NVIDIA A16 : Powering Hybrid Workspaces

Lee Bushen from NVIDIA explains how the new NVIDIA A16 GPU, combined with NVIDIA Virtual PC (vPC) software, enables knowledge workers to tackle any project from anywhere. Purpose-built for high-density, graphics-rich virtual desktop infrastructure (VDI) and leveraging the NVIDIA Ampere architecture, A16 provides double the user density versus the previous generation, while ensuring the best possible user experience.
Further info: www.nvidia.com/en-gb/data-cen...
Virtual GPU Resources: www.nvidia.com/en-gb/data-cen...
Virtual GPU Docs: docs.nvidia.com/grid/
Certified Hardware: www.nvidia.com/en-us/data-cen...
►TIMECODES
0:00 Intro to A16
1:49 Multi-GPU Advantages
4:20 GPU Memory
5:58 Encoders/Decoders
7:50 Tensor/RT Cores
8:17 Performance
9:08 Availability
10:14 Requirements
11:08 Summary

[@diegonayalazo]

Thanks for sharing

[@suthakarthiroshan921]

WoW

[@isamaro_j]

How is this card viewed by the operating system, does it appear like 4 separate GPUs? In regards to the encoders/decoders, is the work load automatically distributed evenly between the GPUs?

[@Aviduduskar]

How many concurrent 1080p30fps streams can the card do with ffmpeg? Would love to see some Tensor / RT benchmarks.

[@lastdoggo251]

yes yes

[@1sh0rt]

If you need GPUs now, how long will the M10 still be produced and supported by citrix, vmware, etc.? Should i rather go with T4s for support and availability reasons, even though price to density is lower?

[@NVIDIADeveloper]

Hi There. Sorry for not getting back to you sooner! Due to the global silicone shortage, we've rationalised our datacenter GPU recommendations for Graphics down to the A40 (high/medium graphical workloads) and A16 (low graphical and office worker workloads). T4 and M10 are still supported our end, but will be more difficult to find. Both A16 and A40 together should satisfy most use cases. Another GPU we're going to be supporting with vGPU in 2022 is the A2. Think of the A2 a little like "a quarter" of an A16. Great for low power and space constrained uses, but the A16 with its 4 x GPUs will win out on density. Hope that helps! Lee.

[@phanigurram]

if any graphic intensive program needs 62 gb ram, do it provide 62 gb ram to it ?

[@sakibkhondaker]

That's a cute little GPU.

[@davidajzhang]

What's the hashrate?

[@NVIDIADeveloper]

Sorry, I don't have those figures, but each GPU individually is much lower powered than something like an A40, A30 or A100. So not optimal for mining! Lee.

[@hundlos1]

Rough price would be nice