So a quick question, does anyone just have the occasional disappearing GPU. Been running a 4060ti with no issue. I usually come home to find one of the dockers with Nvidia set up having errors. Has occurred twice so far. The first time i reconnected the monitor until i found a driver that worked. Put my dummy hdmi back in a different slot but was okay. I think i was using either 580 or 590 drivers. Recently 595 drivers came out and it auto updated, had to restart system and things seemed okay. It has just happened again. Not detecting GPU it dropped while unmanic was processing so gave a swag of errors.

I have tried 590.48.01 and 595.58.03 both production and latest but same error. Any thoughts, nvdia-smi reports nothing system devices shows iommu removed. i doubt it but will test monitor next

14 minutes ago, crazyc said:

I have tried 590.48.01 and 595.58.03 both production and latest but same error. Any thoughts, nvdia-smi reports nothing system devices shows iommu removed. i doubt it but will test monitor next

Do you have Diagnostics or anything after the GPU dropped? Without Diagnostics it's really hard to say anything.

Most of the times it has to do with either a Power Supply which is not up to the task anymore or PCIe Bus issues or even BIOS settings set wrong, sometimes plugging in a Monitor or a HDMI Dummy plug helps and in some cases even PowerTop or Power Saving scripts are being to aggressive or enabling nvidia-persistenced.

I will pull a diagnostic later..Just annoying how its working fine then gone with no apparent reason,, GPU and PSU are basically new motherboard and cpu are old asus Z170pro gaming and an i7 6700k.

About to have dinner then will check gpu is working as monitor is fine....oh the joy

crazytower-diagnostics-20260327-1913.zip

Diagnostic attached.

After three reboots all is working again WTF ????

GPU Stats show 01:00.0:RTX 4060 TI - 0000:01:00.0 when opening the settings cog

NVIDIA Driver is Production Branch V595.58.03

Installed GPU(s):
No devices were found

GPU Driver Support:

01:00.0: VGA compatible controller [0300]: NVIDIA Corporation AD106 [GeForce RTX 4060 Ti]
Detected via: lspci
Chip codename: AD106
Architecture: Ada
Kernel module support: both
Device ID input: 10de:2803
Normalized ID: 2803
Supported candidates: 580.142, 590.48.01, 595.58.03
Recommended Driver: 590.48.01 (open-source)

nvidia-smi says Unable to determine the device handle for GPU0: 0000:01:00.0: Unknown Error

Edited March 27Mar 27 by crazyc

6 minutes ago, crazyc said:

Diagnostic attached.

You don't have to add anything if you post the Diagnostics, I see almost everything you have posted there, however you have an Xid error in your log:

Mar 27 18:22:47 craZyTower kernel: NVRM: GPU at PCI:0000:01:00: GPU-9b9953bb-c437-09fa-c7a0-cb535d0b6baa

Mar 27 18:22:47 craZyTower kernel: NVRM: Xid (PCI:0000:01:00): 79, GPU has fallen off the bus.

Mar 27 18:22:47 craZyTower kernel: NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.

Mar 27 18:22:47 craZyTower kernel: NVRM: A GPU crash dump has been created. If possible, please run

Mar 27 18:22:47 craZyTower kernel: NVRM: nvidia-bug-report.sh as root to collect this data before

Mar 27 18:22:47 craZyTower kernel: NVRM: the NVIDIA kernel module is unloaded.

Mar 27 18:22:47 craZyTower kernel: NVRM: Xid (PCI:0000:01:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (GPU Reset Required)

You can see the full list of Xid errors here, you have #79

However this is a pretty bad issue to troubleshoot since it could be pretty much everything, however I would recommend that you do the following:

1. Since you are on the latest BIOS version, try to reset your BIOS settings to defaults

2. Try to connect a monitor and let it connected

3. Try to add nvidia-persistenced at the very end to your go file

Thanks @ich777

Will check out the bios, it has been untouched for several years.

I connected the monitor and after a reboot it worked. I then reconnected the HDMI dummy and it worked

I will "add nvidia-persistenced at the very end to your go file" Very new to unraid so will work that one out.

And wow with the errors..

Many thanks for your prompt assistance

After adding the nvidia-persistenced it semes to have fallen off again while transcoding or it could of been me editing the file. Doing a reboot, i read if i use the following: nvidia-persistenced && sleep 5 && fuser -k /dev/nvidiactl it is safer if i decide to do any VM's later. So am trying that

UPDATE: Current uptime is 29m and Unaminic compressing without issue....

Edited March 27Mar 27 by crazyc

1hr 49m uptime and dropped gpu yet again again..should i try nvtop instead of gpu stats..some reading has pointed to this

crazytower-diagnostics-20260327-2155.zip

Edited March 27Mar 27 by crazyc

So have removed Dummy Hdmi and reconnected monitor. Have removed gpu stats and loaded NVtop instead will see how long this last. Wonder if the old motherboard has finally bit the dust or is it just Friday ? I am Lost..

Edited March 27Mar 27 by crazyc

1 minute ago, crazyc said:

So have removed Dummy Hdmi and reconnected monitor. Have removed gpu stats and loaded NVtop instead will see how long this last. Wonder if the old motherboard has finally bit the dust or is it just Friday ?

You say that your PSU is nearly new, but is it powerful enough? I was getting drop-outs after upgrading my GPU. Swapped for a higher power PSU and it's been rock-solid ever since - must have been tipping it over the edge. Turns out you can't just keep adding drives and cards and expect it to work 🙃

1 hour ago, crazyc said:

1hr 49m uptime and dropped gpu yet again again..should i try nvtop instead of gpu stats..some reading has pointed to this

Try to remove it entirely for now and see if that helps.

However please don't forget that hardware degrades over time and you actually have a Kernel Panic this time:

Mar 27 21:51:34 craZyTower kernel: NVRM: GPU at PCI:0000:01:00: GPU-9b9953bb-c437-09fa-c7a0-cb535d0b6baa

Mar 27 21:51:34 craZyTower kernel: NVRM: Xid (PCI:0000:01:00): 79, pid=369925, name=ffmpeg, GPU has fallen off the bus.

Mar 27 21:51:34 craZyTower kernel: NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.

Mar 27 21:51:34 craZyTower kernel: NVRM: GPU0 GSP RPC buffer contains function 78 (DUMP_PROTOBUF_COMPONENT) sequence 314407 and data 0x0000000000000000 0x0000000000000000.

Mar 27 21:51:34 craZyTower kernel: NVRM: GPU0 RPC history (CPU -> GSP):

Mar 27 21:51:34 craZyTower kernel: NVRM: entry function sequence data0 data1 ts_start ts_end duration actively_polling

Mar 27 21:51:34 craZyTower kernel: NVRM: 0 103 GSP_RM_ALLOC 314382 0x000000000000c7b5 0x0000000000000008 0x00064e008ab1e719 0x0000000000000000 y

Mar 27 21:51:34 craZyTower kernel: NVRM: -1 103 GSP_RM_ALLOC 314381 0x000000000000c9c0 0x0000000000000010 0x00064e008ab1e5c9 0x00064e008ab1e6fe 309us

Mar 27 21:51:34 craZyTower kernel: NVRM: -2 103 GSP_RM_ALLOC 314380 0x000000000000c56f 0x0000000000000170 0x00064e008ab1e3ea 0x00064e008ab1e5a8 446us

Mar 27 21:51:34 craZyTower kernel: NVRM: -3 103 GSP_RM_ALLOC 314379 0x000000000000c7b5 0x0000000000000008 0x00064e008ab1e267 0x00064e008ab1e341 218us

Mar 27 21:51:34 craZyTower kernel: NVRM: -4 103 GSP_RM_ALLOC 314378 0x000000000000c9c0 0x0000000000000010 0x00064e008ab1e10e 0x00064e008ab1e24c 318us

Mar 27 21:51:34 craZyTower kernel: NVRM: -5 103 GSP_RM_ALLOC 314377 0x000000000000c56f 0x0000000000000170 0x00064e008ab1ded4 0x00064e008ab1e0e5 529us

Mar 27 21:51:34 craZyTower kernel: NVRM: -6 103 GSP_RM_ALLOC 314376 0x000000000000c7b5 0x0000000000000008 0x00064e008ab1db4e 0x00064e008ab1dc2b 221us

Mar 27 21:51:34 craZyTower kernel: NVRM: -7 103 GSP_RM_ALLOC 314375 0x000000000000c9c0 0x0000000000000010 0x00064e008ab1d9ea 0x00064e008ab1db28 318us

Mar 27 21:51:34 craZyTower kernel: NVRM: GPU0 RPC event history (CPU <- GSP):

Mar 27 21:51:34 craZyTower kernel: NVRM: entry function sequence data0 data1 ts_start ts_end duration during_incomplete_rpc

Mar 27 21:51:34 craZyTower kernel: NVRM: 0 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e907f52 0x00064dff6e907f5b 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -1 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e81385b 0x00064dff6e813864 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -2 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e536981 0x00064dff6e53698a 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -3 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e44219d 0x00064dff6e4421a6 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -4 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e34da34 0x00064dff6e34da3d 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -5 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e2592a2 0x00064dff6e2592ab 9us

Mar 27 21:51:34 craZyTower kernel: NVRM: -6 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e164ae6 0x00064dff6e164aee 8us

Mar 27 21:51:34 craZyTower kernel: NVRM: -7 4099 POST_EVENT 0 0x00000000000000a2 0x0000000000000000 0x00064dff6e0702d3 0x00064dff6e0702dc 9us

Mar 27 21:51:34 craZyTower kernel: CPU: 7 UID: 99 PID: 369925 Comm: ffmpeg Tainted: P O 6.12.54-Unraid #1

Mar 27 21:51:34 craZyTower kernel: Tainted: [P]=PROPRIETARY_MODULE, [O]=OOT_MODULE

Mar 27 21:51:34 craZyTower kernel: Hardware name: System manufacturer System Product Name/Z170 PRO GAMING, BIOS 3805 05/16/2018

Mar 27 21:51:34 craZyTower kernel: Call Trace:

Mar 27 21:51:34 craZyTower kernel: <TASK>

Mar 27 21:51:34 craZyTower kernel: dump_stack_lvl+0x47/0x70

Mar 27 21:51:34 craZyTower kernel: _nv013654rm+0x81c/0xd20 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv013558rm+0xe5/0x880 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv051028rm+0x156/0x390 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv052169rm+0x424/0xab0 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv052167rm+0x264/0x4b0 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv050395rm+0x171/0x300 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv050396rm+0x5c/0x90 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv000593rm+0x6b/0x80 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? _nv000739rm+0xa36/0xd00 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? rm_ioctl+0x7f/0x400 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? nvidia_unlocked_ioctl+0x72e/0x7b0 [nvidia]

Mar 27 21:51:34 craZyTower kernel: ? vfs_ioctl+0x20/0x40

Mar 27 21:51:34 craZyTower kernel: ? __do_sys_ioctl+0x53/0x70

Mar 27 21:51:34 craZyTower kernel: ? do_syscall_64+0x68/0xe0

Mar 27 21:51:34 craZyTower kernel: ? entry_SYSCALL_64_after_hwframe+0x76/0x7e

Mar 27 21:51:34 craZyTower kernel: </TASK>

Mar 27 21:51:34 craZyTower kernel: NVRM: Xid (PCI:0000:01:00): 154, GPU recovery action changed from 0x0 (None) to 0x2 (Node Reboot Required)

Mar 27 21:51:34 craZyTower kernel: ------------[ cut here ]------------

Mar 27 21:51:34 craZyTower kernel: WARNING: CPU: 1 PID: 369926 at /tmp/makeself.9LJqJdNd/NVIDIA-Linux-x86_64-595.58.03/kernel/nvidia/nv.c:5250 nvidia_dev_put_uuid+0x33/0x50 [nvidia]

Mar 27 21:51:34 craZyTower kernel: Modules linked in: nft_ct cmac cifs cifs_arc4 nls_ucs2_utils cifs_md4 dns_resolver nft_masq nft_chain_nat algif_rng af_alg nvidia_uvm(PO) br_netfilter xt_nat nf_conntrack_netlink veth xfrm_user xfrm_algo xt_addrtype xt_connmark xt_conntrack ip6table_mangle iptable_mangle xt_MASQUERADE xt_tcpudp xt_mark tun nf_tables nfnetlink ip6table_nat iptable_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 md_mod zfs(PO) spl(O) ntfs3 tcp_diag inet_diag i915 drm_buddy i2c_algo_bit drm_display_helper intel_gtt nct6775 nct6775_core hwmon_vid jc42 regmap_i2c ip6table_filter ip6_tables iptable_filter ip_tables x_tables bridge stp llc bonding tls nvidia_drm(PO) nvidia_modeset(PO) intel_rapl_msr intel_rapl_common iosf_mbi x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel mei_pxp mei_hdcp nvidia(PO) kvm drm_ttm_helper crct10dif_pclmul crc32_pclmul ttm crc32c_intel ghash_clmulni_intel sha512_ssse3 sha256_ssse3 sha1_ssse3 drm_kms_helper aesni_intel crypto_simd cryptd rapl intel_cstate drm intel_uncore i2c_i801

Mar 27 21:51:34 craZyTower kernel: wmi_bmof mxm_wmi e1000e agpgart i2c_smbus mei_me mei i2c_core thermal fan video wmi backlight acpi_pad button

Mar 27 21:51:34 craZyTower kernel: CPU: 1 UID: 99 PID: 369926 Comm: cuda00001400006 Tainted: P O 6.12.54-Unraid #1

Mar 27 21:51:34 craZyTower kernel: Tainted: [P]=PROPRIETARY_MODULE, [O]=OOT_MODULE

Mar 27 21:51:34 craZyTower kernel: Hardware name: System manufacturer System Product Name/Z170 PRO GAMING, BIOS 3805 05/16/2018

Mar 27 21:51:34 craZyTower kernel: RIP: 0010:nvidia_dev_put_uuid+0x33/0x50 [nvidia]

Mar 27 21:51:34 craZyTower kernel: Code: 53 e8 01 c7 ff ff 48 85 c0 74 2f 48 89 c3 48 89 c7 48 89 ee e8 6e d7 ff ff 31 d2 48 89 de 48 89 ef e8 a1 9b c3 00 85 c0 74 02 <0f> 0b 48 8d bb b8 06 00 00 5b 5d e9 8d 99 5e e1 5b 5d c3 cc cc cc

Mar 27 21:51:34 craZyTower kernel: RSP: 0018:ffffc900016dfbc8 EFLAGS: 00010202

Mar 27 21:51:34 craZyTower kernel: RAX: 0000000000000026 RBX: ffff88816ac18000 RCX: 0000000000000001

Mar 27 21:51:34 craZyTower kernel: RDX: 0000000000000000 RSI: 0000000000000246 RDI: ffffc900016dfaf8

Mar 27 21:51:34 craZyTower kernel: RBP: ffff88801e1cb000 R08: 0000000000000000 R09: 0000000000000000

Mar 27 21:51:34 craZyTower kernel: R10: ffffc900016dfb70 R11: 0000000000000080 R12: ffff8881786ea000

Mar 27 21:51:34 craZyTower kernel: R13: ffffc9000b38b078 R14: ffffc9000b38c0f0 R15: 0000000000000000

Mar 27 21:51:34 craZyTower kernel: FS: 0000000000000000(0000) GS:ffff888816c40000(0000) knlGS:0000000000000000

Mar 27 21:51:34 craZyTower kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033

Mar 27 21:51:34 craZyTower kernel: CR2: 00001551ac088000 CR3: 0000000005618004 CR4: 00000000003726f0

Mar 27 21:51:34 craZyTower kernel: Call Trace:

Mar 27 21:51:34 craZyTower kernel: <TASK>

Mar 27 21:51:34 craZyTower kernel: nvUvmInterfaceUnregisterGpu+0x28/0x30 [nvidia]

Mar 27 21:51:34 craZyTower kernel: uvm_gpu_release_locked+0x27/0x40 [nvidia_uvm]

Mar 27 21:51:34 craZyTower kernel: uvm_va_space_destroy+0x365/0x3c0 [nvidia_uvm]

Mar 27 21:51:34 craZyTower kernel: uvm_release.isra.0+0x8d/0x140 [nvidia_uvm]

Mar 27 21:51:34 craZyTower kernel: uvm_release_entry.part.0.isra.0+0x47/0x80 [nvidia_uvm]

Mar 27 21:51:34 craZyTower kernel: ? task_work_run+0x68/0x80

Mar 27 21:51:34 craZyTower kernel: uvm_release_entry+0x21/0x30 [nvidia_uvm]

Mar 27 21:51:34 craZyTower kernel: __fput+0x10a/0x1d0

Mar 27 21:51:34 craZyTower kernel: task_work_run+0x68/0x80

Mar 27 21:51:34 craZyTower kernel: do_exit+0x36f/0x8d0

Mar 27 21:51:34 craZyTower kernel: do_group_exit+0x79/0x80

Mar 27 21:51:34 craZyTower kernel: get_signal+0x617/0x660

Mar 27 21:51:34 craZyTower kernel: ? __seccomp_filter+0x85/0x370

Mar 27 21:51:34 craZyTower kernel: ? vfs_read+0x159/0x1c0

Mar 27 21:51:34 craZyTower kernel: arch_do_signal_or_restart+0x28/0x1d0

Mar 27 21:51:34 craZyTower kernel: syscall_exit_to_user_mode+0x4f/0x80

Mar 27 21:51:34 craZyTower kernel: do_syscall_64+0x82/0xe0

Mar 27 21:51:34 craZyTower kernel: entry_SYSCALL_64_after_hwframe+0x76/0x7e

Mar 27 21:51:34 craZyTower kernel: RIP: 0033:0x148092bea4fd

Mar 27 21:51:34 craZyTower kernel: Code: Unable to access opcode bytes at 0x148092bea4d3.

Mar 27 21:51:34 craZyTower kernel: RSP: 002b:0000148085a1a900 EFLAGS: 00000293 ORIG_RAX: 0000000000000007

Mar 27 21:51:34 craZyTower kernel: RAX: fffffffffffffdfc RBX: 0000000000000000 RCX: 0000148092bea4fd

Mar 27 21:51:34 craZyTower kernel: RDX: 00000000ffffffff RSI: 0000000000000003 RDI: 000055d89b64ec00

Mar 27 21:51:34 craZyTower kernel: RBP: 0000148085a1a920 R08: 0000000000000000 R09: 0000000000000007

Mar 27 21:51:34 craZyTower kernel: R10: 000055d89b68cfe0 R11: 0000000000000293 R12: 0000000000000000

Mar 27 21:51:34 craZyTower kernel: R13: 000055d89b688180 R14: 000055d89b64ec00 R15: 000055d89b688040

Mar 27 21:51:34 craZyTower kernel: </TASK>

Mar 27 21:51:34 craZyTower kernel: ---[ end trace 0000000000000000 ]---

Something is definitely crashing the driver.

As i said i hope it is just a Friday.. GU and PSU are less than 2 years old. Mobo, cpu and ram are 5-7 years at least....as always working fine until they are not.. This time around with Monitor attached Cpu Stats removes and NVtop install current uptime 1hr 6 min happily compressing with unmanic. One more glass of red the =bed and see what happens tomorrow, Will move back to hdmi dummy plug and see what happens

Edited March 27Mar 27 by crazyc

I woke to nvtop saying no GPU and nvidia-smi no devices found it does show under system devices. Any thing different this time in the Diagnocrazytower-diagnostics-20260328-0917.zipstics ?

10 minutes ago, crazyc said:

I woke to nvtop saying no GPU and nvidia-smi no devices found it does show under system devices. Any thing different this time in the Diagnocrazytower-diagnostics-20260328-0917.zipstics ?

Yeah, a lot of things:

Mar 28 02:37:29 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:37:29 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:37:41 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:37:41 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:38:09 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:38:09 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Mar 28 02:38:27 craZyTower kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x22:0x51:884)

Do you have somewhere a dashboard open or something that calls nvidia-smi repeatedly?

However as said, this seems more like a hardware issue.

so strange , reboot server brings it back to life...Nothing that i am aware of that calls nvidia-smi.

I have Homepage on another server that shows a few container statuses on this one . I will have to see if i can check the hardware, The PSU is a 1200w device about 6 months old. I need to do some outside then will come back and attempt reseating everything. This issue has only just started yesterday I do have another motherboard and ram. At a loss but trial and error I guess. Now that I think of it Unmanic had an update maybe 1 to 2 days aga and i have been using it each tim the gpu dropped. Ity just had another update may see if it is anything there, could be grasping at straws

Thanks again

Edited March 28Mar 28 by crazyc

Quick question whiich driver do you recommend for my 4060ti, the production or latest V595.58.03 or the open source v590.48.01 ?

Also having similar issues with my GTX 1070. I have had no issues for the last 2-3 years and now randomly the GPU is dropped and I get nvidia-smi reporting no GPUs found.

1 hour ago, Larvey said:

Also having similar issues with my GTX 1070. I have had no issues for the last 2-3 years and now randomly the GPU is dropped and I get nvidia-smi reporting no GPUs found.

Please always post your Diagnostics.

2 hours ago, crazyc said:

Quick question whiich driver do you recommend for my 4060ti, the production or latest V595.58.03 or the open source v590.48.01 ?

You can use both, both will work fine with your card.

Thank you. I got a solid 2 1/2 hours before it dropped this time. I have another Mobo and CPU and Memory that i can test when time permits. This is frustrating

2 hours ago, crazyc said:

Thank you. I got a solid 2 1/2 hours before it dropped this time. I have another Mobo and CPU and Memory that i can test when time permits. This is frustrating

Sorry about that but as said, this seems like a hardware issue and nothing the driver can solve.

All good, actually got 22hrs before it dropped the gpu...Will start prepping some parts to test

Just tried 7.3 b2 and don't see a way of selecting v580.142 for my Quadro P600 card?

Nvidia Driver Plugin - Upgrade Path Breaks Pascal GPUs (Driver Branch Not Honored)

I ran into an issue after upgrading to the latest Nvidia Driver (and at the same time I had upgraded from 7.2.2 to 7.2.3 - not sure which, but one of them caused my Nvidia GPU (Quadro P2000, Pascal) to disappear entirely from the system because it upgraded the driver to 595.x from 580.142 (which is the latest version that supports the P2000).

Environment:

• Unraid: 7.2.3 (kernel 6.12.54)

• GPU: NVIDIA Quadro P2000 (Pascal, device ID 10de:1c30)

• Plugin: Nvidia Driver

• Original working driver: 580.x branch

Issue observed after upgrade:

• Driver was automatically set to 595.58.03 (Production/latest)

• nvidia-smi failed: “couldn't communicate with the NVIDIA driver”

• /dev/nvidia* devices were not created

• dmesg repeatedly showed:

  • GPU detected as P2000

  • but “supported through 580.xx Legacy drivers”

  • “595.xx driver will ignore”

  • “No NVIDIA GPU found”

Result:

• Docker containers requiring GPU (Plex in this case) failed to start due to missing /dev/nvidia0

• System behavior appeared as if GPU was not present

Resolution:

• Manually changed preferred driver version to 580.142

• Rebooted

• GPU immediately restored:

  • /dev/nvidia* present

  • nvidia-smi functional

  • Plex container started normally

Root cause (as understood):

• Plugin upgrade or Unraid upgrade defaulted driver selection to latest/Production branch

• That branch (595.x) does not support Pascal GPUs properly and ignores them entirely

• The plugin correctly identifies the recommended driver (580.x), but does not preserve or enforce it across upgrades

Request / Suggestion:

Would it be possible for the plugin to honor the currently installed/working driver branch across upgrades, especially when:

1. A GPU is already configured and working

2. The plugin can detect a recommended compatible version (e.g., 580.x for Pascal)

3. A newer branch would result in loss of GPU functionality

At minimum, one of the following would help:

• Preserve the existing driver version on upgrade instead of defaulting to latest

• Automatically constrain upgrades to the compatible branch (e.g., stay within 580.x for Pascal GPUs)

• Provide a warning before switching to an incompatible branch

• Optionally block auto-selection of drivers known to ignore the detected GPU

Right now, the upgrade path can silently move a working system into a non-functional state for older but still widely used GPUs. At least identify the issue and prevent the upgrade or offer a notification of some sort...

Thanks for the work on the recent per-GPU detection enhancements - those made diagnosis much easier. It would be great to extend that logic into upgrade behavior as well.

1 hour ago, Nogami said:

Just tried 7.3 b2 and don't see a way of selecting v580.142 for my Quadro P600 card?

What driver is installed, please always post your Diagnostics so that I can see what's going on.

1 hour ago, jeffreywhunter said:

Root cause (as understood):

• Plugin upgrade or Unraid upgrade defaulted driver selection to latest/Production branch

• That branch (595.x) does not support Pascal GPUs properly and ignores them entirely

• The plugin correctly identifies the recommended driver (580.x), but does not preserve or enforce it across upgrades

No, that's not how it works.

The plugin has a fallback in it's routine that if a 580.x driver is selected that it checks if a 580.x driver is available and downloads that driver.

1 hour ago, jeffreywhunter said:

Right now, the upgrade path can silently move a working system into a non-functional state for older but still widely used GPUs. At least identify the issue and prevent the upgrade or offer a notification of some sort...

That should not happen, however what did your notifications say before you rebooted? Did you see the notifications, if yes which driver was downloaded? The notifications look something like this but in the case for the Nvidia driver it will actually show the version.

1 hour ago, jeffreywhunter said:

Optionally block auto-selection of drivers known to ignore the detected GPU

That should be in there, however if the download was interrupted while upgrading the driver download to the latest version is always in place and will do it's job, do you have Diagnostics from where you upgraded? Before upgrade and after?