NVIDIA GRID Hardware Chart
When Ruben Spruijt and I presented our session “Graphics for Virtual Desktops and Apps – 2016 Edition” at BriForum in London last week, one of our PowerPoint slides showed a chart comparing the different GRID cards. The chart includes lots of information, such as number of GPUs per card, GPU type, clock speed, number of GPU cores, performance (TFLOPS), memory size, maximum number of vGPU concurrent users, form factor, maximum power consumption and more. It’s the kind of charts you can talk about for hours – or just use it as an overview without looking at the details. Ruben and I updated the chart last minute after NVIDIA’s official announcement of the Tesla M10 card only 24 hours prior to our session. As a response to a picture of the chart uploaded to Twitter, we learned that some details were not accurate. Today I ran into Milan Diebel and Erik Bohnhorst from the NVIDIA GRID product team and together we corrected the chart. Well, and now I want to share the updated version with you.
Below is a picture of the chart. Please note that we changed the performance row from GFLOPS to TFLOPS, updated the clock speed specs and modified the way to compare the H.264 1080p stream encoders.