The following tables summarize the host utilization metrics for the Login Enterprise workload and the AI training and validation, and the user density derived from the testing:
| Workload | Density per host | Average CPU % | Average GPU % | Average memory consumed GB | Average memory active GB | Average IOPS / user | Average net mbps/User | EUX score |
| VMware Horizon Knowledge Worker 14G NO vGPU | 52 | 5% | N/A | 302 GB | 104 GB | 8 | 25 Mbps | 8 |
| VMware Horizon Knowledge Worker 14G WITH GPU | 60 | 76% | 15% | 538 GB | 361 GB | 8 | 44 Mbps | 3 |
| VMware Horizon Knowledge Worker 15G NO vGPU | 120 | 75% | N/A | 535 GB | 214 GB | 8 | 65 Mbps | 1 |
| VMware Horizon Knowledge Worker 15G WITH vGPU | 48 | 34% | 15% | 395 GB | 288 GB | 4 | 48 Mbps | 7 |
| Dual AI Training + VMware Horizon Knowledge Worker 15G WITH vGPU | 24 VDI + 1 AI VM | 27% | AI - 82% VDI -14% | 458 GB | 336 GB | 5 | 5 Mbps | 6 |
| Dual AI Validation + VMware Horizon Knowledge Worker 15G WITH vGPU | 24 VDI + 1 AI VM | 19% | AI - 4.9% VDI -14% | 459 GB | 336 GB | 8 | 6 Mbps | 5 |
| Workload | Density per host | Number of epochs | Batch size | Training/validation time | Model accuracy |
| Dual AI Training + VMware Horizon Knowledge Worker 15G WITH vGPU | 24 VDI + 1 AI VM | 15 | 48 | 65 minutes | N/A |
| Dual AI Validation + VMware Horizon Knowledge Worker 15G WITH vGPU | 24 VDI + 1 AI VM | 15 | N/A | 5 minutes | 69% |
The host utilization metrics mentioned in the table are defined as follows:
- User density—The number of users per compute host that successfully completed the workload test within the acceptable resource limits for the host. For clusters, this number reflects the average of the density achieved for all compute hosts in the cluster.
- Average CPU—The average CPU usage over the steady state period. For clusters, this number represents the combined average CPU usage of all compute hosts. On the latest Intel processors, the ESXi host CPU metrics exceed the rated 100 percent for the host if Turbo Boost is enabled, which is the default setting. An additional 35 percent of CPU is available from the Turbo Boost feature. However, this additional CPU headroom is not reflected in the VMware vSphere metrics where the performance data is gathered.
- Average GPU—The figure shown in the table, 'Avg' GPU %', is the combined average GPU usage of all installed GPUs over the test period.
- Average active memory—For ESXi hosts, this is the amount of memory that is actively used, as estimated by the VMkernel based on recently touched memory pages. For clusters, this is the average amount of physical guest memory that is actively used across all compute hosts over the steady state period.
- Average IOPS per user—IOPS calculated from the average cluster disc IOPS over the steady state period divided by the number of users.
- EUX score—The End User Experience score represents the performance of the VDI virtual machines and ranks it between 0 and 10 as experienced by a user.
- Epoch—An epoch is when all the training data is used at once and is defined as the total number of iterations of all the training data in one cycle for training the machine learning model.
- Batch size—Specifies how many samples must be processed before the internal model parameters are updated.
- Model accuracy—Model accuracy is the fraction of predictions the model got right.
Note: The test scenarios on the 14th generation hardware were run on a three-node cluster with a vSAN datastore while the 15G tests were run on a single host using a local datastore. Latency data was not available for the local datastore on the 15G host, thus it is not included in the results data. It is anticipated that with such a low number of VMs and disc operations during testing, latency was not a major factor in these test scenarios.