British semiconductor firm Graphcore has unveiled the first Wafer-on-Wafer (WoW) processor alongside setting out its roadmap for an “ultra-intelligence AI supercomputer”.
The chip unveiled today, the Bow IPU, is the world’s first processor to be based on TSMC’s Wafer-on-Wafer (WoW) technology.
“TSMC has worked closely with Graphcore as a leading customer for our breakthrough SoIC-WoW solution as their pioneering designs in cutting-edge parallel processing architectures make them an ideal match for our technology,” said Paul de Bot, GM of TSMC Europe.
WoW bonds two wafers together to generate a new 3D die:
- The first wafer is for AI processing and is architecturally compatible with the GC200 IPU processor. It has 1,472 independent IPU-Core tiles, is capable of running more than 8,800 threads, and has 900MB of in-processor memory.
- The second wafer is a power delivery die and features deep trench capacitors that enable a large performance increase thanks to being located right next to the processing cores and memory.
Compared to its predecessors, Graphcore claims that its Bow IPU offers up to 40 percent higher performance and 16 percent better power efficiency for real-world AI applications.
In terms of power, Graphcore says the flagship Bow Pod delivers more than 89 petaFLOPS of AI compute. The superscale Bow POD ups that to an incredible 350 petaFLOPS.
Here’s how that power translates into real-world performance across popular AI applications:
Ultra-intelligence AI supercomputer
The most interesting announcement made by Graphcore today is of its plan to develop an “ultra-intelligence AI supercomputer”.
Graphcore points to how the human brain has “approximately 100 billion neurons and more than 100 trillion parameters in a biological-neural-network system that delivers a level of compute yet to be matched by any silicon computers.”
Graphcore says that it’s developing an AI computer that will surpass the parametric capacity of the brain.
The computer will be called the ‘Good’ computer, named after computer science pioneer Jack Good (born Isadore Jacob Gudak).
The achievements of Jack Good – including his pivotal work during the Second World War – are worth reading up on. However, for the purposes of this story, what’s most notable is the fact Good was the first person to describe a machine that is more powerful than a human brain in his 1965 paper Speculations Concerning the First Ultra-Intelligent Machine.
“Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever,” wrote Good.
“Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an ‘intelligence explosion,’ and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make.”
By 2024, Graphcore expects to have delivered the first ultra-intelligence AI computer.
The computer will feature:
- Over 10 exaFLOPS of AI floating point compute
- Up to four petabytes of memory with a bandwidth of over 10 petabytes/second
- Support for AI model sizes of 500 trillion parameters
- 3D wafer on wafer logic stack
- Fully supported by Graphcore’s Poplar® SDK
- Expected cost: ~$120 million (configuration dependent)
Graphcore says it will provide further updates about the Good computer over the coming year.
(Imagery Credit: Graphcore)
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo. The next events in the series will be held in Santa Clara on 11-12 May 2022, Amsterdam on 20-21 September 2022, and London on 1-2 December 2022.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.