Today semiconductor company Tachyum announced today that its Prodigy Processor AI/HPC Reference Design will be used in a supercomputer at an unnamed customer site in 2021. This reference design will provide customers, partners, OEMs and Original Design Manufacturers (ODMs) a proven blueprint for building and deploying ultra-high performance Exascale AI/HPC supercomputers and datacenters in 2021.
We are excited that customers are interested in our technology. There is never a more exciting time for a startup than the first customer, especially one who wants to deploy an AI/HPC supercomputer in 2021. The preparations at the supercomputer site are expected to start later this year, so that infrastructure is ready when Tachyum’s first hardware arrives. Tachyum has responded to its customers’ and partners’ need for an easy blueprint to enable hyperscalers and qualified ODMs and OEMs to begin changing the competitive landscape, while accelerate democratizing AI/HPC in the process,” said Dr. Radoslav “Rado” Danilak, Tachyum founder and CEO. “Every Prodigy-equipped datacenter is also a low-cost, scalable AI/HPC datacenter.”
Tachyum’s Prodigy Universal Processor, slated for commercial availability in 2021, is a 64-core processor with a clock speed in excess of 4GHz that brings to market new “Universal Computing” capabilities. According to the company, Prodigy “handily outperforms the fastest processors while consuming one-tenth the electrical power, and it is one-third the cost and outperforms GPUs and TPUs on neural net training and inference workloads.”
The reference design platform is architected to deliver the power-performance benefits of Prodigy to wide ranging applications in verticals, including AI/HPC supercomputing, Edge Computing, Datacenter operations (public and private cloud), Telecommunications, and Automotive. Today’s datacenters are under relentless pressure to support both regular and AI powered workloads. To address this, a heterogeneous infrastructure is being built, with x86 servers for regular workloads and GPU/TPU accelerators for AI workloads. Additional hardware, plus time and energy are needed to move the enormous amount of data between the two computing silos.
Datacenters are provisioned with enough servers to accommodate peak customer demand; during off-peak hours, more than 50% of most cloud datacenter servers are powered down to save electricity costs. Servers equipped with Prodigy offer the highest performance at the lowest cost to power conventional applications and also provide low-cost AI on demand. Idle Prodigy servers can be seamlessly and dynamically powered up and used for AI training or inference workloads. With Prodigy, provisioning an AI environment becomes CAPEX-free, since the idle servers powered up to handle AI workloads are already on the books as capital equipment
Tachyum’s Prodigy Universal Processor is the smallest and fastest general purpose, 64-core processor developed to date, requiring 10x less processor power, and reducing processor cost by 3x. Prodigy will directly enable a 32-Tensor Exaflop supercomputer and allow the building of machines more powerful than the human brain in 2021, years ahead of industry expectations. Prodigy reduces datacenter annual TCO (total cost of ownership) by 4x, through its disruptive processor architecture and a smart compiler that has made many parts of the hardware found in typical processors redundant. Fewer transistors, fewer and shorter wires, due to a smaller, simpler core, translates into much greater speed and power efficiency for the Prodigy processor.
