Page 1 of 1
[HELP]GTX1070 SLI
Posted: Sun Oct 30, 2016 6:12 am
by xuj7ct
I've got 4790K+Z97+16G DDR3+1070 SLI（PCIE 8X dual）+WIN 7 X64
Driver version is 372.09
FAH version is the latest one
My problem is:
1.When I used only one 1070 , the PPD was about 550000 , is it a little bit low?
2.When I used dual 1070（in SLI）, the PPD was only about 250000, and one GPU got 80-90% TDP, another is 20-25% TDP
But if I turn off SLI, PPD doesn't change, but the TDP changes to 80-90% and 5%(IDLE)
Re: [HELP]GTX1070 SLI
Posted: Sun Oct 30, 2016 7:54 am
by Joe_H
Welcome to the folding support forum.
Someone else will have to comment on the PPD, as I have not used a 1070 or followed what it was getting. PPD will also vary depending on which project your WU comes from.
As for using both, Folding does not use SLI or two GPU's together to process a single WU. To use both you will need to set up two GPU folding slots, one for each GPU. Each slot will fold its own WU on one of the GPU's.
Re: [HELP]GTX1070 SLI
Posted: Sun Oct 30, 2016 8:13 am
by Nathan_P
Hi, The PPD on the 1st 1070 is roughly in line with what you can expect depending on what WU you are folding and what your card is clocked at.
Also - don't upgrade your drivers thinking that that may help - the current ones, 375.xx, do not work with F@H
Re: [HELP]GTX1070 SLI
Posted: Sun Oct 30, 2016 7:32 pm
by toTOW
If you're using the beta client (7.4.15), there is a bug on multi GPU systems and you'll have to manually assign OpenCL/CUDA indexes to each slot ... and you'll have to do this everytime you reboot the system.
7.4.4 client should remember those informations (but I don't know if it will be able to set them up automatically ...
Re: [HELP]GTX1070 SLI
Posted: Mon Oct 31, 2016 5:36 am
by xuj7ct
Thanks ALL ABOVE.
Hoping stanford or nvidia will improve performace of dual 1070.
Re: [HELP]GTX1070 SLI
Posted: Mon Oct 31, 2016 7:28 am
by rwh202
xuj7ct wrote:Thanks ALL ABOVE.
Hoping stanford or nvidia will improve performace of dual 1070.
I'm not sure much needs improving. A pair of 1070 with a standard install of the 7.4.4 client and 373 drivers gives 1,400,000 PPD with a following wind. If you're not seeing that, then something is wrong with your config - please post the top 100 lines or so of your log so someone can try and point out the mistake.
Re: [HELP]GTX1070 SLI
Posted: Mon Oct 31, 2016 3:22 pm
by xuj7ct
rwh202 wrote:xuj7ct wrote:Thanks ALL ABOVE.
Hoping stanford or nvidia will improve performace of dual 1070.
I'm not sure much needs improving. A pair of 1070 with a standard install of the 7.4.4 client and 373 drivers gives 1,400,000 PPD with a following wind. If you're not seeing that, then something is wrong with your config - please post the top 100 lines or so of your log so someone can try and point out the mistake.
thx for advice.
here is the top lines of my log:
Code: Select all
*********************** Log Started 2016-10-31T15:20:00Z ***********************
15:20:00:************************* Folding@home Client *************************
15:20:00: Website: http://folding.stanford.edu/
15:20:00: Copyright: (c) 2009-2014 Stanford University
15:20:00: Author: Joseph Coffland <joseph@cauldrondevelopment.com>
15:20:00: Args: --open-web-control
15:20:00: Config: C:/Users/liulu/AppData/Roaming/FAHClient/config.xml
15:20:00:******************************** Build ********************************
15:20:00: Version: 7.4.4
15:20:00: Date: Mar 4 2014
15:20:00: Time: 20:26:54
15:20:00: SVN Rev: 4130
15:20:00: Branch: fah/trunk/client
15:20:00: Compiler: Intel(R) C++ MSVC 1500 mode 1200
15:20:00: Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
15:20:00: /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
15:20:00: Platform: win32 XP
15:20:00: Bits: 32
15:20:00: Mode: Release
15:20:00:******************************* System ********************************
15:20:00: CPU: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz
15:20:00: CPU ID: GenuineIntel Family 6 Model 60 Stepping 3
15:20:00: CPUs: 8
15:20:00: Memory: 15.93GiB
15:20:00: Free Memory: 9.70GiB
15:20:00: Threads: WINDOWS_THREADS
15:20:00: OS Version: 6.1
15:20:00: Has Battery: false
15:20:00: On Battery: false
15:20:00: UTC Offset: 8
15:20:00: PID: 16224
15:20:00: CWD: C:/Users/liulu/AppData/Roaming/FAHClient
15:20:00: OS: Windows 7 Enterprise
15:20:00: OS Arch: AMD64
15:20:00: GPUs: 4
15:20:00: GPU 0: UNSUPPORTED: NV3 [PCI]
15:20:00: GPU 1: UNSUPPORTED: NV3 [PCI]
15:20:00: GPU 2: NVIDIA:5 GP104 [GeForce GTX 1070]
15:20:00: GPU 3: NVIDIA:5 GP104 [GeForce GTX 1070]
15:20:00: CUDA: 6.1
15:20:00: CUDA Driver: 8000
15:20:00:Win32 Service: false
15:20:00:***********************************************************************
15:20:00:<config>
15:20:00: <!-- Network -->
15:20:00: <proxy v=':8080'/>
15:20:00:
15:20:00: <!-- Slot Control -->
15:20:00: <power v='full'/>
15:20:00:
15:20:00: <!-- User Information -->
15:20:00: <passkey v='********************************'/>
15:20:00: <team v='3213'/>
15:20:00: <user v='yukaribba'/>
15:20:00:
15:20:00: <!-- Folding Slots -->
15:20:00: <slot id='0' type='GPU'>
15:20:00: <cuda-index v='2'/>
15:20:00: <gpu-index v='2'/>
15:20:00: <opencl-index v='2'/>
15:20:00: </slot>
15:20:00: <slot id='2' type='GPU'>
15:20:00: <cuda-index v='3'/>
15:20:00: <gpu-index v='3'/>
15:20:00: <opencl-index v='3'/>
15:20:00: </slot>
15:20:00:</config>
15:20:00:Trying to access database...
15:20:00:Successfully acquired database lock
15:20:00:Enabled folding slot 00: READY gpu:2:GP104 [GeForce GTX 1070]
15:20:00:Enabled folding slot 02: READY gpu:3:GP104 [GeForce GTX 1070]
15:20:00:WU01:FS00:Starting
15:20:00:WU01:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/liulu/AppData/Roaming/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_21.fah/FahCore_21.exe -dir 01 -suffix 01 -version 704 -lifeline 16224 -checkpoint 15 -gpu 2 -gpu-vendor nvidia
15:20:00:WU01:FS00:Started FahCore on PID 14084
15:20:00:WU01:FS00:Core PID:12672
15:20:00:WU01:FS00:FahCore 0x21 started
15:20:00:WU00:FS02:Starting
15:20:00:WU00:FS02:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/liulu/AppData/Roaming/FAHClient/cores/web.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_18.fah/FahCore_18.exe -dir 00 -suffix 01 -version 704 -lifeline 16224 -checkpoint 15 -gpu 3 -gpu-vendor nvidia
15:20:00:WU00:FS02:Started FahCore on PID 16204
15:20:00:WU00:FS02:Core PID:9656
15:20:00:WU00:FS02:FahCore 0x18 started
15:20:00:WU01:FS00:0x21:*********************** Log Started 2016-10-31T15:20:00Z ***********************
15:20:00:WU01:FS00:0x21:Project: 9205 (Run 17, Clone 24, Gen 1)
15:20:00:WU01:FS00:0x21:Unit: 0x00000006ab436c685796c0af05c1d87a
15:20:00:WU01:FS00:0x21:CPU: 0x00000000000000000000000000000000
15:20:00:WU01:FS00:0x21:Machine: 0
15:20:00:WU01:FS00:0x21:Digital signatures verified
15:20:00:WU01:FS00:0x21:Folding@home GPU Core21 Folding@home Core
15:20:00:WU01:FS00:0x21:Version 0.0.17
15:20:00:WU01:FS00:0x21: Found a checkpoint file
15:20:00:WU00:FS02:0x18:*********************** Log Started 2016-10-31T15:20:00Z ***********************
15:20:00:WU00:FS02:0x18:Project: 9675 (Run 0, Clone 65, Gen 179)
15:20:00:WU00:FS02:0x18:Unit: 0x000000beab436c9b56de69beb9081279
15:20:00:WU00:FS02:0x18:CPU: 0x00000000000000000000000000000000
15:20:00:WU00:FS02:0x18:Machine: 2
15:20:00:WU00:FS02:0x18:Digital signatures verified
15:20:00:WU00:FS02:0x18:Folding@home GPU core18
15:20:00:WU00:FS02:0x18:Version 0.0.4
15:20:00:WU00:FS02:0x18: Found a checkpoint file
15:20:02:WU00:FS02:0x18:Completed 1500000 out of 2000000 steps (75%)
15:20:02:WU00:FS02:0x18:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
15:20:11:WU00:FS02:0x18:Completed 1520000 out of 2000000 steps (76%)
15:20:19:WU00:FS02:0x18:Completed 1540000 out of 2000000 steps (77%)
15:20:20:WU01:FS00:0x21:Completed 1000000 out of 2500000 steps (40%)
15:20:20:WU01:FS00:0x21:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
15:20:41:WU00:FS02:0x18:Completed 1560000 out of 2000000 steps (78%)
15:21:06:WU00:FS02:0x18:Completed 1580000 out of 2000000 steps (79%)
Re: [HELP]GTX1070 SLI
Posted: Mon Oct 31, 2016 4:46 pm
by foldy
It is project 9205 which in known to be too slow and get too low PPD.
see viewtopic.php?f=19&t=29288
So your PC is setup OK and when you get other work units than the slow ones then your second gtx 1070 will get the same speed as the first one
Re: [HELP]GTX1070 SLI
Posted: Mon Oct 31, 2016 5:13 pm
by rwh202
foldy wrote:It is project 9205 which in known to be too slow and get too low PPD.
see viewtopic.php?f=19&t=29288
So your PC is setup OK and when you get other work units than the slow ones then your second gtx 1070 will get the same speed as the first one
I'm not so sure. It looks to me like both WUs are being run on the same card. The tpf for 9675 increases massively when the 9205 starts. Also, the GPU indices look wrong to me. How were they set? The defaults when adding a new slot should work, so I'd
1. finish both WUs
2. configure>slots and delete both GPU slots
3. save config
4 configure>slots and add two fresh GPU slots, leaving defaults
5. save config
6. Fold!
Re: [HELP]GTX1070 SLI
Posted: Tue Nov 01, 2016 5:54 am
by xuj7ct
rwh202 wrote:foldy wrote:It is project 9205 which in known to be too slow and get too low PPD.
see viewtopic.php?f=19&t=29288
So your PC is setup OK and when you get other work units than the slow ones then your second gtx 1070 will get the same speed as the first one
I'm not so sure. It looks to me like both WUs are being run on the same card. The tpf for 9675 increases massively when the 9205 starts. Also, the GPU indices look wrong to me. How were they set? The defaults when adding a new slot should work, so I'd
1. finish both WUs
2. configure>slots and delete both GPU slots
3. save config
4 configure>slots and add two fresh GPU slots, leaving defaults
5. save config
6. Fold!
it really works!appreciate that!
Re: [HELP]GTX1070 SLI
Posted: Sun Nov 06, 2016 10:32 pm
by Frodo The Hobbit
You can configure the CUDA and OpenCL adress with a specific value.
On the Screen "Configure Folding Slot"
- "GPU"
--"GPU Core Indices"
Modify the OpenCL and cuda index as the same as the GPU id. It has to be 3 and 4 following by the adress of the GPU. You have to pause before every modification and unpause when it's done.
This parameter is reset every reboot.