MIG gives a job what, compared with time-slicing?
Why D is the answer
Hardware-isolated memory and compute
MIG partitions are isolated in hardware.
In detail
- MIG partitions are isolated in hardware.
- Time-slicing shares a GPU in turns.
Use case
An illustrative scenario, not a real incident.
Two jobs share a GPU, and each has hardware-isolated memory with MIG.
See it in different ways
flowchart LR A["Enable MIG mode"] --> B["Create GPU instances"] B --> C["Create compute instances"] C --> D["Assign to workloads"]
mindmap
root((Sharing))
Run:ai
Project quotas
Over quota, then preempted
Slurm
Fairshare
MIG
GPU instances
Hardware isolationflowchart TD
Q{"Which method?"}
Q -->|"MIG"| A0["Hardware isolation"]
Q -->|"Time-slicing"| A1["Taking turns"]Key point: MIG isolates in hardware.
Why the other options are wrong
- A. Exclusive access to all GPUs
- It divides a GPU.
- B. A higher GPU clock speed
- It does not raise clocks.
- C. A larger host RAM share
- It is not about host memory.