For the performance testing, we designed ten different test cases to illustrate various combinations of AI and medical VDI workloads using both CPU and GPU resources:
- Four AI test cases, for AI training and AI model validation, both run with CPU only and with GPU enhancement
- Two traditional VDI test cases for a medical knowledge worker profile using a MicroDicom medical image viewer application, with the same test case run with CPU only and with GPU enhancement
- Four “dual workload” test cases that ran concurrent workloads for AI training and validation and VDI together, with and without GPUs in several combinations
The list of test cases is:
- AI test cases
- AI Training GPU Enhanced
- AI Validation GPU Enhanced
- AI Training CPU Only
- AI Validation CPU Only
- VDI test cases
- VDI Medical Knowledge Worker GPU Enhanced
- VDI Medical Knowledge Worker CPU Only
- Dual workload AI and VDI test cases
- Dual Workload AI Training and VDI Medical Knowledge Worker Both GPU Enhanced
- Dual Workload AI Validation and VDI Medical Knowledge Worker Both GPU Enhanced
- Dual Workload AI Training CPU Only and VDI Medical Knowledge Worker GPU Enhanced
- Dual Workload AI Validation CPU Only and VDI Medical Knowledge Worker GPU Enhanced
The following sections describe the test environment and methodology, and present a summary of the performance test results, with an emphasis on the dual workload cases that are the most applicable to the overall premise of GPU use in a healthcare environment.