More Power at Lower Cost
Dr Matt Wood, GM of Artificial Intelligence at Amazon Web Services (AWS), outlines how AWS allows customers to do more with less, using the Intel® Xeon® Scalable processor.
Amazon and Intel have been collaborating on new technology for over a decade. Together we've been able to pioneer new EC2 instance families, which provide dramatically more bang for buck. More computational power at ...lower cost, which lets customers do more for less.
The new C5 instance family announced in November last year is powered by a custom Intel® Xeon® Scalable processor, with a new microarchitecture, new hardware acceleration capabilities and support for Intel's AVX-512 instruction set.
The Intel® Xeon® Scalable processor combined with the innovations in networking and acceleration in the C5 instance family bring about a major epoch in the world of high performance computing, machine learning, and artificial intelligence. It's never been easier or more cost effective to collect data at any scale, store it in a broad set of databases and storage services and then build custom machine and deep learning models.
Today business deploy more machine learning workloads on AWS than on any other platform. Including Aon, a large insurance and capital advisory company that is running financial simulations using machine learning on AWS, which reduced the time to complete simulations from days to minutes.
And the computational power of the Intel® Xeon® Scalable processor lets customers use more data to create innovative new products and experiences powered by machine learning.
Together with Intel, we've optimized deep learning engines with the latest version of the Intel® Math Kernel Library (Intel® MKL). And the Intel® Xeon® Scalable processor is available on the new C5 instance family to increase inference performance by over 500x.
We've also worked closely with Intel to demonstrate how you can use familiar tools such as Apache Spark* and BigDL to easily build distributed deep learning applications on AWS.
High performance computing allows engineers and scientists to solve complex problems, compute, and data intensive problems. HPC apps often require high network performance, fast storage, large amounts of memory, and very high compute capabilities, making them an ideal workload for our compute-optimized instances, including the new C5. Running these workloads on AWS gives customers instant access to thousands or tens of thousands of EC2 instances running Intel® Xeon® processors, allowing them to increase the speed of research in a cost-effective way that would not be practical in most on-premises HPC environments.
For example, Novartis used 10,600 EC2 instances and approximately 87,000 compute cores to conduct 39 years of computational chemistry in just 9 hours, screening 10 million compounds against a cancer target. This magnitude of improvement is changing what is possible in industries such as healthcare, life sciences, financial services, scientific research, aerospace, automotive, manufacturing, and energy.
We're excited to see our collaboration with Intel continue and for customers to continue to innovate on AWS through the best performance available anywhere.