list of acceptable supported Nvidia cards

[Fldngbob]

I have been searching and cannot locate an accurate list of supported video cards.

With v7.6.21 I had several Nvidia Quadro K620 folding but, this model now lists as "unsupported" with v8.

I researched and as I have a Quadro P620 that IS supported with v8 working.

So, I bought a Quadro P400 (slightly older than a P620) that I thought would be supported but, sadly is listed as "unsupported" by v8.

With the evolution of FAH v8, is there a list of acceptable supported Nvidia cards (for that matter AMD cards, too?)
The other option to determine supported video cards could be an education in minimum software requirements?
Someone, please help?

[muziqaz]

https://apps.foldingathome.org/GPUs.txt
The ones with zero are not supported.
Looks for species 6 or higher for optimal performance.
For example:
0x1002:0x1305:1:2:Kaveri [Radeon Mobility]
Bold number is your species designator. It is the last digit before the name of the GPU. The higher it is the newer and faster it is.
Please keep in mind this list is constantly updated, to add or remove the GPUs. And Nvidia species designator is not equal to AMD or intel. So AMD species 11 is not faster than nVidia species 8

[Joe_H]

Supported Nvidia cards are anything using Kepler or newer microarchitecture. However Kepler based cards will only use OpenCL code for folding, Maxwell and newer GPUs can use CUDA for processing which provides a speedup over OpenCL.

The same GPUs are supported by v7 and v8. However there are differences in detecting an installation of OpenCL and CUDA support.

So your Quadro K620 should work, it is Maxwell based. It may depend on which drivers are installed and which OS. The P620 and P400 based on the newer Pascal chips both should also work, again proper driver installation and OS support is needed. The entries for all three cards are:

Code: Select all

0x10de:0x13bb:2:4:GM107GL [Quadro K620]
0x10de:0x1cb6:2:5:GP107GL [Quadro P620]
0x10de:0x1cb2:2:5:GP107GL [Quadro P600]

So the K620 is species 4, the other two are species 5.

Ideally if you have the money for a used GTX 1600 series, RTX 2000 series, or equivalent Turing based workstation cards would be better. Driver support is kept more up to date on those. I haven't followed the used market for those series cards though.

In general both Nvidia and AMD tend to drop full support on the older cards in the drivers. AMD especially has done this with older cards and drivers to be used on Linux where you may find video driver support only without support for OpenCL processing. The same card may still work under Windows using OpenCL.

[appepi]

Ideally if you have the money for a used GTX 1600 series, RTX 2000 series, or equivalent Turing based workstation cards would be better.

There is a good article on the used GPU market (US) at https://www.techspot.com/bestof/used-gpu-24/. I followed previous versions of these GPU pricing articles for a while and compared the prices when looking on eBay in AU in 2022, and when the prices became more reasonable I bought some pre-loved ASUS Turbo RTX 2060s in the second half of 2022. Installed in Z440 workstations (which had previously had K620s) they have been running for at least 9 hours/day ever since and have racked up about 690 Mpoints each - see stats for "donors" Z441, Z442 and Z443. I bought the single-fan ASUS Turbo models mainly because I needed a card that fits into a Z440 case without modification, and other variants are too deep. But also they are "uncool" and less likely to have been hammered in someone's gaming rig. I also avoided 2070s because they had been very popular for mining. I use ASUS GPU Tweak III and an aggressive fan profile to keep temperatures around 70 deg C, and I bought some spare fans just in case. The snag is that the RTX 2000 series is now more than two years older. That being so, the eBay.au prices I saw just now for 2060's seem way too high. Another wrinkle is that there are (at least) two chip variants and vendors usually don't spell that out. Mine are your standard TU106 and hate project 18251 - see posts on that topic for details. Good luck.

[Fldngbob]

Having slept on my issue and having read everyone's answers (thank you, btw) I realized I had forgotten that my main desktop (that has the P620 card that IS folding) is a different OS than all my others. So, I thought the problem could be the video drivers difference between Arch Linux OS and Debian OS. So, this machine has two video card slots. I was not using the second video card slot I moved the P400 to that second slot and restarted machine. The P400 did not appear on v8-4.foldingathome.org/machines summary webpage.

Here's where I realized something was weird. The P620 card has always been folding units. In "edit machine settings" when I hover the curse the P620 (that is checked as ENABLED) pops up a Unsupported GPU message as does the P400 card. But, the P620 had been folding since update to v8 now, since adding the P400 the P620 is no longer folding and does not appear v8-4.foldingathome.org/machines summary webpage. So, I removed the P400 card and despite the P620 card is enabled (in settings) it no longer appears on the v8-4.foldingathome.org/machines summary webpage. Now, obviously P620 is no longer folding as it had before.

My Arch Linux has:
Driver version 570.86.16
CUDA version 12.8

My Debian (12) Bookworm has:
Driver vesion 570.86.10
CUDA version 12.8

Both cards show as "Unsupported GPU" however the P620 was folding units but, not any more?

I could be incorrect but, because P620 was and now is not folding, what broke? Is it something in the FAH 8.4.9 program or is it driver related?

Is anyone else experiencing this false report of "Unsupported GPU" when in fact the card WAS folding and installing the P400 broke the driver and stopped the P620?

I'm anxious to see anyone's thoughts?

[Fldngbob]

A followup on my issue. On my Debian (12) client I removed the standard FAH v8.4.9 package and installed the "debug" v8.4.9 listed on the download page. Over at v8-4.foldingathome.org/machines I see that the client I am diagnosing now reports the cpu folding a unit AND a second 'Project' trying to run. The second running Project Status displays an orange outlined hour glass. Hovering (the cursor) over the hourglass displays "Resources not available." The log shows:

18:44:25:I3:Loading default resource group
18:44:25:I1:Listening for HTTP on 127.0.0.1:7396
18:44:25:D3:account = {"pubkey": "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAxwJ02K5FvsXA2TvnYQydzr3roERigrjKx9iTme8OvRB3iSruneI57azPH7NRnKw+urbcY/o+f4TuWbFtvcFfVpVqInCyIcVXyUwWNzRrUQkcbc8Rch3/4a47x0saynsd4u0oW8K9mgRx79DLP8ddov53NHfpgFma9eRFlXuFTldN6cKCdCtl4lvZcBhjqeBj2wJvAMpLXZDGUOHvxV6uaOZU91e9toLyfsZ+UfR9B/S0KTox7yGjh1Ky5gHV0X8DVe2ElK79AJN5HNrjtMsaGtRvRewov+RxUiTidY8LQOyzGMSZ7khLZYsXBjaYY7aMQ+iTyGwTEZV*************************************************************************C+9m7xZraRyTvdJEP7S8oDDgmftbepUvZRly3KbbP7gSGpAJ8vMhsS77anI/Nfc/zhkxVq5hbH1ujhqKHXHAF9yIBb7/YpB1zMiTUwqSeZp79GPzZsZ3bXQEuZ2zi7H3w571MLMD+CbK9TT/OexUGUDGgUgkF2oSYDNAtt3I4dQlvrg5LKyucCAwEAAQ==", "user": "rbrtewdn", "team": 233601, "passkey": "daa3b4a59b7a9********************9b7a9f64", "cause": "alzheimers", "node": "node1.foldingathome.org", "mach_name": "fah004"}
18:44:25:I3:WU24:Loading work unit 24 with ID LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs
18:44:25:I3:WU25:Loading work unit 25 with ID 1mSaHLaHiDVr3DVaEwYMugHBgcwY3y80WhXyBDa0mi4
18:44:25:I3:Loaded 2 wus.
18:44:26:D3:OpenCL not supported: clGetPlatformIDs() returned -1001
18:44:26:D3:CUDA not supported: cuInit() returned 100
18:44:26:I3:gpus = {
18:44:26:I3: "gpu:01:00:00": {"vendor": 4318, "device": 7347, "type": "nvidia", "supported": false, "description": "GP107GL [Quadro P400]"}
18:44:26:I3:}
18:44:26:D3:Default:Unit LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs enabled
18:44:26:D3:Default:Unit 1mSaHLaHiDVr3DVaEwYMugHBgcwY3y80WhXyBDa0mi4 paused
18:44:26:D1:Default:Remaining CPUs: 0, Remaining GPUs: 0, Active WUs: 1
18:44:27:W :CON1:DNS lookup failed for api.foldingathome.org
18:44:27:D3:CON1:Connection failed
18:44:27:E :OUT1:Failed response: CONNECT
18:44:27:I1:Loaded cores/fahcore-a8-lin-64bit-sse2-0.0.12/FahCore_a8
18:44:27:D3:Default:Unit LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs enabled
18:44:27:D3:Default:Unit 1mSaHLaHiDVr3DVaEwYMugHBgcwY3y80WhXyBDa0mi4 paused
18:44:27:D1:Default:Remaining CPUs: 0, Remaining GPUs: 0, Active WUs: 1
18:44:27:I3:Running FahCore: /var/lib/fah-client/cores/fahcore-a8-lin-64bit-sse2-0.0.12/FahCore_a8 -dir LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs -suffix 01 -version 8.4.9 -lifeline 411 -np 3
18:44:27:D3:Executing: /var/lib/fah-client/cores/fahcore-a8-lin-64bit-sse2-0.0.12/FahCore_a8 -dir LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs -suffix 01 -version 8.4.9 -lifeline 411 -np 3
18:44:27:I3:WU24:Started FahCore on PID 732
18:44:27:D3:Default:Unit LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs enabled
18:44:27:D3:Default:Unit 1mSaHLaHiDVr3DVaEwYMugHBgcwY3y80WhXyBDa0mi4 paused
18:44:27:D1:Default:Remaining CPUs: 0, Remaining GPUs: 0, Active WUs: 1

On the client I can manually 'ping' api.foldingathome.org with good response times.

Just for giggles I "trashed" that second running Project to see if the gpu would try again to collect a new project. The project "trashed" as expected however here is the log entry:

19:01:33:I1:WU24:Calling: mdrun -c frame25.gro -s frame25.tpr -x frame25.xtc -cpi state.cpt -cpt 5 -nt 3 -ntmpi 1
19:01:33:I1:WU24:Steps: first=3125000 total=3250000
19:01:34:I1:OUT1:< HTTP/1.1 200 HTTP_OK
19:01:34:D3:account = {"pubkey": "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAxwJ02K5FvsXA2TvnYQydzr3roERigrjKx9iTme8OvRB3iSruneI57azPH7NRnKw+urbcY/o+f4TuWbFtvcFfVpVqInCyIcVXyUwWNzRrUQkcbc8Rch3/4a47x0saynsd4u0oW8K9mgRx79DLP8ddov53NHfpgFma9eRFlXuFTldN6cKCdCtl4lvZcBhjqeBj2wJvAMpLXZDGUOHvxV6uaOZU91e9toLyfsZ+UfR9B/S0KTox7yGjh1Ky5gHV0X8DVe2ElK79AJN5HNrjtMsaGtRvRewov+RxUiTidY8LQOyzGMSZ7khLZYsXBjaYY7aMQ+iTyGwTEZVED83PdvR0IH/I6UKK8OOjAurkAJiraAg5Wkd44v3jvp60glU20KU/Fl50P3SAspdoX3lxUBJ2OfRO3BYm3TcVbd8/7z4bakUe05VTFELmj7dTTq2MuadwK******************************************************************************************zZsZ3bXQEuZ2zi7H3w571MLMD+CbK9TT/OexUGUDGgUgkF2oSYDNAtt3I4dQlvrg5LKyucCAwEAAQ==", "user": "rbrtewdn", "team": 233601, "passkey": "daa3b4a59b7a9*************7a9f64", "cause": "alzheimers", "node": "node1.foldingathome.org", "mach_name": "fah004"}
19:01:34:I1:OUT2:> GET wss://node1.foldingathome.org/ws/client HTTP/1.1
19:01:34:I1:OUT2:< HTTP/1.1 101 HTTP_SWITCHING_PROTOCOLS
19:01:34:I1:Logging into node account
19:01:34:D3:received = {"type": "message", "id": "2toqYJKS06gQJut3tLzMTBTzzRK3Ow4arB3mewfOc2E", "iv": "4PZvOPtlgfODcQY9-GrWlA", "payload": "maXmU5aaoqoOOSDjGIikpAiOrHsgtRwUUQloU08TRSq-c35NWnTyYhY818NdBFzPVY8oFlgcc09v9VpVx2xjhw"}
19:01:34:D3:Adding remote -BD_UixPP5DQGZKz
19:01:34:D3:New client -BD_UixPP5DQGZKz
19:01:35:D3:Default:Unit LRq3QEH1aqdaibmQEVFH8DaLaSfQvH7pt6b6Pq4QlMs enabled
19:01:35:D1:Default:Remaining CPUs: 0, Remaining GPUs: 0, Active WUs: 1
19:01:35:D3:received = {"type": "message", "id": "2toqYJKS06gQJut3tLzMTBTzzRK3Ow4arB3mewfOc2E", "iv": "SaVQbMch7tO8UMxlHipMqg", "payload": "ONCoUYBD1cOMdaGCLA-tdEiKMylN_Ond6Uug5Gp5fxDa9VEW2e1mi6aOkK_AUkWJUqognyoGXrWhM7-fXAZT8LDgqCF88f6yM73JakacDNc6Gt6Rx_VJGA5Do_0x9j7Cj7LozM8pz188wz-hIni4fS-Hy2oRcz4LxO7xLuSH8BA"}
19:01:35:D3:msg: {"enable": true, "cmd": "log", "time": "2025-03-01T19:01:35.069Z"}
19:01:37:I1:WU24:Completed 11152 out of 125000 steps (8%)
19:02:30:I1:WU24:Completed 11250 out of 125000 steps (9%)
19:02:39:D3:received = {"type": "message", "id": "2toqYJKS06gQJut3tLzMTBTzzRK3Ow4arB3mewfOc2E", "iv": "9vZNjdgs0VYLb9RKSxCclQ", "payload": "nIr7I3-iYIgDHHtBLdedoNjYYyoa3PnStTeQgLlQFGtdVFcNnUbXNZ848BShR7KT799danI2WcpyimQn6Nn8ZTuzdYAQOcpvxFLEAFSQlrmAk8QZeSh7W1YH3frNLfXl2h0yXxuem3885zQ7nChlnUQm8l6TLEKy0_hSL0RHfBM"}
19:02:39:D3:msg: {"enable": false, "cmd": "log", "time": "2025-03-01T19:02:39.133Z"}
19:05:59:D3:received = {"type": "message", "id": "2toqYJKS06gQJut3tLzMTBTzzRK3Ow4arB3mewfOc2E", "iv": "ORwOYgTBW62-t6J9fqMogQ", "payload": "X2AHSfouBxlK4JH3LqsOrrWBj8PsCDkbRW1KeHbN8HdQHYgJ8-ZLaKreDJ47QwoX6v5mlqRujJ74R3AD9TNQDqZ9aqRvnLPfLwuXUXFJGHVf3l6fQb4qIAuMyUOzAQ2_SRJd4ttW22s7HigID-LXC5lSK8Wtvh2aPvripwJBHj4"}
19:05:59:D3:msg: {"enable": true, "cmd": "log", "time": "2025-03-01T19:05:59.158Z"}

I am not sure what exactly the "debug" version can debug nor do I know how to collect any info other than the log output. The only thing I found that looked odd was the DNS failure. Anyone's help on that would be greatly appreciated.

Is there a work around to solve this?

Thoughts?

[muziqaz]

You do not need debug version to see what your client is doing and how many things it is folding. Web UI can do it perfectly well. Debug version is for code related debugging.
"Resources not available" means there is no work available for that GPU, and it is perfectly normal, especially on older GPUs.

[toTOW]

I have been searching and cannot locate an accurate list of supported video cards.

With v7.6.21 I had several Nvidia Quadro K620 folding but, this model now lists as "unsupported" with v8.

I researched and as I have a Quadro P620 that IS supported with v8 working.

So, I bought a Quadro P400 (slightly older than a P620) that I thought would be supported but, sadly is listed as "unsupported" by v8.

With the evolution of FAH v8, is there a list of acceptable supported Nvidia cards (for that matter AMD cards, too?)
The other option to determine supported video cards could be an education in minimum software requirements?
Someone, please help?

If you are only buying to run FAH, don't go below Pascal generation.

Kepler has been End Of Life for nVidia for a while and is barely supported on newer FAH cores.

Maxwell and Pascal are already announced as "feature complete" and will no longer get new features. So they are very likely next in the list of unsupported hardware ...

[Fldngbob]

A question for toTOW:

Pascal are already announced as "feature complete" and will no longer get new features.

Who "already announced as "feature complete", Nvidia or FAH?

So they are very likely next in the list of unsupported hardware

Where can I find this "list of unsupported hardware"?

[toTOW]

A question for toTOW:

Pascal are already announced as "feature complete" and will no longer get new features.

Who "already announced as "feature complete", Nvidia or FAH?

https://docs.nvidia.com/cuda/cuda-toolk ... hitectures

1.5.1. Deprecated Architectures

Architecture support for Maxwell, Pascal, and Volta is considered feature-complete and will be frozen in an upcoming release.

Note that nVidia no longer provide driver updates for Kepler based GPUs.

So they are very likely next in the list of unsupported hardware

Where can I find this "list of unsupported hardware"?

There's no such list for FAH ... the only official requirement is that GPU need to support CUDA 11.2 or OpenCL 1.2.

[Fldngbob]

I have been tinkering with this for two weeks (on and off) and have discovered the following:

If I install the OS with the Nvidia card inserted the install will include the Nouveau driver. Install the distro repo Nvidia driver, OpenCL, and CUDA the Nouveau driver is black listed and Nvidia drivers take control. FAH will let the card request and bring to receive a WU. That WU ultimately crashes and appears as an orange (color) hourglass. All the while the CPU has it's own WU and folding is progressing.

All cards I am trying to use are on the white list (this fact was pointed out in an earlier post by Joe H. I think.) Thank you.

Code: Select all

0x10de:0x13bb:2:4:GM107GL [Quadro K620]
0x10de:0x1cb3:2:5:GP107GL [Quadro P400]
0x10de:0x1cb6:2:5:GP107GL [Quadro P620]

I ran 'user@fah014:~# inxi -Fxxxmprz' and saw this:

Code: Select all

Graphics:
  Device-1: NVIDIA GP107GL [Quadro P400] vendor: Hewlett-Packard driver: nvidia v: 535.216.01
    arch: Pascal pcie: speed: 2.5 GT/s lanes: 16 bus-ID: 01:00.0 chip-ID: 10de:1cb3 class-ID: 0300
  Device-2: NVIDIA GP107GL [Quadro P400] vendor: Hewlett-Packard driver: nvidia v: 535.216.01
    arch: Pascal pcie: speed: 2.5 GT/s lanes: 1 bus-ID: 05:00.0 chip-ID: 10de:1cb3 class-ID: 0300
  Device-3: NVIDIA GM107GL [Quadro K620] driver: nvidia v: 535.216.01 arch: Maxwell pcie:
    speed: 2.5 GT/s lanes: 1 bus-ID: 06:00.0 chip-ID: 10de:13bb class-ID: 0300

Either I am missing a piece of software that I am not aware that I need or (please, no) I am the only person experiencing this type of problem?
So, I surely cannot be the ONLY user with this issue?

Tomorrow (Saturday 3-15) I am scheduled to receive a Zotac GTX1060 3Gb card listed on the white list as:

Code: Select all

[0x10de:0x1c02:2:5:GP106 [GeForce GTX 1060 3GB] 3935

As you can see it is a GP106. I'll see what happens.

More soon.

[BobWilliams757]

To be honest, I think a lot of people are troubleshooting issues that seem to only exist in the beta client. It IS a beta client still, and while it might be fine in most cases, that statistic doesn't help when you are not one of the "most cases". There are some little bugs even in the older client, but usually few and far between.

I'd revert back to 7.6.21 and see if the GPU in question works.

[Fldngbob]

Bob, Thanks for the honest viewpoint. I appreciate your thoughts.

Well, I agree that my issue is not one of the "most cases" problem.

Reverting back to v7 for most of my machines has been a plan "C" thought for many weeks now. I think I'll leave one machine on v8 and wait for an update to fahclient v8.4.9 software. The cpu is still folding after all.

I'd revert back to 7.6.21 and see if the GPU in question works.

v7.6.21 worked with my video cards until I upgraded to v8.4.9.

Hi ho, hi ho, it's back to v7 I go.