[ad_1]
Copyright of the image
Cerebras systems
Wafer Scale Engine is the world's largest computer chip
A Californian start-up has unveiled what it says is the largest computer chip in the world.
The Wafer Scale Engine Engine, designed by Cerebras Systems, is slightly larger than a standard iPad.
The company says that a single chip can drive complex artificial intelligence (AI) systems, ranging from driverless cars to surveillance software.
However, one expert suggested that innovation would not be practical to install in many data centers.
Copyright of the image
Cerebras systems
The chip measures 21.5 cm square (3.3 square inches)
Why is development important?
Computer chips have generally become smaller and faster over the years.
Dozens are usually made on a single "wafer" of silicon, which is then separated to separate them from each other.
The most powerful desktop processors (CPUs) have about 30 processor cores, each capable of managing its own set of calculations simultaneously.
GPUs tend to have more cores, even if they are less powerful.
This has traditionally made it the preferred option for artificial intelligence processes, which can be broken down into several parts and run simultaneously, when the result of a calculation does not determine the input of Another.
Examples include voice recognition, image processing, and pattern matching. The most powerful GPUs have up to 5,000 hearts.
But Cerebras' new chip has 400,000 cores, all interconnected by broadband connections.
The company believes this gives it an advantage in managing complex machine learning problems with less delay and less power than combinations of other options.
Copyright of the image
Getty Images
A typical silicon wafer contains about 100 computer chips
Cerebras says the Wafer scale engine will reduce the time required to process complex data from a few months to minutes.
Its founder and general manager, Andrew Feldman, said the company had "overcome decades-old technical challenges" with limited chip size.
"Reducing training time eliminates a major obstacle to the progress of the industry," he said.
Cerebras started shipping the material to a small number of customers.
It has not yet revealed the cost of fleas.
Copyright of the image
Cerebras systems
A worker inspects chips as they are made
What are the disadvantages?
Although fleas process information much more quickly, Dr. Ian Cutress, editor of the AnandTech news site, said that advances in technology would come at a cost.
"One of the advantages of smaller chips is that they consume much less energy and are easier to keep cool," he explained.
"When you start using larger chips like this, companies need a specialized infrastructure to support them, which will limit their usability.
"That is why it is suitable for the development of artificial intelligence, because that is where the big investments are going."
Copyright of the image
Cerebras systems
A close-up of the motor at the scale of wafers during manufacture
Is this the first AI chip?
Cerebras is far from being the first company to develop chips for powering artificial intelligence systems.
In 2016, Google developed TPU (Tensor Processing Unit) chips to power software, including its translation application, and sells them to third parties.
The following year, the Chinese Huawei announced that its Kirin smartphone chips had acquired a Neural Processing Unit (NPU) to speed up the calculation of dot matrix multiplications – a type of calculation commonly used in tasks related to l & # 39; AI.
But all these efforts have not succeeded.
In the early 1980s, the US company Trilogy received funds of several hundred million dollars to create its own super-chip.
However, the processors became too hot during testing and were less powerful than originally planned.
In the face of technical and personal challenges, the company abandoned the project five years later.
[ad_2]
Source link