Question

A hospital's data science team runs two workloads on shared infrastructure. Workload 1 trains a convolutional neural network on millions of MRI images, performing enormous numbers of identical matrix operations across the dataset. Workload 2 is a legacy scheduling script with heavily branching, sequential logic and frequent unpredictable decision points. A consultant recommends moving one workload to a GPU-accelerated node. Which workload benefits most from the GPU, and why?

Accepted Answer

Workload 1, because GPUs have thousands of cores optimized for executing the same operation across large amounts of data in parallel

Answer

Workload 2, because GPUs excel at branching, sequential logic with unpredictable control flow

Answer

Workload 1, because GPUs have a small number of very high-clock-speed cores ideal for serial tasks

Answer

Workload 2, because GPUs provide larger caches and faster single-thread performance than CPUs