Hi,
I'm running a test of Horizon View (5.2 build-987719) on a Dell R720 with an Nvidia GRID K2 card. The Nvidia driver is version 304.76.
I've set up a pool on this node to test out the hardware accelerated graphics, but am encountering an odd issue. I'm toying with the molecular modeling package Schrodinger Maestro, loading and rotating 3d protein models. In a fresh session, the fps is initially very poor, to the point that the software would not be usable by our users (< 10fps). On the esxi host, 'nvidia-smi' shows low to no GPU utilization, in the 0-5% range. Using the PCoIP log viewer, I see low-ish PCoIP server utilization (<20%) and relatively low Image Tx bandwidth utilization (~5000 Kb/sec).
At some point, and what triggers this I'm not entirely sure (it has happened a couple of times after I opened the task manager, though this seems an odd cause. It has happened on its own after 30 seconds or so), the GPU kicks in; 'nvidia-smi' shows 25-30% utilization, PCoIP server utilization jumps to ~90%, and Image Tx bandwidth bumps to ~20000Kb/sec. The protein rotation smooths out, and everything appears to work as it should from that point on.
The VMs run with 6 cpus and 32 GB ram. I've run through the PCoIP optimization guide (and disabled 'build to lossless' as well as tuned audio bandwidth), and also ran through the Windows 7 View optimization guide (and made use of the recommended optimization script). I'm seeing this issue with only one View desktop running on the host.
Any ideas?
Thanks,
Eric