System unusable whilst GPU folding

[JimboPalmer]

12:21:15:WU03:FS00:0xa7:Completed 240000 out of 500000 steps (48%)
12:25:07:WU03:FS00:0xa7:Completed 245000 out of 500000 steps (49%)

0xa7 is the CPU folding on Core_a7

12:26:03:WU01:FS01:FahCore 0x22 started

Core_22 starts on the GPU

12:26:05:WU01:FS01:0x22: Found a checkpoint file

It finds a checkpoint from a previous attempt to fold.

12:27:27:WU01:FS01:0x22:Completed 60000 out of 500000 steps (12%)

the previous attempt made it to 12% done.

13:20:33:WU01:FS01:0x22:Completed 105000 out of 500000 steps (21%)
13:22:08:FS01:Shutting core down
13:22:08:WU01:FS01:0x22:WARNING:Console control signal 1 on PID 36704
13:22:08:WU01:FS01:0x22:Exiting, please wait. . .
13:22:08:WU01:FS01:0x22:Folding@home Core Shutdown: INTERRUPTED
13:22:11:WU01:FS01:FahCore returned: INTERRUPTED (102 = 0x66)

And it fails again at 21% done.

13:31:22:WU01:FS01:0x22: Found a checkpoint file
13:32:52:WU01:FS01:0x22:Completed 100000 out of 500000 steps (20%)

It finds a check point at 20% done and starts again

16:38:35:WU01:FS01:0x22:Completed 290000 out of 500000 steps (58%)
16:42:45:FS01:Shutting core down
16:42:45:WU01:FS01:0x22:WARNING:Console control signal 1 on PID 25364
16:42:45:WU01:FS01:0x22:Exiting, please wait. . .
16:42:45:WU01:FS01:0x22:Folding@home Core Shutdown: INTERRUPTED
16:42:49:WU01:FS01:FahCore returned: INTERRUPTED (102 = 0x66)

And it fails again at 58% done. and never attempts to fold in the logs as we see it here.

As always a Log that contains the configuraation portion will give us the most clues why it might be doing this.

My guess and it will always just be a guess is that you have not found a way to be sure you are never folding on the Intel GPU.

Old laptops https://forums.autodesk.com/t5/hardware ... -p/4363279

Newer laptops use the 3d portion of the Nvidia control panel https://www.youtube.com/watch?v=HE6WA2l-14w



The underlying GROMACS science code has issues with large prime numbers (3 is never large, 5 is sometimes large, 7 is always large) and their multiples. In this sense, 13 is prime and 12 is not. So we recommend 12.

[IanWorthington]

Is it not being interrupted because I have the gpu set to run "ON IDLE"?

From the top of the previous post:

Code: Select all

E: SELECT ALL
12:12:31:<config>
12:12:31:  <!-- Network -->
12:12:31:  <proxy v=':8080'/>
12:12:31:
12:12:31:  <!-- Slot Control -->
12:12:31:  <power v='full'/>
12:12:31:
12:12:31:  <!-- User Information -->
12:12:31:  <passkey v='*****'/>
12:12:31:  <team v='10813'/>
12:12:31:  <user v='Ian_Worthington'/>
12:12:31:
12:12:31:  <!-- Folding Slots -->
12:12:31:  <slot id='0' type='CPU'/>
12:12:31:  <slot id='2' type='GPU'>
12:12:31:    <gpu-index v='0'/>
12:12:31:    <idle v='true'/>
12:12:31:  </slot>
12:12:31:  <slot id='1' type='GPU'>
12:12:31:    <gpu-index v='1'/>
12:12:31:    <idle v='true'/>
12:12:31:  </slot>
12:12:31:</config>

Ah, ok. Will set cpu to 12. Any reason I shouldn't try 14 or 15?

[JimboPalmer]

If you are using 'on idle', the the above behavior may be what you want! (I am a 24/7 kinda guy and forget other kinds of folders exist)

14 is 2 * 7 and 7 is bad
15 is 3 * 5 and 5 may be bad

As the client gets newer, it does a better and better job of adjusting automatically. I hope someday not to worry about prime numbers.

[IanWorthington]

I'd be happy to run it 24x7 but the system's unusable if GPU folding is active, which was kind of my initial problem. Then I found that "ON IDLE" maybe just seemed to transfer the work to the built-in GPU, which was then I added the second otherwise useless definition to make /that/ stop.

Seems a somewhat circuitous route. Am I doing the right thing?

Ah, /factors/. With you.

[Joe_H]

Not quite sure why you're saying I'm only folding on the CPU. This (below) seems, to me, to say that I'm folding on the Nvidia, no?

One of your log segments showed GPU folding, most did not. To be honest, what you posted was overkill, and I and probably others only looked at part of what you posted. For future reference to you and others that may come across this topic looking for help, we just need the first section of the log showing the system info and client configuration and enough from other log sections to show successful processing or a problem and the messages proceeding and following that error.

Now there are a few reasons why your system may be less responsive when GPU folding is active beyond what has already been mentioned. You have a mobile GPU and CPU, so I am assuming a laptop or similar non-desktop configuration. With both active you may be reaching the limits of the laptop cooling system and you may be experiencing throttling to stay within this limits. That will decrease responsiveness in a number of ways.

Using the "On Idle" setting for the GPU slot should help, but may only be part of the fix

As mentioned, 1 CPU thread is reserved for the GPU to handle processing,, on Full that will result in a maximum of 15. Medium reduces that to 14, assignments requesting for 14 CPU threads will always be reduced to 12 active threads. Depending on the OS, more threads than that may be associated with the CPU folding core, such as you saw in Resource Manager but the ones beyond those shown as -nt 12 in the log are not involved actively in processing the WU data.

Bruce suggested setting for 12 threads for the CPU folding slot, I am going to make a suggestion for 8. Here are my reasons. Your i9 processor has 8 CPU cores that provide 16 CPU threads through HT. The heat generated by the processor increases by a substantial amount when the circuitry involved in providing the virtual cores through HT becomes active. Your i9 processor is listed with a 45 W TDP, I suspect it has to throttle down to a lower clock speed to stay within that when the processor is continuously active while doing CPU folding. You may not be seeing much if any additional throughput using the 12 CPU thread setting as compared to just using the main CPU cores on the chip. Running at a higher clock speed may also help on responsiveness.

Finally, some of the logs show repeated requests for WU downloads for a Folding Slot 2 (FS02). I did not see that in any of the configurations I spotted while scanning through the logs. FS00 is your CPU slot, FS01 is your GPU slot for the GTX 1650M. What is or was defined for FS02?

[PantherX]

This is the most recent configuration that you have posted:

Code: Select all

<config>
  <!-- Network -->
  <proxy v=':8080'/>

  <!-- Slot Control -->
  <power v='full'/>

  <!-- User Information -->
  <passkey v='*****'/>
  <team v='10813'/>
  <user v='Ian_Worthington'/>

  <!-- Folding Slots -->
  <slot id='0' type='CPU'/>
  <slot id='2' type='GPU'>
    <gpu-index v='0'/>
    <idle v='true'/>
  </slot>
  <slot id='1' type='GPU'>
    <gpu-index v='1'/>
    <idle v='true'/>
  </slot>
</config>

This is the hardware that you have:

Code: Select all

*********************** Log Started 2020-05-24T18:17:52Z ***********************
18:17:52:************************* Folding@home Client *************************
18:17:52:        Website: https://foldingathome.org/
18:17:52:      Copyright: (c) 2009-2018 foldingathome.org
18:17:52:         Author: Joseph Coffland <joseph@cauldrondevelopment.com>
18:17:52:           Args:
18:17:52:         Config: C:\Users\IanWo\AppData\Roaming\FAHClient\config.xml
18:17:52:******************************** Build ********************************
18:17:52:        Version: 7.5.1
18:17:52:           Date: May 11 2018
18:17:52:           Time: 13:06:32
18:17:52:     Repository: Git
18:17:52:       Revision: 4705bf53c635f88b8fe85af7675557e15d491ff0
18:17:52:         Branch: master
18:17:52:       Compiler: Visual C++ 2008
18:17:52:        Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
18:17:52:       Platform: win32 10
18:17:52:           Bits: 32
18:17:52:           Mode: Release
18:17:52:******************************* System ********************************
18:17:52:            CPU: Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
18:17:52:         CPU ID: GenuineIntel Family 6 Model 158 Stepping 13
18:17:52:           CPUs: 16
18:17:52:         Memory: 31.73GiB
18:17:52:    Free Memory: 25.07GiB
18:17:52:        Threads: WINDOWS_THREADS
18:17:52:     OS Version: 6.2
18:17:52:    Has Battery: true
18:17:52:     On Battery: false
18:17:52:     UTC Offset: 2
18:17:52:            PID: 17668
18:17:52:            CWD: C:\Users\IanWo\AppData\Roaming\FAHClient
18:17:52:             OS: Windows 10 Enterprise
18:17:52:        OS Arch: AMD64
18:17:52:           GPUs: 1
18:17:52:          GPU 0: Bus:1 Slot:0 Func:0 NVIDIA:7 TU117M [GeForce GTX 1650 Mobile /
18:17:52:                 Max-Q]
18:17:52:  CUDA Device 0: Platform:0 Device:0 Bus:1 Slot:0 Compute:7.5 Driver:10.2
18:17:52:OpenCL Device 0: Platform:0 Device:0 Bus:1 Slot:0 Compute:1.2 Driver:441.45
18:17:52:OpenCL Device 1: Platform:1 Device:0 Bus:NA Slot:NA Compute:2.1 Driver:26.20
18:17:52:  Win32 Service: false
18:17:52:***********************************************************************

Based upon all the information provided above, this is what your optimized configuration would look like:

Code: Select all

<config>
  <!-- Network -->
  <proxy v=':8080'/>

  <!-- Slot Control -->
  <power v='full'/>

  <!-- User Information -->
  <passkey v='*****'/>
  <team v='10813'/>
  <user v='Ian_Worthington'/>

  <!-- Folding Slots -->
  <slot id='0' type='CPU'>
    <cpus v='8'/>
  </slot>
  <slot id='1' type='GPU'>
    <idle v='true'/>
  </slot>
</config>

You can change the 8 to any other number mentioned in this topic if you wanted too. I have removed the duplicate GPU Slot as that is unnecessary since you have a single GPU supported for folding.