Intel Tackles Next-gen Computing with Quantum and Neuromorphic Innovations
At Intel Innovation 2022, Intel is announcing its "software first" approach to quantum and a new stackable, multi-board platform based on the Loihi 2.
With traditional device scaling and the forces of Moore’s Law coming to a halt, modern approaches to computing are beginning to reach their limits. Instead, many developers have turned their sights to promising new computing methods, including quantum computing and neuromorphic computing. While both of these technologies hold exciting potential, there is still significant work to be done in each field.
Intel believes it is taking a full-stack approach to quantum computing. Image used courtesy of Intel
Today, Intel is pushing the state of both quantum and neuromorphic computing with new releases at Intel Innovation 2022. All About Circuits heard from Anne Matsuura, director of the quantum and molecular technologies at Intel, and Mike Davies, director of Intel’s neuromorphic computing lab, to hear about the new releases firsthand.
Intel Envisions a Quantum Future with New SDK
Many quantum computing researchers are focusing on qubits and hardware for advancing the technology, but significant breakthroughs are needed across the full software stack.
Intel has announced the release of its new software development kit (SDK) for quantum computing, called the Intel Quantum SDK. The new SDK is designed to optimize quantum-classical algorithms, known as variational algorithms, which are among the most popular quantum workloads today.
Current and future state of Intel’s Quantum SDK. Image used courtesy of Intel
Describing the SDK, Matsuura explained, “The Intel Quantum SDK includes a compiler that targets a binary quantum instruction set and a quantum runtime to manage the execution of the quantum program. This then enables users to find quantum operations and decomposes them into operations that are available on the Intel Quantum Dot Qubit chip.”
To do this, a key tool in the Quantum SDK is a C++, low-level virtual machine (LLVM)-based compiler toolchain. Intel has judiciously chosen to enhance the industry standard LLVM representation with quantum extensions using industry standards to ultimately open-source the compiler’s front end.
According to Intel, the Quantum SDK has already experienced positive reviews from beta users who have utilized the tool in applications such as material and fluid dynamic simulations. To help drive further adoption, Intel has already awarded grants to several universities, including the University of Pennsylvania and Ohio State University, to use and teach the Quantum SDK in their curriculum.
Intel Targets Neuromorphic Computing with Kapoho Point
Intel has historically been a leader in the neuromorphic computing hardware world, with its flagship Loihi 2 neuromorphic chip released over a year ago. Now, the company is taking Loihi one step further with the release of Kapoho Point, an 8-chip Loihi 2 board designed for large-scale workloads.
The Kapoho Point board can be easily stacked. Image used courtesy of Intel
Coming in a compact form factor, a single Kapoho Point board has the ability to hold up to 8 million neurons and 1 billion parameters. Further, Kapoho Point is stackable, meaning it can be easily scaled.
Davies explained, “Since it's stackable, you can just layer these on through this expansion port and build out to a larger capability. We've assembled stacks of as many as four boards here, which get up to 32 million neurons. Sometimes a couple of thousand variables can define a very difficult problem, so we're excited to see the scale of optimization problems Loihi 2 is going to be able to tackle with this kind of a form factor.”
Compared to a state-of-the-art CPU, Kapoho boasts 1,000x better energy efficiency. Further, improvements in inter-chip communication bandwidth and an update to the Lava framework work together to enable deep learning applications to be 12x faster and 15x lower energy than Loihi 1 systems.
Quantum and Neuromorphic Computing Charge Ahead
As conventional computing begins to slow down, the industry has turned its attention to new technologies such as quantum computing and neuromorphic computing. Intel is aiming to establish itself as a leader in both of these areas.
"We need to encourage a larger group of people who are confident in programming a quantum computer," Matsuura remarked. "I think IBM had it right in that sense when they got users using the software toolchains to get used to programming quantum. Intel's Quantum SDK is our entry into that race—but we've made it a little bit different based on C++."
Speaking about Kapoho Point, Davies added, "The way we see neuromorphic computing is enabling AI in motion, or movement in the real world."