Sign up for our newsletter and get the latest HPC news and analysis.
Send me information from insideHPC:


Tachyum Moves Forward with Prodigy Chip for Extreme Power Efficiency

Today Tachyum announced that it has achieved a significant engineering milestone in its Prodigy product development, by proving that all critical components of its Prodigy core meet or exceed its stated design goals of unprecedented compute performance, power efficiency and cost advantages.

Achieving engineering-proof of our Prodigy design is an important step to show that our chips will perform to the theoretical limits for which they were designed,” said Dr. Radoslav Danilak, Co-founder and CEO of Tachyum. “While this is not a full prototype, we have built the most difficult parts of the chip. This achievement shows that what we have developed is doable. Since we have exceeded design goal targets for the critical components, we fully expect to hit all targets for the non-critical components as well.”

Tachyum achieved engineering proof of performance using the Cadence Innovus Implementation System (industry standard CAD tool) to place & route (incl. parasitic extraction), its design of all critical core components. This proof point, which cannot be challenged, has confirmed that the core design to be utilized in its final Prodigy product, has achieved Tachyum performance design goals using only standards cells. Tachyum used 7nm standard cells provided by the world’s largest silicon fabricator to build all critical circuits of the Prodigy core – arithmetic logic units, register files and data caches, etc. – Tachyum met its initial design goal of 4 GHz with a healthy margin. This not only validates the Prodigy architecture’s enormous headroom, it also demonstrates that the company’s engineering team has all the necessary capabilities required to complete the Prodigy processor. Complete results are detailed in a 35,000-page post-layout timing report indicating the speed Prodigy should achieve when fabricated in a state-of-the-art 7nm process. This represents a major risk reduction milestone for Tachyum product development.

Engineering-proof of the Prodigy Universal Processor Chip comes a year after the company showed that its architecture was viable by running industry-standard benchmark test suites on the Tachyum processor simulation in software. The company is expected to expand equity investment opportunities, and plans on releasing an FPGA prototype in 2019, with volume chip production in 2020.

This milestone overcomes any engineering risk that may have been counted against Tachyum and shows that they are a company that had better be taken seriously,” said Adrian Vycital, IPM Managing Partner and a Tachyum Board Member. “While we were fortunate to have bought into Tachyum’s vision early, we anticipate that there will be additional interest from the broader investment community and potential OEM partners, who will look to leverage Tachyum’s success to build out their own data center, next generation AI, and HPC solutions.”

Tachyum’s Prodigy Universal Processor Chip is the smallest and fastest general purpose, 64-core processor developed to date, requiring 10x less power and reducing processor cost by 3x. Tachyum’s Prodigy processor reduces data center server power by an order of magnitude, through a disruptive hardware architecture and a smart compiler that has made many parts of the hardware found in a typical processor redundant. Fewer wires and shorter wires, due to a smaller, simpler core, translates into much greater speed and power efficiency for the processor. The ultra-low power Prodigy processor will allow system integrators to build a 32 Tensor Exaflops AI supercomputer beginning in 2020. This will enable users to simulate, in real-time, human brain-sized Neural Networks well ahead of the EU’s stated goal of 2028.

Sign up for our insideHPC Newsletter

Leave a Comment

*

Resource Links: