A new AI startup has launched what he believes is the largest semiconductor chip ever built.
Cerebras Systems has removed the wraps from its Wafer Scale Engine (WSE), which contains 1.2 trillion transistors, as a new type of computer that is exclusively optimized for AI and in-depth learning opportunities.
This is way ahead of other chips currently on the market, with recent launches from other chip leaders such as AMD and Nvidia packing between 20 and 30 billion transistors.
- AMD processors: the best AMD CPU ' s in 2019
- What is AI? Everything you need to know
- Is AI the new personal test?
Cerebras says its hardware will benefit from AI processing capabilities, which are often blocked by a lack of available computing capacity.
The company says testing a single new hypothesis or training a new model is both expensive and time-consuming and often takes weeks or even months on currently available systems.
Cerebras believes its new chip can help meet the "enormous computational requirements" of deep learning, which is specifically built to enable better calculations than any other hardware currently on the market.
Unlike many other powerful chips currently on the market, which are actually a collection of chips that are stacked on top of a silicon wafer, the WSE is a single chip on a single wafer, resulting in much higher transfer speeds and processing capacity.
The WSE is more than 56X larger than the largest graphics processing unit, contains 3,000X more memory on the chip and is able to reach more than 10,000X the memory bandwidth, with high-speed memory close to each core so cores are always busy calculating .
Over its 46.225 square millimeters of silicon (making it slightly smaller than a normal tablet), the chip can deliver 400,000 AI-optimized cores, and its specialized memory architecture ensures that each of these cores operates at maximum efficiency.
The WSE offers 18 gigabytes of fast, on-chip memory distributed over the cores in a one-level memory hierarchy, one clock cycle away from each core.
There is no news yet about when the hardware will be available for wider use, with Cerebras simply saying that updates will be released soon.
- The best laptops for programming in 2019