
Digital cable link key to quantum supercomputing deal
SEEQC has teamed up with Nvidia to link its quantum processors directly to Grace Hopper GPUs.
This would create the first chip-based quantum-to-GPU computing system to bring quantum supercomputing from SEEQC closer to datacentre-scale with infrastructure for quantum AI.
When completed, this will be the first time an active multi-chip module quantum processor will be directly linked with both a GPU- and a CPU-operational platform, creating a tightly coupled, fully digital integration of quantum and classical computing technologies.
“We are connecting quantum and classical computing in a highly efficient way, taking advantage of a tight integration of both technologies to power systems that will run the most important quantum and classical applications,” said Matthew Hutchings, chief product officer and co-founder of SEEQC. “The only way to do this and accommodate high speed, low latency and high bandwidth is to do so on a chip-to-chip basis. This collaboration creates not just the possibility for scalable, real-time error correction in the near-term, but future quantum applications, many of which are unimaginable today.”
“We will develop a proprietary digital chip-to-chip link that does not require high fidelity, superconducting RF cables like competing readout and quantum-classical interface technology,” he said. “Instead it will use high bandwidth digital ribbon cables to link the chip modules.”
- First fully digital chips for full stack quantum computers
- Italy gets its first quantum computer from SEEQC
- Superconductor IC spinoff eyes ‘fully digital’ quantum chjip
The SEEQC Single Flux Quantum (SFQ) superconducting quantum processor sits just above absolute zero at 20mK with a digital interface chip, the DQM, to provide the data link. The ribbon cable will link to an interface chip at an intermediate temperature before connecting to the GH200 Grace Hopper system on chip at room temperature. This is key as the high performance combined GPU and ARM chips are built on CMOs that is characterised at room temperature.
“The Grace Hopper will remain at room temperature,” said Hutchings. “The key bottlenecks to scaling the quantum classical interconnect is no longer the temperature difference once you eliminate the analogue bottlenecks of existing interfaces. It is the high fidelity, high bandwidth and slow analogue interface that creates the scaling challenges between cryogenic and room temperature in conventional systems. An interface chip at intermediate temperature will act as a digital amplifier.”
As SEEQC’s technology is entirely digital, it will remove several analog steps and noisy hardware that limits the bandwidth and latency of quantum supercomputing systems and the company has a co-development partnership with Dutch quantum startup QuantWare.
SEEQC’s digital chip is also compatible with all quantum computing technologies from superconducting and silicon spin to photonics, trapped ion, neutral and cold atom systems.
“This all-digital integration will take advantage of each system for a low-latency interface while maintaining the highest possible bandwidth performance from each individual system,” said John Levy, CEO and co-founder of SEEQC. “The development we’re taking on with Nvidia represents the best of breed in both quantum and classical; and, together, both core technologies create unprecedented compute power.”
- £6.5m project to develop cryogenic CMOS IP for quantum chips
- Cryogenic CMOS design for quantum chips on 22FDX
- Cryogenic CMOS IC operates at -270ºC, controls up to 128 qubits
Integrating quantum and GPUs will advance the NVIDIA CUDA Quantum platform for hybrid quantum-classical computing. This collaboration is a major milestone in SEEQC’s technology roadmap. With GPU acceleration and a software platform that is unique to each industry, SEEQC will now be able to develop a full-stack quantum computing architecture that can enable hybrid quantum AI and ML programs, as well as scalable real-time error correction — all powered by SEEQC’s proprietary SFQ technology.
“Tight integration of quantum with GPU supercomputing is essential for progress toward useful quantum computing,” said Tim Costa, Director, HPC and Quantum Computing Product at NVIDIA. “Coupling the NVIDIA GH200 Grace Hopper Superchip with SEEQC’s digital chip architecture — tied together by the CUDA Quantum programming model — will provide a major step toward that goal.”
