new donor, don't know where to start

Post requests to add new GPUs to the official whitelist here.

Moderators: Site Moderators, FAHC Science Team

Post Reply
RAFTN
Posts: 4
Joined: Tue Mar 31, 2020 12:59 pm

new donor, don't know where to start

Post by RAFTN »

Hi, I just got F@H installed and working. I'm now able to donate CPU and GPU work for some projects.

I'm running an AMD Ryzen 5 3550H CPU with 8 logical processors and a Radeon RX 560X GPU on Win10 Pro and it takes forever to complete WUs on either one. I have the latest Adrenalin software (20.3.1) from AMD.

1) The client is detecting that I have an RX 460 instead -> does this mean my GPU is not yet in the list?
2) I have two OpenCL devices -> is it possible to have these two work on different WU's simultaneously?
3) I can see from my Task Manager that my CPU runs at 100%, but my GPU0 is running at 0% total, but Compute 2 graphs shows 60% utilization, and my GPU1 averages only 3%, shows about 7% on the 3D graph -> Is there a way to increase GPU utilization to speed up work?
4) Is there a way to change the UTC offset to match the timestamp in the Log to match my time?
davidcoton
Posts: 1102
Joined: Wed Nov 05, 2008 3:19 pm
Location: Cambridge, UK

Re: new donor, don't know where to start

Post by davidcoton »

Please post your log if you need help with the configuration. In Advanced Control, Log tab, press Refresh then Copy, and post here in Code tags.

You cannot change the log time (UTC), since that would make it much harder for people here to determine what happened when in sequence with server events.
Image
Joe_H
Site Admin
Posts: 7856
Joined: Tue Apr 21, 2009 4:41 pm
Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2
Location: W. MA

Re: new donor, don't know where to start

Post by Joe_H »

AMD reuses their chips and Device IDs across different series of cards. In your case the RX 460 and RX 560 are almost identical cards, with a variant 560 that has more shaders. The description field in the GPUs.txt file is based on the first report of the Device ID we received, and does not effect how the client uses the card.

Depends on what the other OpenCL device is, but usually not.

Task Manager is not the best tool for seeing GPU utilization. Something like GPU-Z or other GPU monitoring tool may be better. However, how many CPU cores are you folding on? You need to leave at least one CPU thread available to be used to prepare data to be sent to the GPU for processing, and receiving it back.

Greater detail on answers can be given after you post your log to see how your system is configured for folding.
Image

iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
RAFTN
Posts: 4
Joined: Tue Mar 31, 2020 12:59 pm

Re: new donor, don't know where to start

Post by RAFTN »

davidcoton wrote:Please post your log if you need help with the configuration. In Advanced Control, Log tab, press Refresh then Copy, and post here in Code tags.

You cannot change the log time (UTC), since that would make it much harder for people here to determine what happened when in sequence with server events.
this is what my log shows:

Code: Select all

*********************** Log Started 2020-04-02T05:14:35Z ***********************
05:14:35:************************* Folding@home Client *************************
05:14:35:        Website: https://foldingathome.org/
05:14:35:      Copyright: (c) 2009-2018 foldingathome.org
05:14:35:         Author: Joseph Coffland <joseph@cauldrondevelopment.com>
05:14:35:           Args: --open-web-control
05:14:35:         Config: C:\Users\*******\AppData\Roaming\FAHClient\config.xml
05:14:35:******************************** Build ********************************
05:14:35:        Version: 7.5.1
05:14:35:           Date: May 11 2018
05:14:35:           Time: 13:06:32
05:14:35:     Repository: Git
05:14:35:       Revision: 4705bf53c635f88b8fe85af7675557e15d491ff0
05:14:35:         Branch: master
05:14:35:       Compiler: Visual C++ 2008
05:14:35:        Options: /TP /nologo /EHa /wd4297 /wd4103 /Ox /MT
05:14:35:       Platform: win32 10
05:14:35:           Bits: 32
05:14:35:           Mode: Release
05:14:35:******************************* System ********************************
05:14:35:            CPU: AMD Ryzen 5 3550H with Radeon Vega Mobile Gfx
05:14:35:         CPU ID: AuthenticAMD Family 23 Model 24 Stepping 1
05:14:35:           CPUs: 8
05:14:35:         Memory: 7.81GiB
05:14:35:    Free Memory: 2.43GiB
05:14:35:        Threads: WINDOWS_THREADS
05:14:35:     OS Version: 6.2
05:14:35:    Has Battery: true
05:14:35:     On Battery: false
05:14:35:     UTC Offset: 8
05:14:35:            PID: 9748
05:14:35:            CWD: C:\Users\*******\AppData\Roaming\FAHClient
05:14:35:             OS: Windows 10 Enterprise
05:14:35:        OS Arch: AMD64
05:14:35:           GPUs: 1
05:14:35:          GPU 0: Bus:1 Slot:0 Func:0 AMD:5 Baffin XT [Radeon RX 460]
05:14:35:           CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': The
05:14:35:                 specified module could not be found.
05:14:35:
05:14:35:OpenCL Device 0: Platform:0 Device:0 Bus:1 Slot:0 Compute:1.2 Driver:3004.8
05:14:35:OpenCL Device 1: Platform:0 Device:1 Bus:4 Slot:0 Compute:1.2 Driver:3004.8
05:14:35:  Win32 Service: false
05:14:35:***********************************************************************
I'm currently averaging about 2.2167min TPF on CPU and a little under 10min TPF on GPU. is my setup optimal or is there a way to speed it up or dedicate more resources to the client?

Mod Edit: Added Code Tags - PantherX
Last edited by RAFTN on Thu Apr 02, 2020 5:59 am, edited 1 time in total.
RAFTN
Posts: 4
Joined: Tue Mar 31, 2020 12:59 pm

Re: new donor, don't know where to start

Post by RAFTN »

Joe_H wrote:AMD reuses their chips and Device IDs across different series of cards. In your case the RX 460 and RX 560 are almost identical cards, with a variant 560 that has more shaders. The description field in the GPUs.txt file is based on the first report of the Device ID we received, and does not effect how the client uses the card.

Depends on what the other OpenCL device is, but usually not.

Task Manager is not the best tool for seeing GPU utilization. Something like GPU-Z or other GPU monitoring tool may be better. However, how many CPU cores are you folding on? You need to leave at least one CPU thread available to be used to prepare data to be sent to the GPU for processing, and receiving it back.

Greater detail on answers can be given after you post your log to see how your system is configured for folding.
On my folding slots window, it shows cpu:7 and gpu:0, so I am assuming i am using 7 CPU cores and using OpenCL device 0 on my GPU, is that the correct interpretation?
Joe_H
Site Admin
Posts: 7856
Joined: Tue Apr 21, 2009 4:41 pm
Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2
Location: W. MA

Re: new donor, don't know where to start

Post by Joe_H »

That is probably correct. The client will not actually run any WUs on 7 threads of your CPU, but will automatically step that down to 6.

As for the second OpenCL device, that appears to be the integrated Vega Mobile GPU on the Ryzen processor. Most integrated GPUs can not be used for GPU folding, some research would need to be done on the Vega variant used in yours. But even if usable, probably would process about as fast as the Ryzen on 6 CPU threads.
Image

iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
RAFTN
Posts: 4
Joined: Tue Mar 31, 2020 12:59 pm

Re: new donor, don't know where to start

Post by RAFTN »

alright, thank you. so, my current rate is basically the fastest it can go? I can't squeeze any more power out of it without overclocking my processors?
Joe_H
Site Admin
Posts: 7856
Joined: Tue Apr 21, 2009 4:41 pm
Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2
Location: W. MA

Re: new donor, don't know where to start

Post by Joe_H »

Yes, basically you are doing as much as you can on the CPU. They used to allow running on 7 threads, but too many projects had higher error at that. 7 requires slicing up the simulation system into 7 slices, and the thickness may end up too thin. 6 can be used to slice it up in half in one direction and in thirds in another.
Image

iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
extide
Posts: 11
Joined: Fri Apr 27, 2012 9:41 pm

Re: new donor, don't know where to start

Post by extide »

I would actually bet that your dGPU is the second OpenCL device and the iGPU is the first one -- so you can either add another GPU slot and use both, or change the OpenCL index to 1 on your existing slot to use only the dGPU.
Image
Joe_H
Site Admin
Posts: 7856
Joined: Tue Apr 21, 2009 4:41 pm
Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2
Location: W. MA

Re: new donor, don't know where to start

Post by Joe_H »

extide wrote:I would actually bet that your dGPU is the second OpenCL device and the iGPU is the first one -- so you can either add another GPU slot and use both, or change the OpenCL index to 1 on your existing slot to use only the dGPU.
Changing OpenCL index is not needed as long as the folding slot is associated with the correct OpenCL device.

As for adding the iGPU, the 3550H has a TDP of 35 W. That power can go to either or both, but in the case of both that may lead to throttling to keep within the TDP. Someone else posted their results using the Vega 8 iGPU on a different Ryzen, basically about the same speed as processing on the CPU.
Image

iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
Post Reply