Tachyum unveils a monster processor that does everything

Audio participant loading…

The wait is lastly over and the US-startup Tachyum has now launched its Prodigy common processor which mixes the performance of a CPU, GPU and TPU in a single processor.

With a view to develop its revolutionary new processor, the corporate first got down to conquer the processor efficiency plateau in nanometer-class chips and the programs they energy.

Along with being Tachyum’s first industrial product, the Prodigy Cloud/AI/HPC supercomputer processor chip presents 4 instances the efficiency of the quickest Intel Xeon processor. Nonetheless, it additionally has 3 times extra uncooked efficiency than the Nvidia H100 on HPC in addition to six instances extra uncooked efficiency on AI coaching and inference workloads.

With its new processor, Tachyum believes it’s poised to beat the challenges of accelerating knowledge heart energy consumption, low server utilization and stalled efficiency scaling.

Tachyum Prodigy

The Tachyum Prodigy options 128 high-performance unified 64-bit cores working at as much as 5.7 GHz with 16 DDR5 reminiscence controllers and 64 PCIe 5.0 lanes. All this uncooked energy can simply be deployed in a knowledge heart because of the truth that the corporate has additionally developed rack options for each air-cooled and liquid-cooled knowledge facilities.

Not like different CPU and GPU options, Tachyum’s Prodigy was designed to deal with matrix and vector processing from the bottom up and the brand new processor helps a variety of information sorts together with FP64, FP32, TF32, BF16, Int8, FP8 and TAI.

By rising efficiency whereas utilizing much less energy, Prodigy additionally solves the issue of sustainable data center growth by providing unparalleled carbon footprint discount in addition to unprecedented knowledge heart TCO financial savings.

Founder and CEO of Tachyum, Dr. Radoslav Danilak defined in a press release how the corporate’s Prodigy processor will help in reworking hyperscale knowledge facilities into common computing facilities, saying:

“We have now lengthy believed in our capability to beat Moore’s Regulation to rework hyperscale knowledge facilities into true common computing facilities. With the launch of Prodigy, we’ve begun the revolution. Prodigy’s capability to allow human brain-scale AI whereas concurrently decreasing knowledge heart energy consumption and decreasing the TCO of hyperscale knowledge facilities and supercomputer programs, is a breakthrough for a projected $100 billion business. By launching Prodigy, we’re advancing not solely the state of expertise however making the world a greener place as nicely.”

Sampling for Tachyum’s Prodigy will start later this 12 months and the processor is anticipated to enter quantity manufacturing throughout the first half of 2023.

Source

Leave a Reply

Your email address will not be published.