New NVIDIA Pascal GPUs Accelerate Deep Learning Inference

BEIJING, CHINA -- (Marketwired) -- Sep 12, 2016 -- GPU Technology Conference China - NVIDIA (NASDAQ: NVDA) today unveiled the latest additions to its Pascal™ architecture-based deep learning platform, with new NVIDIA® Tesla® P4 and P40 GPU accelerators and new software that deliver massive leaps in efficiency and speed to accelerate inferencing production workloads for artificial intelligence services.

Modern AI services such as voice-activated assistance, email spam filters, and movie and product recommendation engines are rapidly growing in complexity, requiring up to 10x more compute compared to neural networks from a year ago. Current CPU-based technology isn't capable of delivering real-time responsiveness required for modern AI services, leading to a poor user experience.

The Tesla P4 and P40 are specifically designed for inferencing, which uses trained deep neural networks to recognize speech, images or text in response to queries from users and devices. Based on the Pascal architecture, these GPUs feature specialized inference instructions based on 8-bit (INT8) operations, delivering 45x faster response than CPUs(1) and a 4x improvement over GPU solutions launched less than a year ago.(2)

The Tesla P4 delivers the highest energy efficiency for data centers. It fits in any server with its small form-factor and low-power design, which starts at 50 watts, helping make it 40x more energy efficient than CPUs for inferencing in production workloads.(3) A single server with a single Tesla P4 replaces 13 CPU-only servers for video inferencing workloads,(4) delivering over 8x savings in total cost of ownership, including server and power costs.

The Tesla P40 delivers maximum throughput for deep learning workloads. With 47 tera-operations per second (TOPS) of inference performance with INT8 instructions, a server with eight Tesla P40 accelerators can replace the performance of more than 140 CPU servers.(5) At approximately $5,000 per CPU server, this results in savings of more than $650,000 in server acquisition cost.

"With the Tesla P100 and now Tesla P4 and P40, NVIDIA offers the only end-to-end deep learning platform for the data center, unlocking the enormous power of AI for a broad range of industries," said Ian Buck, general manager of accelerated computing at NVIDIA. "They slash training time from days to hours. They enable insight to be extracted instantly. And they produce real-time responses for consumers from AI-powered services."

Software Tools for Faster Inferencing
Complementing the Tesla P4 and P40 are two software innovations to accelerate AI inferencing: NVIDIA TensorRT and the NVIDIA DeepStream SDK.

TensorRT is a library created for optimizing deep learning models for production deployment that delivers instant responsiveness for the most complex networks. It maximizes throughput and efficiency of deep learning applications by taking trained neural nets -- defined with 32-bit or 16-bit operations -- and optimizing them for reduced precision INT8 operations.

NVIDIA DeepStream SDK taps into the power of a Pascal server to simultaneously decode and analyze up to 93 HD video streams in real time compared with seven streams with dual CPUs.(6) This addresses one of the grand challenges of AI: understanding video content at-scale for applications such as self-driving cars, interactive robots, filtering and ad placement. Integrating deep learning into video applications allows companies to offer smart, innovative video services that were previously impossible to deliver.

Leap Forward for Customers
NVIDIA customers are delivering increasingly more innovative AI services that require the highest compute performance.

"Delivering simple and responsive experiences to each of our users is very important to us," said Greg Diamos, senior researcher at Baidu. "We have deployed NVIDIA GPUs in production to provide AI-powered services such as our Deep Speech 2 system and the use of GPUs enables a level of responsiveness that would not be possible on un-accelerated servers. Pascal with its INT8 capabilities will provide an even bigger leap forward and we look forward to delivering even better experiences to our users."

Specifications of the Tesla P4 and P40 GPUs include:

Specification                             Tesla P4              Tesla P40   
Single Precision FLOPS*                   5.5                   12          
INT8 TOPS* (Tera-Operations Per Second)   22                    47          
CUDA Cores                                2,560                 3,840       
GPU GDDR5 Memory                          8GB                   24GB        
Memory Bandwidth                          192GB/s               346GB/s     
Power                                     50 Watt (or higher)   250 Watt    

* With boost clock on

The NVIDIA Tesla P4 and P40 are planned to be available in November and October, respectively, in qualified servers offered by ODM, OEM and channel partners.

Supporting Resources

Keep Current on NVIDIA
Subscribe to the NVIDIA blog, follow us on Facebook, Google+, Twitter, LinkedIn and Instagram, and view NVIDIA videos on YouTube and images on Flickr.

NVIDIA (NASDAQ: NVDA) is a computer technology company that has pioneered GPU-accelerated computing. It targets the world's most demanding users -- gamers, designers and scientists -- with products, services and software that power amazing experiences in virtual reality, artificial intelligence, professional visualization and autonomous cars. More information at

(1) Comparing latency using VGG-19 neural network, batch size=4. CPU: Xeon E5-2690v4 using Intel MKL 2017. GPU: Tesla P40 using TensorRT internal version. Intel optimized VGG-19 from
(2) Comparing img/sec using Caffe GoogLeNet neural network, batch size = 128. GPU server with 8x P40 compared to GPU server with 8x M40. Both using TensorRT internal version.
(3) Comparting img/sec/watt using Caffe AlexNet neural network, batch size = 128. CPU: E5-2690v4 using Intel MKL 2017. Using Intel-optimized Caffe and AlexNet from GPU: Tesla P4 measuring GPU power.
(4) Using Intel optimized GoogLeNet, dual-socket CPU server, Xeon E5-2650v4 using Intel MKL 2017. GPU server with 1x Tesla P4 using DeepStream SDK. Video streaming at 720p @ 30FPS.

1 | 2  Next Page »

Review Article Be the first to review this article


Featured Video
Peggy AycinenaWhat Would Joe Do?
by Peggy Aycinena
Acquiring Mentor: Four Good Ideas, One Great
More Editorial  
Sr. staff ASIC Design Engineer -2433 for Microchip at San Jose, CA
Manager, Field Applications Engineering for Real Intent at Sunnyvale, CA
SENIOR ASIC Design Engineer for TiBit Communications at Petaluma, CA
Upcoming Events
DeviceWerx - 2016 at Green Valley Ranch Casino & Resort Las Vegas NV - Nov 3 - 4, 2016
2016 International Conference On Computer Aided Design at Doubletree Hotel Austin TX - Nov 7 - 10, 2016
ICCAD 2016, Nov 7-10, 2016 at Doubletree Hotel in Austin, TX at Doubletree Hotel Austin TX - Nov 7 - 10, 2016
Electric&Hybrid Aerospace Technology Symposium 2016 at Conference Centre East. Koelnmesse (East Entrance) Messeplatz 1 Cologne Germany - Nov 9 - 10, 2016
S2C: FPGA Base prototyping- Download white paper

Internet Business Systems © 2016 Internet Business Systems, Inc.
595 Millich Dr., Suite 216, Campbell, CA 95008
+1 (408)-337-6870 — Contact Us, or visit our other sites:
AECCafe - Architectural Design and Engineering TechJobsCafe - Technical Jobs and Resumes GISCafe - Geographical Information Services  MCADCafe - Mechanical Design and Engineering ShareCG - Share Computer Graphic (CG) Animation, 3D Art and 3D Models
  Privacy Policy