Tachyum Inc. today declared that it has successfully finished a demonstration showing its Prodigy Universal Processor running faster than some other processor, HPC or AI chips, including ones from NVIDIA and Intel. This is the latest of numerous ongoing milestones accomplished by Tachyum as the organization continues its walk towards Prodigy's item release one year from now.
Tachyum demonstrated how its computational activity and the speed of its item design, using an industry-standard Verilog simulation of the genuine Prodigy post format equipment, is the superior solution to current serious offerings. In addition to the fact that Prodigy executes instructions at exceptionally high speeds, yet Tachyum now has an infrastructure actualized for consequently checking right results from the Verilog RTL. These computerized tests check Verilog yield for correctness contrasted with Tachyum's C-model, which was used to measure execution, and is presently the 'Brilliant Model' for the Verilog equipment simulation to ensure it produces indistinguishable, step-by-step results.
This confirmation milestone drastically increases Tachyum's profitability and its capacity to test the Prodigy equipment design productively to discover bugs and right them before tape-out. With this latest accomplishment, Tachyum now has robotized the constrained irregular test age ability, which further adds to its efficiency.
Tachyum's previous equipment design milestone was to fabricate components and interconnect them, which was successfully finished in April. The most late equipment design milestone - and resulting device - is about the Prodigy processor creating right results and its exhibition on test programs. Prodigy is currently dealing with branch mispredictions, or compiler misprediction of memory reliance, whereupon it detects, recovers and produces right results.
Thanks to Tachyum's IP suppliers, the organization is currently ready to do peruse/writes from Prodigy communications mesh to its DDR5 DIMMs equipment memory models. The worldwide clock is currently associated from the PLL to Prodigy cores. RAMBIST and other manufacturability features are currently coordinated into the Prodigy equipment design in enormous part because of Tachyum's physical design accomplice.
"This latest equipment milestone is a testament to the industrious work of our designing group and the vast HR we have had the option to assemble to finish a progressive solution at no other time seen," said Dr. Radoslav Danilak, Tachyum author and CEO. "We set out to deliver the highest exhibition, lowest energy and most cost-proficient processor for the hyperscale, HPC and AI commercial center, and these milestones are demonstrating that we are accomplishing those goals. With an item that is faster than the fastest Intel Xeon or NVIDIA A100 Chips, Prodigy is approaching the entirety of its stated objectives and remains on target to make its introduction as arranged one year from now."
Tachyum's Prodigy can run HPC applications, convolution AI, reasonable AI, general AI, bio AI and spiking neural networks, as well as would be expected server farm workloads on a single homogeneous processor stage with its simple programming model. Using CPU, GPU, TPU and different accelerators in lieu of Prodigy for these various types of workloads is wasteful. A heterogeneous processing texture, with one of a kind equipment devoted to each sort of outstanding task at hand (for example server farm, AI, HPC), results in underutilization of equipment resources, and an all the more testing programming climate. Prodigy's capacity to seamlessly switch among these various workloads drastically changes the serious landscape and the economics of server farms.
0 Comments