What does an AI-optimised server platform look like
You'll hear a lot about AI and the value it can bring to a business. But when it comes to hardware and software, what does it mean for hardware to be AI-optimized?
AI is the new computing revolution. Analyst IDC predicts [1] that companies worldwide will increase spending on AI by 19.6% in 2022, with the total market growing to more than $500bn next year.
That's because AI is showing its value in a huge range of data analysis roles – and the latest server platforms provide everything businesses need to unlock its full potential.
One major new feature is Intel® Deep Learning Boost for deep learning tasks. This ensures the most efficient use is made of computing resources and caches, while minimising bandwidth bottlenecks. Compared to last-generation processors, it delivers a 1.56x improvement in AI inference for image classification [2].
Since Deep Learning Boost is built into every 3rd Gen Xeon Scalable processor, it can accelerate applications at scales ranging from individual workstations to multi-node supercomputer configurations. 4S and 8S models also support the bfloat16 number format, which can hugely improve training speed without compromising accuracy.
On the software side, the oneAPI AI Analytics Toolkit lets developers use familiar Python tools and frameworks to create AI and analytics pipelines, while the Intel® Distribution of OpenVINO™ Toolkit comes with pre-trained models and enables teams to deploy AI apps across any combination of processors and accelerators.
HPC systems with powerful processors and massively parallel architectures are perfect for such tasks. And Intel's new AI enhancements provide a transformative benefit. For example, CERN researchers have used Intel's 3rd Gen Xeon Scalable processors to move from software simulations of particle collisions to AI-accelerated models, getting results up to 68,000 times faster – with a slight improvement in accuracy [3].
For more demanding roles, the Tyan Thunder SX TS65-B7126 is a 2U dual-socket server that features massive storage capabilities (up to 216TB if equipped with 12x 18TB SATA HDDs) and large amounts of memory. Powered by the newest Intel Xeon Scalable processors, the SX TS65-B7126 is a flexible solution for challenging Big Data workloads and AI-inference applications.
At the top of the range, the 4U Tyan Thunder HX FT83A-B7129 supports not only dual Xeon Scalable processors, but can host up to 10 DW GPU cards (with one spare PCIe x16 expansion slot) for powering massively parallel workloads. It's a true supercomputer in a rack, allowing you to unlock the full power of AI-optimised computing. Its flexible design gives users the option of using GPUs alongside other accelerator hardware if required, while PCIe Gen 4.0 support is a welcome addition.
Built for the most demanding HPC workloads, the Tyan Thunder series is a line of world class barebones server systems based on Intel's latest technology, including 3rd Gen Intel Xeon Scalable Processors. For more information about our newest products, view the Tyan/Intel Product Catalog 2022.
AI is the new computing revolution. Analyst IDC predicts [1] that companies worldwide will increase spending on AI by 19.6% in 2022, with the total market growing to more than $500bn next year.
That's because AI is showing its value in a huge range of data analysis roles – and the latest server platforms provide everything businesses need to unlock its full potential.
The benefits of built-in AI acceleration
3rd Gen Intel® Xeon® Scalable processors are the only CPUs on the market designed with built-in AI acceleration, to enable faster results across a wide range of workloads.One major new feature is Intel® Deep Learning Boost for deep learning tasks. This ensures the most efficient use is made of computing resources and caches, while minimising bandwidth bottlenecks. Compared to last-generation processors, it delivers a 1.56x improvement in AI inference for image classification [2].
Since Deep Learning Boost is built into every 3rd Gen Xeon Scalable processor, it can accelerate applications at scales ranging from individual workstations to multi-node supercomputer configurations. 4S and 8S models also support the bfloat16 number format, which can hugely improve training speed without compromising accuracy.
On the software side, the oneAPI AI Analytics Toolkit lets developers use familiar Python tools and frameworks to create AI and analytics pipelines, while the Intel® Distribution of OpenVINO™ Toolkit comes with pre-trained models and enables teams to deploy AI apps across any combination of processors and accelerators.
How HPC workloads benefit from an AI-optimised architecture
With the rise of technologies such as cloud computing, IoT and 5G, companies across a huge range of industries are capturing more and more data. Given the right analytical capability, this data can provide actionable insights that hold the key to gaining a competitive edge and staying ahead of evolving markets.What does an AI-optimised server look like?
The wide range of 3rd Gen Intel Xeon Scalable processor models means that there are AI-ready options at almost every level and for every budget. At one end of the scale, an AI-optimized solution might look like Tyan's Thunder HX FT65T-B5642. This is a deskside HPC workstation that can be specified with a wide range of processors, ranging from the eight-core Intel Xeon Silver 4309Y processor right up to the 40-core Intel Xeon Platinum 8380 CPU.For more demanding roles, the Tyan Thunder SX TS65-B7126 is a 2U dual-socket server that features massive storage capabilities (up to 216TB if equipped with 12x 18TB SATA HDDs) and large amounts of memory. Powered by the newest Intel Xeon Scalable processors, the SX TS65-B7126 is a flexible solution for challenging Big Data workloads and AI-inference applications.
At the top of the range, the 4U Tyan Thunder HX FT83A-B7129 supports not only dual Xeon Scalable processors, but can host up to 10 DW GPU cards (with one spare PCIe x16 expansion slot) for powering massively parallel workloads. It's a true supercomputer in a rack, allowing you to unlock the full power of AI-optimised computing. Its flexible design gives users the option of using GPUs alongside other accelerator hardware if required, while PCIe Gen 4.0 support is a welcome addition.
Built for the most demanding HPC workloads, the Tyan Thunder series is a line of world class barebones server systems based on Intel's latest technology, including 3rd Gen Intel Xeon Scalable Processors. For more information about our newest products, view the Tyan/Intel Product Catalog 2022.