Unverified50% confidenceFactExact time
MPS's driver_allocated_memory count and the process's phys_footprint are two independent ledgers, and both dropping in sync does not mean the memory has been released.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSystem memory and VRAM are two different memory pools and cannot simply be added together to equal total available VRAM.66% similarUnverifiedThe fact that 96GB of VRAM can fit two models simultaneously does not mean two tasks can run at full speed simultaneously, because GPU compute resources are shared and competed over.63% similarUnverifiedOn an Apple Silicon Mac, after moving a PyTorch tensor on the MPS device to the CPU via .to("cpu"), the process's phys_footprint immediately drops to near zero, but the tensor data remains fully in memory.60% similarUnverifiedOn a 1 vCPU / 2 GB instance, not explicitly calling gc.collect() may cause Python's cyclic-reference garbage collection to lag, with peak memory exceeding 2 GB and being terminated by the OOM Killer.59% similarUnverifiedLibTorch内部使用引用计数的智能指针（intrusive_ptr）管理张量的存储生命周期，在张量不再被引用时自动释放GPU显存58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1015595API
curl https://kongchang.com/api/v1/knowledge/claims/1015595MCP
get_claim(id=1015595)