Last month NVIDIA launched GRID 2.0 architecture supporting virtualized graphics.
So what's new in NVIDIA GRID 2.0?
- They've doubled the user density, now allowing for up to 128 vGPU users per physical server. This comes by increasing the amount of GPU memory per physical GPU to 8GB, meaning they can get more vGPUs per physical GPU.
- They've moved to the Maxwell (from Kepler) architecture. This means the individual GPUs are much more powerful than the previous generation. You can have either more GPU resources per user, or have the GPU resources to support those additional vGPU instances enabled by the extra GPU memory. Either way it's a win.
- New form factors will allow GRID enabled GPUs to be installed in blade servers without requiring a sidecar! This is a major win for customers who have standardized on blade architectures for their datacenter. No longer will they be required to deploy a different set of servers and infrastructure to support graphics needs.
- Last, the new release will support the Linux VDI offerings from VMware and Citrix. So if you need graphics and you need Linux you can finally do both under vGPU!
If you just purchased GRID 1.0, don't panic, it isn't going away - it's still valid and supported. 2.0 simply gives you more options to consider for future purchases.