Presentation Summary
I recently (April 23rd) presented “Considerations for the deployment of GPUs in virtual desktop environments” at the Tennessee Higher Education Information Technology Symposium (THEITS). In addition to being well received material, the topic also prompted some lively and informative discussion among the attendees.
My goals were to cover three main topics within the space: 1) motivations for looking at GPUs in virtual desktop environments, 2) what solutions were available, and 3) evaluation of graphic resource needs and usage. All of that packed into a 45 minute presentation, plus a desire to allow time for discussion and questions, resulted in lots of things getting quick mentions and then moving on to something else. The aim of this blog post is to provide a list of all the resources and tools that were referenced in the talk to allow for more thorough research.
In making the slide deck available for download, I’ve replaced a few of the images that weren’t my making with placeholder links and added more prominent citations in places.
Presentation Resources
Below are links to relevant resources that were one of 1) the source for what was said, 2) directly referenced, or 3) reinforcement of what was covered. Items are grouped by presentation section and listed by order of reference in the presentation.
Motivation
Samsung Galaxy S9
- Qualcomm Snapdragon 845 Mobile Platform use in Samsung Galaxy S9
- and Qualcomm Snapdragon 845 Mobile Platform with Adreno 630 Visual Processing Subsystem (Tech Specs)
Forthcoming Intel NUC Hardware
- 8th Gen Intel Core i7 processor with Radeon RX Vega M Graphics
- and Intel NUC8i7HVK and NUC8i7HNK (NUC models)
Lakeside Software Windows 10 analysis
- What You Need to Know About GPUs for Windows 10 (blog post)
- and Elevating User Experience through GPU Acceleration (white paper)
Example GPU applications
- Tennessee Tech HPC Impulse NAMD results
- Ansys Discovery Live via Citrix XenDesktop video by Miranex on YouTube
- Amitai Rottem tweet concerning Windows Defender’s use of Intel GPU:
Solutions
Architectures for graphics virtualization
Hardware Vendors (listed alphabetically)
Evaluation
NVIDIA-SMI
- Useful nvidia-smi Queries (NVIDIA Knowledgebase)
- NVIDIA Virtual GPU Software User Guide
Experience/Environment Monitoring Tools
- uberAgent(product page), list of metrics
- Lakeside Systrack
- Stratosphere UX
- vRealize Operations for Horizon (product page), NVIDIA GRID monitoring
Other tools
Testing at Scale
Workload Benchmarks
- PCMark
- SolidWorks Performance Test (built-in benchmark)
- Matlab bench command (built-in benchmark)
- Standard Performance Evaluation Corporation (SPEC) Graphics and Workstation Performance Group (GWPG) benchmarks:
- SPECviewperf (version 12 at time of presentation)
- SPECwpc (currently version 2.1)
- Application Performance Characterization (SPECapc) work also has application specific benchmarks.
Unigine benchmarks
- Heaven
- Valley
- Note: Newest benchmark, Superposition, moves command line automation to professional license (vs advanced license for others).
