'Universal' processor integrates CPU, GPU, and TPU capabilities

April 23, 2021 // By Rich Pell
'Universal' processor integrates CPU, GPU, and TPU capabilities
AI supercomputer company Tachyum has unveiled what it says is the world's first universal processor - a single processor that delivers industry leading performance for data center, AI, and HPC workloads.

The company says it has achieved engineering proof of performance for all critical core components of its Prodigy Universal Processor prototype - which integrates CPU, GPU, and TPU capabilities into a single homogeneous architecture - which was built using field-programmable gate array (FPGA) emulation boards. The processor, with 2 x 64 cores, greater than 4 GHz speed, and less than 400 W power consumption, is claimed to outperform the fastest Xeon @ 10x lower power, 3x lower cost per MIPS (millions of instructions per second), and 4x lower data center annual total cost of ownership while minimizing space requirements, and enabling existing legacy applications to run without changes.

In addition, says the company, Prodigy outperforms NVIDIA’s fastest GPU by more than two times in HPC, as well as AI training and inference. The company's Government business entity is seeking customers to use the processors’s fully functional FPGA emulation for product evaluation, performance measurements, as well as software development, debug and compatibility testing.

A fully functional chip emulation (in FPGA-based hardware) is typically the last step, before chip tape-out when the final design is submitted for fabrication in silicon. The Prodigy FPGA emulation system will help customers smooth the adoption curve for Prodigy in their existing or new data center and/or HPC systems that demand high performance, high utilization and low power.

Specifically, says the company, Prodigy helps to connect the dots, convert data into insights, and enable a decision cycle that is faster than the competition. Although the company is focused on the data center and some edge computing solutions, the microprocessors that will deliver tensor petaflops of AI performance per chip can be used to address many challenges in many vertical markets.

Some government areas of interest include, but are not limited to: ultra-high performance supercomputers, ultra-low power satellites, computing intensive applications, UAVs, cell towers, 5G adoption, and a broad array of sensor applications for AI fusion at the


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.