NVIDIA Unveils New Hyperscale Accelerators to Boost Machine Learning Throughput for Data Centers

NVIDIA has unveiled new additions to its NVIDIA Tesla platform – an end-to-end hyperscale data center platform that lets web hosting services companies accelerate their huge machine learning workloads and increase the throughput of data centers. The new NVIDIA hyperscale accelerator line consists of two accelerators, Tesla M40 GPU and Tesla M4 GPU.

The Tesla M40 GPU accelerator lets researchers more quickly innovate and design new deep neural networks for each of the increasing number of applications they want to power with artificial intelligence (AI). The Tesla M4 GPU accelerator is a low-power accelerator designed to deploy these networks across the data center. The line also includes a suite of GPU-accelerated libraries.

These new hardware and software products are designed specifically to accelerate the flood of web applications that are racing to incorporate AI capabilities. Machine learning is being used to make voice recognition more accurate. It enables automatic object and scene recognition in video or photos with the ability to tag for later search. It makes possible facial recognition in videos or photos, even when the face is partially obscured. And it powers services that are aware of individual tastes and interests, which can organize schedules, deliver relevant news stories and respond to voice commands accurately and in a conversational tone.

Cloud Services, Automotive, Health Care

nvidia-tesla-accelerators“The artificial intelligence race is on,” said Jen-Hsun Huang, co-founder and CEO of NVIDIA. “Machine learning is unquestionably one of the most important developments in computing today, on the scale of the PC, the Internet and cloud computing. Industries ranging from consumer cloud services, automotive and health care are being revolutionized as we speak. We created the Tesla hyperscale accelerator line to give machine learning a 10X boost. The time and cost savings to data centers will be significant.”

The NVIDIA hyperscale accelerator line was created to accelerate these workloads and dramatically increase the throughput of data centers. These new additions to the NVIDIA Tesla platform include:

  • NVIDIA Tesla M40 GPU – a “very powerful” accelerator designed for training deep neural networks
  • NVIDIA Tesla M4 GPU – a low-power, small form-factor accelerator for machine learning inference, as well as streaming image and video processing
  • NVIDIA Hyperscale Suite – a “rich suite” of software optimized for machine learning and video processing

NVIDIA Tesla M40 GPU Accelerator

The NVIDIA Tesla M40 GPU accelerator would allow data scientists to save days, even weeks, of time while training their deep neural networks against massive amounts of data for higher overall accuracy. Key features include:

  • Optimized for Machine Learning – Reduces training time by 8X compared with CPUs (1.2 days vs. 10 days for a typical AlexNet training).
  • Built for 24/7 reliability – Designed and tested for high reliability in data center environments.
  • Scale-out performance – Support for NVIDIA GPUDirect allowing fast multi-node neural network training.

NVIDIA Tesla M4 GPU Accelerator

nvidia-teslaThe NVIDIA Tesla M4 accelerator is a low-power GPU purpose-built for hyperscale environments and optimized for demanding, high-growth web services applications, including video transcoding, image and video processing, and machine learning inference. Key features include:

  • Higher throughput – Transcodes, enhances and analyzes up to 5X more simultaneous video streams compared with CPUs.
  • Low power consumption – With a user-selectable power profile, the Tesla M4 consumes 50-75 watts of power, and delivers up to 10X better energy efficiency than a CPU for video processing and machine learning algorithms.
  • Small form factor – Low-profile PCIe design fits into enclosures required for hyperscale data center systems.

NVIDIA Hyperscale Suite

The new NVIDIA Hyperscale Suite includes tools for both developers and data center managers, specifically designed for web services deployments, including:

  • cuDNN – Popular algorithm software for processing deep convolutional neural networks used for AI applications.
  • GPU-accelerated FFmpeg multimedia software – Harnesses widely used FFmpeg software to accelerate video transcoding and video processing.
  • NVIDIA GPU REST Engine – Enables the “easy creation and deployment of high-throughput, low-latency” accelerated web services spanning dynamic image resizing, search acceleration, image classification and other tasks.
  • NVIDIA Image Compute Engine – GPU-accelerated service with REST API that would provide image resizing “5 times faster” compared to a CPU.

The NVIDIA Tesla M40 GPU accelerator and Hyperscale Suite software will be available later this year, while the NVIDIA Tesla M4 GPU will be available in the first quarter of 2016.