Tachyum Prodigy Hits Milestone With 96 Percent Of Silicon Designed


Tachyum™ Inc. introduced that it has reached one other milestone in assembly its objective of quantity manufacturing of the Prodigy Universal Processor in 2021 by attaining 96 p.c of silicon designed and format accomplished, with solely a steady netlist format to go earlier than the ultimate netlist and tape out.

The firm has been making regular progress in its march towards Prodigy’s product launch subsequent yr. While advancing the state of Prodigy’s design to 96 p.c completion, Tachyum additional confirmed in verification that Prodigy has accurately executed directions straight from DDR5 reminiscence via coherent mesh with the Prodigy core producing right outcomes. The firm additionally confirmed that this newest Prodigy post-layout netlist maintains clock pace targets with no die measurement development from its earlier netlist format milestone. With Prodigy’s first bundle pin-out accomplished, cache miss dealing with over coherent interconnect has been verified, as have nearly all of directions.

Tachyum has efficiently compiled the design to its FPGA emulation of Prodigy and has ensured that compiled processor tiles match into the FPGA emulation system. Full chip FPGA emulation will probably be launched to format and manufacturing throughout the subsequent few weeks. These achievements intently observe Tachyum’s earlier milestones in efficiency testing and {hardware} connectivity, and software program compatibility, they usually proceed to reveal that Prodigy is on its approach to market in 2021.

“We continue to successfully meet every key metric of our aggressive timeline for producing Prodigy Universal Processor chips for mass deployment in data center, AI and HPC workload environments,” stated Dr. Radoslav Danilak, Tachyum founder and CEO. “By nearing 100 percent completion of the silicon design of our product, we are ensuring that organizations will finally have the performance, power efficiency and cost advantages they need to solve the most challenging issues facing them years earlier than previously expected.”

Tachyum’s Prodigy can run HPC functions, convolutional neural web AI, explainable AI, basic AI, bio AI and spiking neural networks, in addition to regular knowledge heart workloads on a single homogeneous processor platform, with a easy and acquainted programming mannequin. Using CPU, GPU, TPU and different accelerators in lieu of Prodigy for these several types of workloads is grossly inefficient. A heterogeneous processing cloth, with distinctive {hardware} devoted to every sort of workload (e.g. knowledge heart, AI, HPC), leads to a big underutilization of invaluable {hardware} assets, and creates a tougher programming setting. Prodigy’s capability to seamlessly change amongst these numerous workloads dramatically adjustments the aggressive panorama and the economics of information facilities, and Big-AI.

Prodigy considerably improves computational efficiency, power consumption, {hardware} (server) utilization and area necessities in comparison with current processors provisioned in hyperscale knowledge facilities at the moment. Prodigy may even enable Edge builders for IoT to take advantage of its low energy / excessive efficiency, together with its easy programming mannequin to ship AI to the sting.
Prodigy is really a common processor. In addition to native Prodigy code, it additionally runs legacy x86, ARM and RISC-V binaries. And, with a single, extremely environment friendly processor structure, Prodigy delivers industry-leading efficiency throughout knowledge heart, AI, and HPC workloads. Prodigy, the corporate’s flagship Universal Processor, will enter quantity manufacturing in 2021. In April the Prodigy chip efficiently proved its viability with an entire full chip format, exceeding clock pace design targets. In August the processor was in a position to accurately execute brief applications, with outcomes mechanically verified towards the golden software program mannequin, whereas exceeding the goal clock pace. The subsequent step is to get a completely useful FPGA emulation of the Prodigy chip later this yr, which is the final prototype milestone earlier than closing tape-out.

Prodigy outperforms the quickest Xeon processors at 10x decrease energy on knowledge heart workloads, in addition to outperforming NVIDIA’s quickest GPU on HPC, AI coaching and inference. The 125 HPC Prodigy racks can ship a 32 tensor EXAFLOPS. Prodigy’s 3X decrease price per MIPS and 10X decrease energy interprets to a 4X decrease knowledge heart Total Cost of Ownership (TCO), allows billions of {dollars} of financial savings for hyperscalers resembling Google, Facebook, Amazon, Alibaba, and others. Since Prodigy is the world’s solely processor that may change between knowledge heart, AI and HPC workloads, unused servers can be utilized as CAPEX-free AI or HPC cloud, as a result of the servers have already been amortized.

Sign up for the free insideBIGDATA e-newsletter.


Source hyperlink

Write a comment