Closed davidmi closed 3 years ago
Ricks-Lab
commented
3 years ago Can you provide the output of gpu-ls? It provides total amount of GTT memory and VRAM. Are either of those what you are looking for?
I have never tried a Renoir GPU. Can you run gpu-ls --debug and post the log file? That would help me understand if there any additional sensors that I have not considered.
Drizzt321
commented
3 years ago My 4800H Renoir, in KDE Plasma (via their Neon Ubuntu 20.04 based distro). The debug log file is at Gist https://gist.github.com/Drizzt321/8abaf6bbbc39d1b1031fedd7f6b1dc2d Happy to run any other debug/dump if you need more info.
# gpu-mon
┌─────────────┬────────────────┐
│Card # │card0 │
├─────────────┼────────────────┤
│Model │Renoir │
│GPU Load % │0 │
│Mem Load % │None │
│VRAM Usage % │81.297 │
│GTT Usage % │7.462 │
│Power (W) │None │
│Power Cap (W)│None │
│Energy (kWh) │0.0 │
│T (C) │40.0 │
│VddGFX (mV) │nan │
│Fan Spd (%) │None │
│Sclk (MHz) │0 │
│Sclk Pstate │1 │
│Mclk (MHz) │1600Mhz │
│Mclk Pstate │0 │
│Perf Mode │ │
└─────────────┴────────────────┘
# gpu-ls --debug
Neon: Unverified
Package addon [clinfo] executable not found. Use sudo apt-get install clinfo to install
OS Command [clinfo] not found. Use sudo apt-get install clinfo to install
Detected GPUs: AMD: 1
Can not access package read utility to verify AMD driver.
AMD: Wattman features not enabled: 0xffffbfff, See README file.
Warning: Can not read parameter: mem_loading, disabling for this GPU: 0
Warning: Can not read parameter: power_cap_range, disabling for this GPU: 0
Warning: Can not read parameter: power, disabling for this GPU: 0
Warning: Can not read parameter: power_cap, disabling for this GPU: 0
Warning: Can not read parameter: voltages, disabling for this GPU: 0
1 total GPUs, 0 rw, 1 r-only, 0 w-only
Card Number: 0
Vendor: AMD
Readable: True
Writable: False
Compute: True
GPU UID: None
Device ID: {'device': '0x1636', 'subsystem_device': '0x109f', 'subsystem_vendor': '0x1d05', 'vendor': '0x1002'}
Decoded Device ID: Renoir
Card Model: Advanced Micro Devices, Inc. [AMD/ATI] Renoir (rev c6)
Display Card Model: Renoir
PCIe ID: 04:00.0
Link Speed: 16.0 GT/s PCIe
Link Width: 16
##################################################
Driver: amdgpu
vBIOS Version: 113-RENOIR-026
Compute Platform: None
GPU Type: APU
HWmon: /sys/class/drm/card0/device/hwmon/hwmon4
Card Path: /sys/class/drm/card0/device
System Card Path: /sys/devices/pci0000:00/0000:00:08.1/0000:04:00.0
##################################################
Current Power (W): None
Power Cap (W): None
##################################################
Current GPU Loading (%): 0
Current Memory Loading (%): None
Current GTT Memory Usage (%): 7.002
Current GTT Memory Used (GB): 0.210
Total GTT Memory (GB): 3.000
Current VRAM Usage (%): 81.073
Current VRAM Used (GB): 0.405
Total VRAM (GB): 0.500
Current Temps (C): {'edge': 43.0}
Critical Temps (C): {'edge': 0.0}
Current Voltages (V): None
Current Clk Frequencies (MHz): {'sclk': 400.0}
Current SCLK P-State: [1, '400Mhz']
Current MCLK P-State: [0, '1600Mhz']
Power Profile Mode: None
Power DPM Force Performance Level: auto
# uname -a
Linux darklaptop 5.8.0-45-generic #51~20.04.1-Ubuntu SMP Tue Feb 23 13:46:31 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
# cat /proc/cmdline
BOOT_IMAGE=/BOOT/ubuntu_1lc8c8@/vmlinuz-5.8.0-45-generic root=ZFS=rpool/ROOT/ubuntu_1lc8c8 ro
davidmi
commented
3 years ago Thank you for suggesting gpu-ls, here are the results. This is a Ryzen 4500u.
$ sudo gpu-ls
Detected GPUs: AMD: 1
amdgpu/rocm version: UNKNOWN
AMD: Wattman features not enabled: 0xffffbfff, See README file.
Warning: Error reading parameter: mem_loading, disabling for this GPU: 0
Warning: Error reading parameter: power_cap_range, disabling for this GPU: 0
Warning: Error reading parameter: power, disabling for this GPU: 0
Warning: Error reading parameter: power_cap, disabling for this GPU: 0
Warning: Error reading parameter: voltages, disabling for this GPU: 0
1 total GPUs, 0 rw, 1 r-only, 0 w-only
Card Number: 0
Vendor: AMD
Readable: True
Writable: False
Compute: False
GPU UID: None
Device ID: {'device': '0x1636', 'subsystem_device': '0x0a1e', 'subsystem_vendor': '0x1028', 'vendor': '0x1002'}
Decoded Device ID: Renoir
Card Model: Advanced Micro Devices, Inc. [AMD/ATI] Renoir (rev c3)
Display Card Model: Renoir
PCIe ID: 03:00.0
Link Speed: 16.0 GT/s PCIe
Link Width: 16
##################################################
Driver: amdgpu
vBIOS Version: 113-RENOIR-025
Compute Platform: None
GPU Type: APU
HWmon: /sys/class/drm/card0/device/hwmon/hwmon4
Card Path: /sys/class/drm/card0/device
System Card Path: /sys/devices/pci0000:00/0000:00:08.1/0000:03:00.0
##################################################
Current Power (W): None
Power Cap (W): None
##################################################
Current GPU Loading (%): 0
Current Memory Loading (%): None
Current GTT Memory Usage (%): 1.273
Current GTT Memory Used (GB): 0.089
Total GTT Memory (GB): 7.000
Current VRAM Usage (%): 30.291
Current VRAM Used (GB): 0.151
Total VRAM (GB): 0.500
Current Temps (C): {'edge': 48.0}
Critical Temps (C): {'edge': 0.0}
Current Voltages (V): None
Current Clk Frequencies (MHz): {'sclk': 400.0}
Current SCLK P-State: [1, '400Mhz']
Current MCLK P-State: [3, '400Mhz']
Power Profile Mode: None
Power DPM Force Performance Level: auto
Looks like what I was looking for is the GTT memory usage! It would be helpful to have that in gpu-mon, I think, since that's the "real" total VRAM.
I've manually overridden that in modprobe now using gttsize since the amdgpu driver calculates it incorrectly. I found the problem in another way, but it would be useful if gpu-mon made it visible. I was investigating why I seemed to have less VRAM on Linux than Windows, and it looks like that was the problem. I overrode the gttsize parameter and it seems to have worked https://www.kernel.org/doc/html/v4.19/gpu/amdgpu.html
davidmi
commented
3 years ago Looking again, I see that gpu-mon does report GTT! So it's my fault for missing it/not understanding what it meant, apologies. Hopefully this thread can help anyone who is similarly confused.
Ricks-Lab
commented
3 years ago Cool! I will go ahead and close the issue. You can reopen if needed.
Hi, thank you so much for writing gpu-utils!This is the only monitoring utility I could find that worked for my Ryzen iGPU on Linux.
I'm sorry if this is not the right place to ask a question, but I thought asking it here might help others if they encounter the same thing.
I'm running a Ryzen 4000 series dual-booting Ubuntu laptop, with an integrated Vega GPU. The iGPU has 512 MB dedicated vram, but it also uses up to 4 GB of shared system memory. On Windows 10, I can see the usage of the shared memory in Task Manager, but in gpu-mon it only reports the usage of the dedicated vram. I am running with compositing disabled on KDE, so it's using very little VRAM -- when I run on GNOME 2 with compositing enabled, it is always at 97% or so -- the excess is not reported.
Is there any way to see the amount of shared video memory being used?
Once again, thank you for this really useful set of tools!