GPU time-slicing itself adds neither memory nor fault isolation; separate VM/host controls have their own boundaries
0observed
| Value kind | observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range. |
|---|---|
| Scope | NVIDIA comparison concerns GPU sharing mechanisms. Volta-and-later MPS has separate GPU address spaces but shares scheduling and memory-system resources. Assess the surrounding VM, host and management boundary separately. |
| As of | 2026 |
| Source | NVIDIA, Multi-Instance GPU User Guide, Concepts: CUDA Concurrency Mechanisms (rolling documentation, no publication date). · CUDA Concurrency Mechanisms, Table 3: Streams, MPS and MIG; MPS address protection, shared memory-system resources and absence of error isolation. |
| Review | checking…review by 2026-11-23 · standard cadence |
| Recorded changes | last 2026-07-10 · 2 revisions tracked |
| Claim id | memory-and-fault-isolation-guarantees-provided |
Where the guide uses it
Not quoted in a chapter yet; it is kept in the curated register.
← Full numbers register — every date-stamped figure in the guide, with revision history.