Graphcore, an ambitious and well-funded British chip designer focused solely on artificial intelligence applications, has unveiled what it says is the world’s most complex chip – the Colossus MK2 or GC200 IPU.
The processor has 59.4 billion transistors and offers an eight-fold performance boost over the company’s Colossus MK1, Graphcore says. It has more than the 54 billion transistors found in Nvidia’s A100, which previously held the title of world’s largest processor, and which the American firm announced earlier this year.
Each GC200 chip has 1,472 independent processor cores and 8,832 separate parallel threads, all supported by 900MB of RAM in the processor. Graphcore will make the GC200 available through its new IPU machine, the M2000, which contains four GC200 chips in a package the size of a pizza box and offers 1 petaflop of total computation. The firm says its new hardware is “completely plug-and-play” and that customers will be able to connect up to 64,000 IPUs together for a total of 16 exaflops of computing power.
The announcement comes at a time when the world of chips is still rocked by the advent of artificial intelligence. AI model training requires highly parallel processors, a lawsuit that attracted new players to the market (like Graphcore) and even encouraged some existing tech giants (like Google) to make their own specialized chips.
So far, however, Nvidia has dominated the market, as its GPUs, originally designed to speed up graphics rendering in video games, have proven to be perfect for AI processing. Graphcore is trying to challenge that dominance and has already attracted significant amounts of funding and veteran sponsors in the tech industry like Microsoft and Dell.
Earlier this year, Graphcore announced that it had attracted $ 150 million in R&D funding in its latest round of financing, giving it a total valuation of $ 1.95 billion. The company, which was founded in 2012, at the time when the trend towards deep learning really took off, claims that its biggest advantage is that its chips have been designed from scratch with AI in mind.
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/20083266/GC010_IPU2_004_W4K.jpg)
Karl Freund, analyst at Moor Insights & Strategy, said The edge He was impressed by Graphcore’s latest offering, particularly updates to its software, which is key to properly harnessing the enormous parallel processing power required for AI.
“What Graphcore focuses on is not just the chip, but the system,” says Freund. “Meaningful neural network training cannot be done on a single chip, it must be done on hundreds, thousands, perhaps tens of thousands of them, and that scalability factor is really what, in my opinion, makes Graphcore stand out. ”
He notes, for example, that the Graphcore support software is “very complete for a startup,” capable of interacting with a wide range of AI frameworks and offering the kind of workload monitoring tools that allow researchers to take advantage of your hardware to the max.
It seems that even in a hardware market, where companies compete for the number of transistors that can fit on a chip, it is software that will still make or break a company’s fortune.
