Fundamental new approaches to computing don’t arrive every year. Heck, they don’t even arrive every decade. Last year, Intel released a research paper detailing how its new MESO (Magneto-Electric Spin Orbit) architecture could reinvigorate performance and power scaling. While promising — MESO could dramatically reduce switching voltage, switching energy, and allow for superior density scaling — Intel also stated that the technology could be as much as a decade away. We covered MESO and the paper Intel released here, for those looking for more details on the technology.
The company is singing a different tune at ISSCC this week. VentureBeat recently sat down with Intel’s CTO of AI, Amir Khosrowshahi, and Ian Young, Senior Fellow, circuit designer, and lead researcher on the MESO project. According to both men, MESO is well-suited to AI applications — so much so, that we may see the technology popping up in those spaces before we’d see it in a mainstream CPU or other product.
MESO has two features that make it potentially superior to conventional silicon for two reasons. First, it allows for significantly lower input voltages compared with CMOS (Young notes that CMOS scaling stops at 0.5v). The second advantage has to do with the types of structures that can be built with MESO. Here’s Young:
When I’m looking at finding another transistor, I have to pay attention to AI. That’s going to be a huge part of the chips that we sell. And it’s a big, big interesting coincidence that I get together with Amir, and there is an intersection. MESO can build majority gates. Majority gates is the next door neighbor to the neuron. Deep neural networks is about neurons and weights. We’ve found that this MESO technology and things that can do majority gates is very attractive in AI.
In addition to this, the power supply voltage of the next transistor has to go down. And that’s the other thing that AI is challenged by: moving data. The data movement is where all of the energy is consumed. This is where Amir really sees it: The future big-step growth in AI, if we’re really going to make it, is in what he calls the substrate, or the basic integrated circuit.
Khosrowshahi notes that we’re much more likely to see MESO debut in AI products first before jumping to CPUs.
“CPU’s, which are the most commonplace when you’re building silicon, are oddly enough the hardest thing to build,” Khosrowshahi said in the interview. “But in AI, it’s a simpler architecture. AI has regular patterns, it’s mostly compute and interconnect, and memories. Also, neural networks are very tolerant to inhomogeneities in the substrate itself. So I feel this type of technology will be adopted sooner than expected in the AI space. By 2025, it’s going to be the biggest thing.”
The Intel team’s excitement could well be justified. Reports have suggested that AI could account for 83 percent of the expected growth in semiconductor spending over the next six years. That level of growth would suggest a great deal of funding flowing into the field from every direction, and that’s the kind of environment that would make an architecture like MESO more likely to stand out in a crowd — or enable a critical advantage for the company that can bring it to market.
- With Spintronics, Intel Sees Efficiency, Density Scaling Far Beyond CMOS
- Beyond Traditional Computing: Expectations for Intel in 2019
- Intel Details Its Upcoming Sunny Cove CPU Architecture