lmstudio-ai
Failed to Load LLM on Old PC with blank system resources
It get Failed to load the model Attempt to pull a snapshot of system resources failed. Error: 'Cannot read properties of undefined (reading 'pullReport')'
As, I guess, there are no system resources listed on the System Reroutes page
perhaps because my PC (Win10, 64bit 16GB plus 1GB ancient GPU) is too old Specs below
But then I hear LMStudio works on laptops. Perhaps my PC is even slower.
Speccy says Operating System Windows 10 Home 64-bit CPU AMD Phenom II X6 1060T Thuban 45nm Technology RAM 16.0GB Dual-Channel DDR3 @ 802MHz (11-11-11-28) Motherboard ASUSTeK Computer INC. M4A89GTD-PRO/USB3 (AM3) 30 °C Graphics HP P221 (1080x1920@60Hz) L1942 (1280x1024@60Hz) 1023MB NVIDIA GeForce GTS 450 (MSI) 46 °C Storage 465GB Western Digital WDC WDS500G2B0A ATA Device (SATA (SSD)) 33 °C Optical Drives ELBY CLONEDRIVE SCSI CdRom Device Audio Realtek High Definition Audio
Someone on Youtube kindly informed me that
"The GPU architecture is far too old. I think the oldest Llama.cpp gpu arch supported is Ada. That is way way too old. However, you can do cpu, which will work but very very very slow and probably awful. But it will generate tokens"
@timtak the AMD Phenom II X6 1060T CPU is quite old and lacks AVX2 instructions which are a hard requirement for running models in LM Studio:
https://lmstudio.ai/docs/system-requirements
Thank you very much.
I skimmed over the AVX2, probably because I had no idea what it meant. I have a two year old PC at work, I will try that.
Tim
@timtak the
AMD Phenom II X6 1060TCPU is quite old and lacks AVX2 instructions which are a hard requirement for running models in LM Studio:https://lmstudio.ai/docs/system-requirements
Is there a reason why LM Studio set AVX2 instead of AVX as minimum requirement? This is opting out some servers with dual Intel Xenon CPU's that only support AVX and could be used for lower AI powered projects... Jan.ai project (https://github.com/janhq/jan) sets AVX as minimum requirement and works very well, so LM Studio should perform similar (there was already a beta version with AVX as minimum requirement, why is it gone?)...
I was getting Attempt to pull a snapshot of system resources failed. Error: 'Cannot read properties of undefined (reading 'pullReport')' on my M1 Max (64 GB) with every model I tried (all MLX).
A restart fixed it for me — I think a suspended user session was tying up resources.
