SUNNYVALE, Calif.–(Enterprise WIRE)–Nov 16, 2021–
Cerebras Methods Inc. (“Cerebras”), the pioneer in accelperiodting synthetic intelligence (AI) compute, right now introduced at SC21 A mannequin new Cerebras Computer software Enchancment Package (SDK) Which will allow the developer group to create new wafer-scale purposes throughout computational fluid dynamics, molecular dynamics, signal processing, and extra. The SDK is in Beta now and Shall be out tright here yr finish.
The SDK will permit A Quantity of AI and HPC constructers to invent and look at new ideas on the CS-2 system, and its Industry main Wafer Scale Engine (WSE-2), at a extra versatile, decrease diploma. The WSE-2 is the Industry’s faslook at and Strongest processor. With 850,000 cores, 40 Gigabytes of on-chip reminiscence, 20 Petabytes/s of reminiscence bandwidth, and 220 Petabits/s Of supplies bandwidth, the WSE-2 has 123 occasions extra cores, 1,000 occasions extra on chip reminiscence, 12,733 occasions extra reminiscence bandwidth, and 45,833 occasions extra supplies bandwidth than the main graphics processing unit. With the SDK, and A mannequin new area-particular programming language based mostly on C and acquainted parallel programming ideas, the large computational resupplys of the WSE-2 are put immediately Inside the palms of the developer, restricted solely by their imaginations.
“We’re excited To work together with the developer group, particularly right here at SC21, To assist accelperiodte their work On the vanguard of HPC and AI,” said Andy Hock, Vice chairman of Product, Cerebras. “Powered by our 850,000 core, second period wafer-scale engine with extreme bandwidth on-chip reminiscence and interconnect, our CS-2 system is uniquely-suited To assist buyers scale current workloads and unlock new ones. We will not wait to see what We will construct collectively.”
Primitives will embrace a library of neural community kernels and a library of BLAS kernels. Using this, constructers can program standalone purposes, or They will mix new gadgets with current kernels in a modular style, using The identical devices that engineers at Cerebras use for inner low-diploma development. The objective of the SDK is allow buyers to create issues which have by no means been carried out earlier than, to discover the boundaries of our Industry, and to shortly And merely look at these ideas on the WSE-2.
Additionally at SC21, Cerebras buyer Argonne Nationwide Labs introduced A mannequin new ALCF AI Testbed that consists of the Cerebras CS-2 to assist pioneering evaluation at the intersection of AI and extreme efficiency compute (HPC). The Nationwide Power Know-how Laboratory (NETL) also shared particulars on work with Cerebras to accelperiodte each simulations for improved cybersecurity and computational fluid dynamics to mannequin climate and earth systems dynamics, astrophysical simulations and electrodynamics.
Cerebras has expanded its footprint past the U.S. with new workplaces in Tokyo, Japan and Toronto, Canada. Its roster Of consumers has grown signalificantly To incorporate Argonne Nationwide Laboratory, Lawrence Liverextra Nationwide Laboratory, Pittsburgh Supercomputing Center (PSC) for its groundbreaking Neocortex AI superpc, EPCC, the supercomputing center at the College of Edinburgh, Tokyo Electron System, GlaxoSmithKline, and AstraZeneca.
About Cerebras Methods
Cerebras Methods is a group of pioneering pc architects, pc scientists, deep studying evaluationers, and Material specialists of All varieties. We now have come collectively To assemble A mannequin new class of pc system, designaled for the singular objective of accelperiodting AI and altering The biggest method forward for AI work eternally. Our Cerebras CS-2 system, powered by the world’s largest processor – the WSE-2, permits buyers to accelperiodte their deep studying by orders of magnitude over widespread objective compute.
View supply mannequin on Industrywire.com:https://www……..