Posted by: Andy Trevor
My last post was based around a rant I had at a recent GRID partner event held in the UK, it became obvious to me that NVIDIA were missing a trick hence the last piece here
It looks like I was not the only one to have a rant about it as NVIDIA to their credit have reacted pretty quickly to address the core issues I and the others appear to have raised. It is always good to see the vendors listening to us at the coal face and for this I tip my hat to NVIDIA.
So what have they done? Well with the Kepler based cards nearing EOL we have the new M series that was the M6 and M60, they have just announced the M10 board that will be available a little later in the year. What does this bring to the party? Well it fits in as a replacement for the K1 card to provide lower end vGPU goodness for lightweight users. This is a good thing as we now have the entry point back. The board specs are impressive but I will not bore you with all the numbers (these will be available on the NVIDIA site) but I will bring 2 important numbers to the fore. Cost is estimated to be around the current K1 cost which is good, even better is the density (which is what was needed). The boards on the lowest profile will be able to run a maximum of 64 vGPU users. This is the key. Greater density at the low end will lead to a greater adoption of vGPU supported desktops. I still have a small issue with the fact there are associated license costs to deal with over and above buying the boards, but I suspect this is not going to change unless there is a massive customer kickback.
I am still doing cost modelling around K1 and no license v M10 and licenses over a 5 year lifetime and due to the density it is looking around cost neutral so I may have to wind my neck in a little on this (especially as the performance hike looks substantial)
All in all, I am pretty pleased in the response from NVIDIA and look forward to continuing to work with the team in all the projects we have on the go.