AMD GPU not folding

[Bonven]

Hello,
I have 7 GPUs of different types, all of them folding except one.
This one GPU receives WUs, then it stays at 0.0% until time expires and finally it downloads the next WU just to do the same.
I just received a new WU, so I thought I would investigate.
This is the end of the log:

Code: Select all

08:44:20:WU00:FS03:Requesting new work unit for slot 03: RUNNING gpu:2:Curacao Pro [Radeon Rx 270/370] from 128.252.203.10
08:44:20:WU00:FS03:Connecting to 128.252.203.10:8080
08:45:51:WU00:FS03:Downloading 50.73MiB
08:45:57:WU00:FS03:Download 5.54%
08:46:03:WU00:FS03:Download 9.61%
08:46:09:WU00:FS03:Download 20.45%
08:46:15:WU00:FS03:Download 40.28%
08:46:21:WU00:FS03:Download 57.41%
08:46:27:WU00:FS03:Download 61.97%
08:46:33:WU00:FS03:Download 64.55%
08:46:39:WU00:FS03:Download 67.02%
08:46:45:WU00:FS03:Download 69.36%
08:46:51:WU00:FS03:Download 72.31%
08:46:57:WU00:FS03:Download 74.90%
08:47:03:WU00:FS03:Download 77.12%
08:47:09:WU00:FS03:Download 79.58%
08:47:15:WU00:FS03:Download 83.03%
08:47:21:WU00:FS03:Download 85.37%
08:47:27:WU00:FS03:Download 89.68%
08:47:33:WU00:FS03:Download 93.01%
08:47:38:WU00:FS03:Download complete
08:47:39:WU00:FS03:Received Unit: id:00 state:DOWNLOAD error:NO_ERROR project:11762 run:0 clone:1353 gen:59 core:0x22 unit:0x0000007980fccb0a5e6d80d6eda7e2eb

What I am investigating is, using GPU-Z the GPU is a Pitcairn, while it is being listed as Curacao Pro.
Device ID is 0x1002:0x6811, which is listed as Rx270/370 in the Whitelist, while it is correctly Named R9 200 series (I think it's a R9 270) in GPU-Z.
So my question is, besides any correction that can be made server side, is there a command in config to force the type of GPU?

[JimboPalmer]

Searching Pitcairn I get this, near the bottom are cards made with the chip, none named RX 270

https://www.techpowerup.com/gpu-specs/amd-pitcairn.g420

Searching Curacao, I get this, again the named card are near the bottom, and none ar.e RX 270

https://www.techpowerup.com/gpu-specs/amd-curacao.g758

Showing the Configuration portion of the log will help, the first 200 lines, say.

[bruce]

We can adjust the man-readable information between the [..] but that doesn't change the device or it's performance. the 0x6811 device ID is all that matters. See https://www.techpowerup.com/gpu-specs/r ... -270.c2458.

AMD used the same processing chip on a variety of different GPUs, including the R9 270.

How are you determining whether it's processing or not? GPU-Z sensor page should show a high percentage on GPU Load and a reasonable temperature. Video Engine Load will probably be close to 0.

If you're looking a the Windows Task Monitor, it generally shows CPU resources being used just to move data, not the work that the GPU is doing.

[Bonven]

Thank you!
Actually the GPU runs at 97% but it is now attempting a cleanup, retried 14 times already, so I guess it will go on indefinitely.

Code: Select all

*********************** Log Started 2020-05-19T11:23:19Z ***********************
11:23:19:****************************** FAHClient ******************************
11:23:19:        Version: 7.6.9
11:23:19:         Author: Joseph Coffland <joseph@cauldrondevelopment.com>
11:23:19:      Copyright: 2020 foldingathome.org
11:23:19:       Homepage: https://foldingathome.org/
11:23:19:           Date: Apr 17 2020
11:23:19:           Time: 11:13:06
11:23:19:       Revision: 398c2b17fa535e0cc6c9d10856b2154c32771646
11:23:19:         Branch: master
11:23:19:       Compiler: Visual C++ 2008
11:23:19:        Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
11:23:19:       Platform: win32 10
11:23:19:           Bits: 32
11:23:19:           Mode: Release
11:23:19:           Args: --open-web-control
11:23:19:         Config: C:\Users\Ariel\AppData\Roaming\FAHClient\config.xml
11:23:19:******************************** CBang ********************************
11:23:19:           Date: Apr 17 2020
11:23:19:           Time: 11:10:09
11:23:19:       Revision: 2fb0be7809c5e45287a122ca5fbc15b5ae859a3b
11:23:19:         Branch: master
11:23:19:       Compiler: Visual C++ 2008
11:23:19:        Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
11:23:19:       Platform: win32 10
11:23:19:           Bits: 32
11:23:19:           Mode: Release
11:23:19:******************************* System ********************************
11:23:19:            CPU: Intel(R) Celeron(R) CPU G3900 @ 2.80GHz
11:23:19:         CPU ID: GenuineIntel Family 6 Model 94 Stepping 3
11:23:19:           CPUs: 2
11:23:19:         Memory: 7.96GiB
11:23:19:    Free Memory: 5.97GiB
11:23:19:        Threads: WINDOWS_THREADS
11:23:19:     OS Version: 6.2
11:23:19:    Has Battery: false
11:23:19:     On Battery: false
11:23:19:     UTC Offset: 2
11:23:19:            PID: 2064
11:23:19:            CWD: C:\Users\Ariel\AppData\Roaming\FAHClient
11:23:19:             OS: Windows 10 Home
11:23:19:        OS Arch: AMD64
11:23:19:           GPUs: 5
11:23:19:          GPU 0: Bus:15 Slot:0 Func:0 AMD:5 R575A [Radeon R7 250X/HD 7700/8760]
11:23:19:          GPU 1: Bus:14 Slot:0 Func:0 AMD:5 Hawaii [Radeon R7 200 Series/HD
11:23:19:                 8670]
11:23:19:          GPU 2: Bus:19 Slot:0 Func:0 AMD:5 Curacao Pro [Radeon Rx 270/370]
11:23:19:          GPU 3: Bus:1 Slot:0 Func:0 NVIDIA:0 GT218 [GeForce 210]
11:23:19:          GPU 4: Bus:22 Slot:0 Func:0 AMD:5 Ellesmere XT [Radeon RX
11:23:19:                 470/480/570/580/590]
11:23:19:  CUDA Device 0: Platform:0 Device:0 Bus:1 Slot:0 Compute:1.2 Driver:6.5
11:23:19:OpenCL Device 0: Platform:0 Device:0 Bus:0 Slot:0 Compute:1.2 Driver:2527.7
11:23:19:OpenCL Device 1: Platform:0 Device:1 Bus:0 Slot:0 Compute:1.2 Driver:2527.7
11:23:19:OpenCL Device 2: Platform:0 Device:2 Bus:0 Slot:0 Compute:1.2 Driver:2527.7
11:23:19:OpenCL Device 3: Platform:0 Device:3 Bus:0 Slot:0 Compute:1.2 Driver:2527.7
11:23:19:OpenCL Device 5: Platform:1 Device:0 Bus:1 Slot:0 Compute:1.0 Driver:342.1
11:23:19:  Win32 Service: false
11:23:19:******************************* libFAH ********************************
11:23:19:           Date: Apr 15 2020
11:23:19:           Time: 14:53:14
11:23:19:       Revision: 216968bc7025029c841ed6e36e81a03a316890d3
11:23:19:         Branch: master
11:23:19:       Compiler: Visual C++ 2008
11:23:19:        Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
11:23:19:       Platform: win32 10
11:23:19:           Bits: 32
11:23:19:           Mode: Release
11:23:19:***********************************************************************
11:23:19:<config>
11:23:19:  <!-- Folding Core -->
11:23:19:  <checkpoint v='30'/>
11:23:19:
11:23:19:  <!-- Folding Slot Configuration -->
11:23:19:  <cause v='HIGH_PRIORITY'/>
11:23:19:
11:23:19:  <!-- HTTP Server -->
11:23:19:  <allow v='0.0.0.0/0'/>
11:23:19:  <deny v=''/>
11:23:19:
11:23:19:  <!-- Network -->
11:23:19:  <proxy v=':8080'/>
11:23:19:
11:23:19:  <!-- Remote Command Server -->
11:23:19:  <command-allow-no-pass v='0.0.0.0/0'/>
11:23:19:  <command-deny-no-pass v=''/>
11:23:19:
11:23:19:  <!-- Slot Control -->
11:23:19:  <power v='FULL'/>
11:23:19:
11:23:19:  <!-- User Information -->
11:23:19:  <passkey v='*****'/>
11:23:19:  <team v='264503'/>
11:23:19:  <user v='Stefano_Ariel_Bonvini'/>
11:23:19:
11:23:19:  <!-- Folding Slots -->
11:23:19:  <slot id='0' type='CPU'/>
11:23:19:  <slot id='1' type='GPU'>
11:23:19:    <opencl-index v='0'/>
11:23:19:  </slot>
11:23:19:  <slot id='2' type='GPU'>
11:23:19:    <opencl-index v='1'/>
11:23:19:  </slot>
11:23:19:  <slot id='3' type='GPU'>
11:23:19:    <opencl-index v='2'/>
11:23:19:  </slot>
11:23:19:  <slot id='6' type='GPU'>
11:23:19:    <client-type v='advanced'/>
11:23:19:    <cuda-index v='0'/>
11:23:19:  </slot>
11:23:19:  <slot id='5' type='GPU'>
11:23:19:    <opencl-index v='3'/>
11:23:19:  </slot>
11:23:19:</config>
11:23:19:Trying to access database...
11:23:19:Successfully acquired database lock
11:23:19:Enabled folding slot 00: READY cpu:1
11:23:19:Enabled folding slot 01: READY gpu:0:R575A [Radeon R7 250X/HD 7700/8760]
11:23:19:Enabled folding slot 02: READY gpu:1:Hawaii [Radeon R7 200 Series/HD 8670]
11:23:19:Enabled folding slot 03: READY gpu:2:Curacao Pro [Radeon Rx 270/370]
11:23:19:Enabled folding slot 06: READY gpu:3:GT218 [GeForce 210]
11:23:19:Enabled folding slot 05: READY gpu:4:Ellesmere XT [Radeon RX 470/480/570/580/590]

Here is the setup: I had to build it manually because it wouldn't find any GPU in auto.
I might have switched openCl devices somehow, but all other GPUs work, except for the GeForce 210 which is not supported.
I am now leaving office, so I won't be able to provide further data until monday, except what I already know.

[JimboPalmer]

OpenCL device 5 is using the Nvidia Driver
0 to 3 are using AMD drivers (a current driver would be over 3000)

F@H requires 1 CPU thread per supported GPU, so you are over booked by 2, more if you leave in the GT210 slot.
That should only hurt performance, but honestly, performance is what F@H is about.

11:23:19:Enabled folding slot 01: READY gpu:0:R575A [Radeon R7 250X/HD 7700/8760]
11:23:19: <slot id='1' type='GPU'>
11:23:19: <opencl-index v='0'/>
11:23:19:OpenCL Device 0: Platform:0 Device:0 Bus:0 Slot:0 Compute:1.2 Driver:2527.7

11:23:19:Enabled folding slot 02: READY gpu:1:Hawaii [Radeon R7 200 Series/HD 8670]
11:23:19: <slot id='2' type='GPU'>
11:23:19: <opencl-index v='1'/>
11:23:19:OpenCL Device 1: Platform:0 Device:1 Bus:0 Slot:0 Compute:1.2 Driver:2527.7

11:23:19:Enabled folding slot 03: READY gpu:2:Curacao Pro [Radeon Rx 270/370]
11:23:19: <slot id='3' type='GPU'>
11:23:19: <opencl-index v='2'/>
11:23:19:OpenCL Device 2: Platform:0 Device:2 Bus:0 Slot:0 Compute:1.2 Driver:2527.7

11:23:19:Enabled folding slot 05: READY gpu:4:Ellesmere XT [Radeon RX 470/480/570/580/590]
11:23:19: <slot id='5' type='GPU'>
11:23:19: <opencl-index v='3'/>
11:23:19:OpenCL Device 3: Platform:0 Device:3 Bus:0 Slot:0 Compute:1.2 Driver:2527.7

11:23:19:Enabled folding slot 06: READY gpu:3:GT218 [GeForce 210]
11:23:19: <slot id='6' type='GPU'>
11:23:19: <client-type v='advanced'/>
11:23:19: <cuda-index v='0'/>
Neither advanced nor CUDA is helping you here, It should be Opencl index 5

But really, just delete the slot.

[Bonven]

So I looked into your recommendations and solved the issue completely, so I thought I just recap everything here as successful actions in my case.
The basic was I did not understand the relationship btw GPU and OpenCl in the System Info.
Anyhow in the meantime I removed all GPUs, tested them one by one and reinstalled only the AMD ones.
I actually had an Intel onboard VGA, which I switched on for display.
When trying to update the AMD driver, I encountered all sort of difficulties.
But in the end I found out that there is an AMD driver in windows 10, which was automatically installed for each GPU.
So this is now the situation:

Code: Select all

OS: Windows 10 Home
OS Arch: AMD64
GPUs: 4
GPU 0: Bus:18 Slot:0 Func:0 AMD:5 Curacao Pro [Radeon Rx 270/370]
GPU 1: Bus:12 Slot:0 Func:0 AMD:5 Ellesmere XT [Radeon RX470/480/570/580/590]
GPU 2: Bus:13 Slot:0 Func:0 AMD:5 Hawaii [Radeon R7 200 Series/HD8670]
GPU 3: Bus:21 Slot:0 Func:0 AMD:5 R575A [Radeon R7 250X/HD 7700/8760]
CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': Impossibile trovare il modulo specificato.

OpenCL Device 0: Platform:0 Device:0 Bus:12 Slot:0 Compute:1.2 Driver:3004.8
OpenCL Device 1: Platform:0 Device:1 Bus:21 Slot:0 Compute:1.2 Driver:3004.8
OpenCL Device 2: Platform:0 Device:2 Bus:18 Slot:0 Compute:1.2 Driver:3004.8
OpenCL Device 3: Platform:0 Device:3 Bus:13 Slot:0 Compute:1.2 Driver:3004.8
OpenCL Device 4: Platform:1 Device:0 Bus:NA Slot:NA Compute:2.1 Driver:26.20

Now what I did, which I didn't understand before, was setting on slot 1 GPU 0 and OpenCl 2, because they refer to the same Bus.
This is my setup:
Slot 0: 1 CPU
Slot 1: GPU 0 OpenCl 2 (Bus 18)
Slot 2: GPU 1 OpenCl 0 (Bus 12)
Slot 3: GPU 2 OpenCl 3 (Bus 13)
Slot 4: GPU 3 OpenCl 1 (Bus 21)

As soon as I got this, the machine started folding at top speed and I do almost 500k points per day.
I know this whole explaination is redundant, but as I have seen some confusion on this matter, I hope a very plain explaination will help someone else.

As I installed the GPUs, the system booked 1 CPU thread for all of them (luckily!).
And I didn't install any video driver, leaving it all to Windows 10.

Thank you all for your support.

[JimboPalmer]

I am glad you got it sorted! And I am glad you reported back, sometimes I wonder if my 'help' actually helped.