppd: 134360   67C
Sun Nov 1 17:01:01 EST 2015   9704   0x21   Progress: 61% [||||||____]   ppd: 134279   67C
Sun Nov 1 18:01:02 EST 2015   9704   0x21   Progress: 77% [|||||||___]   ppd: 122884   67C
Sun Nov 1 19:01:02 EST 2015   9704   0x21   Progress: 94% [|||||||||_]   ppd: 134106   67C
Thu Nov 5 11:01:01 EST 2015   9704   0x21   Progress: 11% [|_________]   ppd: 138774   66C
Thu Nov 5 12:01:01 EST 2015   9704   0x21   Progress: 28% [||________]   ppd: 138637   66C
Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Re: Low PPD with a 970

Post by Hughesy26 »

Last project finished, new one started. New one runing at 114k PPD, GPU @ 49C

Code: Select all

17:38:23:WU00:FS01:0x21:Completed 608000 out of 640000 steps (95%)
17:40:32:WU00:FS01:0x21:Completed 614400 out of 640000 steps (96%)
17:42:40:WU00:FS01:0x21:Completed 620800 out of 640000 steps (97%)
17:44:49:WU00:FS01:0x21:Completed 627200 out of 640000 steps (98%)
17:46:57:WU00:FS01:0x21:Completed 633600 out of 640000 steps (99%)
17:46:58:WU01:FS01:Connecting to 171.67.108.45:80
17:46:59:WU01:FS01:Assigned to work server 140.163.4.234
17:46:59:WU01:FS01:Requesting new work unit for slot 01: RUNNING gpu:0:GM204 [GeForce GTX 970] from 140.163.4.234
17:46:59:WU01:FS01:Connecting to 140.163.4.234:8080
17:47:01:WU01:FS01:Downloading 3.63MiB
17:47:03:WU01:FS01:Download complete
17:47:03:WU01:FS01:Received Unit: id:01 state:DOWNLOAD error:NO_ERROR project:10477 run:1 clone:290 gen:111 core:0x18 unit:0x00000091538b3dba548b2b6fcea97900
17:49:05:WU00:FS01:0x21:Completed 640000 out of 640000 steps (100%)
17:49:32:WU00:FS01:0x21:Saving result file logfile_01.txt
17:49:32:WU00:FS01:0x21:Saving result file checkpointState.xml
17:49:36:WU00:FS01:0x21:Saving result file checkpt.crc
17:49:36:WU00:FS01:0x21:Saving result file log.txt
17:49:36:WU00:FS01:0x21:Saving result file positions.xtc
17:49:37:WU00:FS01:0x21:Folding@home Core Shutdown: FINISHED_UNIT
17:49:37:WU00:FS01:FahCore returned: FINISHED_UNIT (100 = 0x64)
17:49:37:WU00:FS01:Sending unit results: id:00 state:SEND error:NO_ERROR project:9704 run:27 clone:15 gen:89 core:0x21 unit:0x0000008bab404162553ebff90a48397f
17:49:37:WU00:FS01:Uploading 14.97MiB to 171.64.65.98
17:49:37:WU01:FS01:Starting
17:49:37:WU00:FS01:Connecting to 171.64.65.98:8080
17:49:37:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/Ellie/AppData/Roaming/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_18.fah/FahCore_18.exe -dir 01 -suffix 01 -version 704 -lifeline 2452 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
17:49:37:WU01:FS01:Started FahCore on PID 7276
17:49:37:WU01:FS01:Core PID:1360
17:49:37:WU01:FS01:FahCore 0x18 started
17:49:38:WU01:FS01:0x18:*********************** Log Started 2015-11-06T17:49:37Z ***********************
17:49:38:WU01:FS01:0x18:Project: 10477 (Run 1, Clone 290, Gen 111)
17:49:38:WU01:FS01:0x18:Unit: 0x00000091538b3dba548b2b6fcea97900
17:49:38:WU01:FS01:0x18:CPU: 0x00000000000000000000000000000000
17:49:38:WU01:FS01:0x18:Machine: 1
17:49:38:WU01:FS01:0x18:Reading tar file state.xml
17:49:38:WU01:FS01:0x18:Reading tar file system.xml
17:49:38:WU01:FS01:0x18:Reading tar file integrator.xml
17:49:38:WU01:FS01:0x18:Reading tar file core.xml
17:49:38:WU01:FS01:0x18:Digital signatures verified
17:49:38:WU01:FS01:0x18:Folding@home GPU core18
17:49:38:WU01:FS01:0x18:Version 0.0.4
17:49:43:WU00:FS01:Upload 9.60%
17:49:49:WU00:FS01:Upload 20.46%
17:49:55:WU00:FS01:Upload 31.31%
17:49:57:WU01:FS01:0x18:Completed 0 out of 5000000 steps (0%)
17:49:57:WU01:FS01:0x18:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
17:50:01:WU00:FS01:Upload 42.58%
17:50:07:WU00:FS01:Upload 54.27%
17:50:13:WU00:FS01:Upload 66.79%
17:50:19:WU00:FS01:Upload 78.48%
17:50:25:WU00:FS01:Upload 90.17%
17:51:02:WU00:FS01:Upload complete
17:51:02:WU00:FS01:Server responded WORK_ACK (400)
17:51:02:WU00:FS01:Final credit estimate, 44197.00 points
17:51:02:WU00:FS01:Cleaning up
17:53:57:WU01:FS01:0x18:Completed 50000 out of 5000000 steps (1%)
17:57:52:WU01:FS01:0x18:Completed 100000 out of 5000000 steps (2%)
18:01:53:WU01:FS01:0x18:Completed 150000 out of 5000000 steps (3%)
18:05:47:WU01:FS01:0x18:Completed 200000 out of 5000000 steps (4%)
18:09:41:WU01:FS01:0x18:Completed 250000 out of 5000000 steps (5%)
18:13:43:WU01:FS01:0x18:Completed 300000 out of 5000000 steps (6%)
18:17:37:WU01:FS01:0x18:Completed 350000 out of 5000000 steps (7%)
bollix47
Posts: 2957
Joined: Sun Dec 02, 2007 5:04 am
Location: Canada

Re: Low PPD with a 970

Post by bollix47 »

Not sure where you're seeing the PPD figures that you're quoting but if you check http://www.linuxforge.net/bonuscalc2.php you'll see that you're getting more than you think you are, assuming you're running 24/7. If your figures are coming from FAHControl then you need to wait a few frames (I've seen it take as many as 10 frames) before the estimate settles down to where it should be. Besides using the online calculator you can install HFM.net (in the 3rd party software forum) as it does a pretty good job of estimating PPD usually as soon as the first frame is completed depending on how you set it up.

According to the online calculator and using the TPFs in your latest log the 9704 is actually earning around 300K PPD which is very close to my 970 on that project. Your 10477 is currently earning 252K PPD. Other projects can and do vary from the low 200s to just above 300K ... don't think I've ever seen anything in the low 100s on my 970 but I don't use the Advanced control software to look at the PPD estimates even though they do improve as the WU progresses.

Another possibility you might look at is the GPU you replaced. Was it earning much lower PPD on the same projects? FAHControl may need time before it accurately shows the PPD for the 970. You could 'Finish' the current work and uninstall the client on that computer including data so the old estimates are removed and then re-install so the estimates are only for the 970 work. Whenever I replace a GPU with a different model I always re-install the driver and the client.
Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Re: Low PPD with a 970

Post by Hughesy26 »

Thanks bollix - v interesting and insightful!
I was quoting my PPD from the FAH client itself (although I was waiting until jobs were 10%+ done before paying any attention, I know how inaccurate the early stages can be)
I've tried reinstalling FAH and drivers and no improvement sadly.
I've also downloaded HFM and that is reporting the same ballpark PPD as the client itself - currently 130k PPD on P10486. Interestingly the link you sent suggests that should actually equate to a PPD of 260k
Wierd, right? Any ideas?
I noticed the link states the project is a beta. I've not 'signed up' to betas specifically in the advanced options. Could that be relevant?
bruce
Posts: 20824
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Low PPD with a 970

Post by bruce »

Signing up for beta gives you write access to the beta forum, along with improved opportunities to get support by others running the same project. The actual PPD is always based on the project, itself, plus your passkey.

The PG may re-align the project settings (changing the PPD) if they get reports that those settings are too high or too low.
bollix47
Posts: 2957
Joined: Sun Dec 02, 2007 5:04 am
Location: Canada

Re: Low PPD with a 970

Post by bollix47 »

Okay, let's take it one step at a time:
1. The client itself does not display PPD ... either you're looking at FAHControl (aka Advanced Control) or Web Control. Which one?
2. When you uninstalled FAH did you check the box for Data removal? Did you have to re-enter your folding username, team # and passkey when you installed it again?
3. In HFM go to Edit Preferences, click on the Options tab and set the Calculate PPD option to use Last Frame. After at least 10 percent what is the Estimated TPF that is showing on FAHControl's Status tab? What does HFM show in it's TPF column? Lastly, look at the log tab in FAHControl and manually calculate the time between the last two %s.
4. P10486 is not beta ... it is listed in the regular Project Summary (link at the top of this page). Again, what exactly are you looking at that says beta?
Weird? Yes, so far it's not making a lot of sense.

EDIT: see the next post!!!
Last edited by bollix47 on Fri Nov 06, 2015 1:57 am, edited 1 time in total.
DutchForce
Posts: 60
Joined: Sun Sep 08, 2013 12:43 pm
Location: Netherlands

Re: Low PPD with a 970

Post by DutchForce »

I noticed that your logs show November 06, but you posted on November 05, so it looks like the date is not set correctly.

14:09:15:WU00:FS01:0x21:*********************** Log Started 2015-11-06T14:09:15Z ***********************
14:09:15:WU00:FS01:0x21:Project: 9704 (Run 27, Clone 15, Gen 89)

17:49:38:WU01:FS01:0x18:*********************** Log Started 2015-11-06T17:49:37Z ***********************
17:49:38:WU01:FS01:0x18:Project: 10477 (Run 1, Clone 290, Gen 111)
Image
bollix47
Posts: 2957
Joined: Sun Dec 02, 2007 5:04 am
Location: Canada

Re: Low PPD with a 970

Post by bollix47 »

Great catch DutchForce! That would explain the weirdness ... thanks, I'll sleep better tonight. :wink:
Hughesy26
Posts: 36
Joined: Mon Jul 06, 2015 12:25 pm

Re: Low PPD with a 970

Post by Hughesy26 »

Thanks Dutchforce and Bollix - that fixed it! Much appreciated!! :)
Clocker
Posts: 4
Joined: Fri Dec 04, 2015 12:56 pm

Re: Low PPD with a 970

Post by Clocker »

I'm having a similar but more severe problem with one of my (brand new) GTX 970s. It is only at about 60-65% GPU utilization and only netting about 60-70K PPD. It is a brand new Asus Strix 970. System is fresh install of Windows 10 x64. The CPU is not doing any folding, it is a Celeron J1900. It is headless and I remote into it via Teamviewer.

My other system, using the same type of work unit, downloaded the same day even, is getting closer to 300K PPD. It is an EVGA FTW GTX970, clocked just slightly higher.

The EVGA card is using some older nvidia drivers so I tried those as well as the latest with the Asus card but with no change to the weak production.

The date in Windows appears to be correct. Is there a different date I should be looking at?

The only other thing I can think of is that the Asus GPU power connector is the 8-pin style and I am currently using a 6-pin connector because my 400W PSU does not have the 8-pin style. Right now the entire system is drawing about 110W from the wall, so I am sure the PSU is capable from a wattage perspective.

Looking for any help you guys can provide!

Thanks,
Clocker
Post Reply