NVIDIA Launches Tesla M10 Quad Maxwell Graphics Card – 32 GB Memory, 2560 Cores Aimed at Virtual Computing

May 21, 2016

NVIDIA has launched their latest Tesla M10 graphics card which features four Maxwell based GPUs on a single PCB. The Tesla M10 graphics card is aimed at Virtual PCs and Virtual application workloads with a design that’s user optimized meaning that it can support several virtual users on a single graphics card. Remote enterprise market will benefit the most from this graphics card.

NVIDIA Tesla M10 With Quad Maxwell GPUs Launched – Could Support Up To 64 Virtual Users On A Single Graphics Card

Unlike the Tesla M60 which is the performance optimized solution or the Tesla M6 which is the blade optimized solution, the Tesla M10 will be NVDIA’s user density optimized solution which aims to provide the maximum no of virtual users on a single graphics card. NVIDIA is also partnering with several virtualization companies to offer to offer remote desktop sessions to users at $2 per month and virtual PCs at $6 per month.

“We have had good success with our GPUs in high-performance computing, deep learning in hpc, deep learning, data analytics, remote work stations,” said McHugh, a former Cisco executive who joined Nvidia six months ago. “Now we have VDI and app virtualization. We are changing the economics for graphics-accelerated virtualization.” via VentureBeat

“While the need for advanced GPU technology has commonly been associated with the usage of 3D applications, as enterprises make the move to software like Windows 10, Office 365, and other SaaS and web apps, IT departments will increasingly seek the benefits of GPU acceleration to provide everyday business tools to all of their users, ”said Robert Young, analyst for IT Service Management and Client Virtualization Software at IDC, in a statement. via VentureBeat

NVIDIA Tesla M10 Graphics Card Specifications In Detail

The NVIDIA Tesla M10 graphics card features a large PCB that deploys four Maxwell based GPUs. All four GPUs are based on NVIDIA’s efficient Maxwell architecture. The specific SKU used is part of GM107 family which was NVIDIA’s most efficient GPU of 2014. Each GPU features 640 CUDA cores and 8 GB of GDDR5 GPU. The graphics card features 2560 CUDA cores and 32 GB of GDDR5 memory on a single board.

NVIDIA Quad GPU solution For Tesla Cards. (Image Credits: IYD.KR)

The entire GPU is passive cooled with a total of 225W TDP. This is provided through a single 8-Pin connector and is designed for dense rack servers. The GPU can support up to 64 Virtual GPU instances and 28 H.264 1080p 30 FPS streams. Each user can access 512 MB of VRAM with the entire GPU filled with vGPU instances. A Dual Tesla M10 configuration can support 124 vGPU instances.

There is no word on the price of this graphics card but it is expected to cost a few thousand bucks. You can learn more about the other two Tesla cards, the Tesla M60 and Tesla M6 over here.

NVIDIA Tesla Maxwell Lineup:

Grid 2.0 Board Name NVIDIA Tesla M60 NVIDIA Tesla M40 NVIDIA Tesla M10 NVIDIA Tesla M6 NVIDIA Tesla M4
GPU GM204 GM200 GM107 GM204 GM206
GPU Cores 2048 x 2 (Dual Config)
4096 CUDA Cores
3072 CUDA Cores 2560 CUDA Cores 1536 CUDA Cores 1024 CUDA Cores
Memory 16 GB GDDR5 (8 GB x 2) 12 GB GDDR5 32 GB GDDR5 8 GB GDDR5 4 GB GDDR5
Memory Bus 256-bit x 2 384-bit 128-bit x 4 256-bit 128-bit
Max Users 36 Deep Learning Focused 64 18 Deep Learning Focused
H.264 (1080P @ 30 FPS) Streams 2-32 Deep Learning Focused 28 1-16 Deep Learning Focused
Form Factor Dual-Slot PCI-Express Dual Slot PCI-Express (Passive Cooling) Dual Slot PCI-Express (Passive Cooling) MXM Card Single Slot PCI-Express (Low Profile Passive Cooling)
TDP 300W 250W 225W 100W 50-75W