This section summarizes the SPECviewperf 13 benchmark scores obtained from the three test cases that we ran. A higher SPEC score indicates greater speed and better performance for the simulated graphics application. We performed the first two test cases by running four concurrent virtual workstations on the MX740c compute host and the third test case by running a single workstation on the MX740c compute host. In the first test case, we connected to one of the virtual workstations through a thin client and collected additional endpoint experience metrics with the NVIDIA nVector Lite tool. The three test cases that we performed were:

• Test Case 1: NVIDIA nVector SPECviewperf 13 + NVIDIA nVector Lite, Multi VM test results
• Test Case 2: NVIDIA nVector SPECviewperf 13 Multi VM test results
• Test Case 3: NVIDIA nVector SPECviewperf 13 Single VM test results

Test Case 1: NVIDIA nVector SPECviewperf 13 + nVector Lite, Multi VMs test results

For this test case, the MX740c GPU-enabled compute host was populated with four vGPU-enabled virtual workstation using the NVIDIA T4-8Q vGPU profile. Three of the endpoints used in this test were VMs, while the fourth endpoint was a Wyse 5070 Thin Client. We used the Wyse Thin Client to collect nVector Lite endpoint metrics, including image quality, frame rate, and end-user latency. Because the nVector Lite tool cannot collect endpoint metrics from virtual machine endpoints, we used the thin client option. You can compare SPEC scores from our performance testing with other published scores on the SPEC website. SPEC scores from our tests indicate an excellent graphics performance by the virtual workstations. The following table summarizes the SPEC scores for this test case (larger scores indicate greater speed):

Table 5.      SPEC score summary for Test Case 1

Test Case 2: NVIDIA nVector SPECviewperf 13, Multi VMs test results

For this test case, we populated the MX740c GPU-enabled compute host with four vGPU-enabled virtual workstations, each with an NVIDIA T4-8Q vGPU profile. The following table summarizes the SPEC scores. For host utilization metric graphs for this test case, see Appendix C: NVIDIA nVector, SPECviewperf 13, Multi VMs test.

Table 6.      SPEC score summary for Test Case 2

Test Case 3: NVIDIA nVector SPECviewperf 13, Single VM test results

For this test case, we populated the MX740c GPU-enabled compute host with only one vGPU-enabled virtual workstation that had an NVIDIA T4-8Q vGPU profile. The following table summarizes the SPEC scores. For host utilization metric graphs for this test case, see Appendix D: NVIDIA nVector, SPECviewperf 13, Single VM test.

The single virtual workstation was able to consume more than its default allocated resources due to the “Best Effort” vGPU scheduler being used. Because there was only one VM per physical GPU, there was no contention for GPU resources. For additional information, see NVIDIA's virtual GPU software documentation.

Also, because there was no CPU oversubscription, the single virtual workstation got more CPU time slices. Because no concurrent VMs were running on the host, the single workstation was able to consume allocated resources in a way that delivered a better performance than VMs running concurrently on the host. As a result, you can see a relatively higher SPEC score in this test compared to the multi VM tests.

Table 7.      SPEC score summary Test Case 3

Note: You can find the results and raw data for the SPECperfview 13 benchmark testing here: https://dell.box.com