A new AI startup has released what it says is the largest semiconductor chip ever constructed. Cerebral Systems has taken the wraps off its Wafer Scale Engine (WSE), which packs in 1.2 trillion transistors. It is a new laptop optimized completely for AI and deep studying abilities. This places it way ahead of different chips currently available on the market, with recent launches from other chip leaders, AMD and Nvidia, packing between 20 and 30 billion transistors.
Enormous
Cerebral says that its hardware will benefit AI processing abilities that are often held back by a loss of computing capacity. The employer says testing a new hypothesis or schooling a new version is costly and time-consuming, often taking weeks or months on currently available structures. Cerebral believes its new chip can help meet the “substantial computational demands” of deep learning, being purpose-constructed to permit more computation than other hardware available today.
Unlike many excessive-powered chips on the market today, which are simply a group of chips piled on top of a silicon wafer, the WSE is a single chip on a single wafer, bearing in mind plenty of faster switch speeds and processing ability. The WSE is more than 56X larger than the biggest photo processing unit, containing three 000X greater on-chip memory and capable of reaching more than 10,000X the reminiscence bandwidth, with high-speed memory close to each middle, ensuring that cores are constantly occupied doing calculations.
Across its 46,225 rectangular millimeters of silicon (making it slightly smaller than a regular tablet), the chip can deliver 400,000 AI-optimised cores, and its specialized memory architecture guarantees each of those cores operates at maximum performance. The WSE affords 18 gigabytes of speedy, on-chip memory disbursed among the cores in an unmarried-degree reminiscence hierarchy, one clock cycle far from each center. There’s no information on whether the hardware could be used for wider use, with Cerebral simply saying that updates may be coming quickly.
