Tachyum Corporation announced today that the final version of its Prodigy FPGA emulation system will be released next year before the chip goes into production and is generally available. As part of the announcement, the company has decided to terminate the prototype system purchase program previously offered to commercial and federal customers.
The latest version of the hardware FPGA prototype unit will ensure that Tachyum achieves its extremely high reliability test goals of more than 10 petaflop cycles before the first Prodigy chips hit the market. Tachyum's software emulation system is expanding its capabilities to make open source software ported prior to the introduction of Prodigy chips more accessible.
The final release also includes the following improvements:
- After the number of cores increased to 192, signals were added between boards to support expansion of more than 128 cores.
- Minor fixes to support for high-capacity DIMMs.
- Other debugging optimizations have been made.
- Adjust the BMC-UEFI hardware to simplify the communication process.
- Replace board-to-board connectors to improve user experience.
Pictured: Tachyum builds the last batch of FPGA prototypes
Dr. Radoslav Danilak, Founder and CEO of Tachyum, said: "We are very proud to have reached this milestone in our development process as Prodigy processors are about to be taped out and in mass production. We have always been committed to providing the world's smallest, fastest, and most environmentally friendly general-purpose chips. Ensuring that this goal is achieved from day one is our top priority, and we are excited to launch this industry-changing product."
As a general-purpose processor, Prodigy delivers industry-leading performance for a wide range of workloads. It supports data center servers and enables seamless and dynamic switching between computing domains such as AI/ML, high-performance computing (HPC), and cloud computing through a single, homogeneous architecture. Prodigy dramatically reduces capital and operating costs by eliminating the need for expensive dedicated AI hardware and dramatically improving server utilization, while delivering breakthroughs in performance, power, and economics for data centers. Designed to handle complex workloads in AI and ML, Prodigy's high-frequency and multi-core architecture provides the necessary computing power to support deep learning and other machine learning algorithms. The AI performance of the Prodigy processor is several times that of the leading products on the market, which means that it can handle large-scale neural networks and data analysis tasks more efficiently.
Prodigy integrates 192 high-performance, custom-designed 64-bit compute cores that deliver 4.5x the performance of the highest-performance x86 processors for cloud workloads, 3x the performance of the highest-performance GPUs for HPC, and 6x the performance for AI applications. This marks a significant step forward for Tachyum in advancing computing efficiency and performance.
