I was on a call with an HPE rep discussig the merits of L4 vs L40 and L40s. The engineer said that L40 and L40s were not a good fit for vPC. I had planned on using L40s for the L40s-2B profile and fit up to 24 VDI on each card with a 2GB GPU profile. The engineer said that L4 was better for vPC due to some queueing mechanism but he couldn’t go any deeper than that. Based on everything I see, the L40s is compatible and provides a B series profile, so besides Power, Form Factor and Cost, why should I not use it over the L4. It takes 4 L4s to run same density as 2 L40S and not many chassis support 4 L4 (besides HPE GPU Chassis, but that might not be an option for other reasons).
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Mix L4 and L40 for vGPU POC | 5 | 1089 | October 26, 2023 | |
| Tesla P40 24GB GPU Profile setting | 1 | 7852 | April 4, 2019 | |
| GU H100/L40S Performance | 4 | 230 | November 25, 2024 | |
| Tensor core differences between L40 and L40S? (and RTX 6000 Ada?) | 0 | 9159 | August 14, 2023 | |
| Consumer motherboard options for 2 or 4, P40s? | 1 | 965 | May 22, 2023 | |
| Question Regarding A40 vs A6000 for VMWare vGPU Server | 1 | 5094 | January 14, 2022 | |
| NVIDIA VGPU for L40s baremetal | 0 | 184 | October 28, 2024 | |
| L40 vs. RTX 6000 Ada FP16/FP8 throughput? | 7 | 13074 | April 4, 2023 | |
| L40S VMware Linux Setup and Licensing | 1 | 32 | February 8, 2025 | |
| Multi GPU mobo certified by NVIDIA | 1 | 273 | July 8, 2024 |