In virtualizing AI infrastructure, which two considerations are crucial for supporting GPU-accelerated workloads? (Select two)

Utilizing NVIDIA vGPU technology for partitioning GPUs is a crucial consideration in virtualizing AI infrastructure to support GPU-accelerated workloads. This technology allows multiple virtual machines to share a single GPU while ensuring that each one has access to dedicated GPU resources. By partitioning the GPU, it improves workload performance, as it enables better resource utilization, scalability, and efficiency for demanding AI tasks. This is particularly important for applications that require high computational power, which is typically provided by GPUs.

Additionally, ensuring GPU pass-through capability is enabled in the hypervisor is another essential consideration. GPU pass-through allows a virtual machine to take direct control of a physical GPU, thereby avoiding the overhead associated with shared GPU resources. This capability is vital for achieving optimal performance in AI workloads, which often rely on heavy parallel processing capabilities that only dedicated access to a GPU can provide.

The other options relate to CPU allocation and resource limits that do not directly impact the GPU's ability to perform effectively in a virtualized environment. While adequate CPU resources can support GPU workloads, they are secondary to the actual GPU resource management strategies such as vGPU technology and pass-through capabilities for maximizing GPU performance.