HPE is creating a rapid AI supercomputer driven by the tour’s largest PROCESSOR


(Image credit: Shutterstock / Timofeev Vladimir)

Hewlett Packard Enterprise (HPE) has introduced it is creating a powerful brand new AI supercomputer in cooperation with Cerebras Systems, manufacturer of the sides largest nick.

The brand new system is going to be made up of a mixture of HPE Superdome Flex servers and Cerebras CS-2 accelerators, which are driven by the gigantic Wafer-Scale Motor 2 (WSE-2) processor.

The nameless supercomputer is anticipated to go reside later come july 1st at the Leibniz Supercomputing Middle (LRZ) within Bavaria, delivering researchers with a brand new resource to assist accelerate studies on subjects ranging from healthcare imaging in order to aerospace anatomist.

New AI supercomputer

Unveiled simply by Cerebras in April final year, the particular WS2-E was created expressly in order to accelerate AI training plus inference workloads. The nick houses an astounding 2 . six trillion transistors and 850, 000 AI cores distribute across 46, 225 mm(2) of silicon, supposedly providing the AI performance associated with hundreds of GPUs.

The wafer-sized chip furthermore boasts 40GB of on-chip memory plus 20PB/s associated with memory band width, which allows for many parameters associated with large-scale AI models to become held on-chip at the same time, accelerating computation.

Comparison of WSE2 with the largest GPU

(Image credit: Cerebras)

The launch from the new program in Australia will indicate the first time the particular WSE-2 nick has been used inside a Western supercomputer.

“We founded Cerebras to better compute, inch explained Toby Feldman, TOP DOG and Co-Founder of Cerebras Systems. “Were proud in order to partner with LRZ and HPE to give Bavarias researchers entry to blazing quick AI, allowing them to consider new ideas, train huge language versions and eventually advance medical discovery. “

The arrival from the new program has also been recognized by scientists at the LRZ, who state the machine will certainly significantly boost the speed which they can carry out important AI and general-purpose HPC workloads.

“Currently, we all observe that AI compute requirement is duplicity every 3 to 4 months with the users, mentioned Prof. Doctor Dieter Kranzlmller, Director from the LRZ.

With the high incorporation of cpus, memory plus on-board systems on a single nick, Cerebras allows high performance plus speed. This particular promises much more efficiency within data digesting and thus quicker breakthrough associated with scientific results.

Joel Khalili

Joel Khalili will be the News plus Features Publisher at TechRadar Pro, addressing cybersecurity, information privacy, impair, AI, blockchain, internet facilities, 5G, information storage plus computing. He is responsible for curating our information content, along with commissioning plus producing functions on the systems that are changing the way the planet does company.

Read More

Recent Articles


Related Stories

Leave A Reply

Please enter your comment!
Please enter your name here

Stay on op - Ge the daily news in your inbox