Bittide - Data center application

Home | Projects | Bittide – Data center application

The challenge

Google operates numerous large data centers worldwide to power its products and platforms. To significantly cut energy use, Google is researching moving away from conventional, asynchronous computer networks instead opting for distributed, synchronous hardware. To do so, they need to develop specialized hardware and software, designed to seamlessly synchronize their data centers’ computers at a nanosecond level. What they needed was a partner skilled in implementing this vision, for both the software and hardware aspects.

The solution

We recognize that R&D projects demand flexibility and openness to navigate unexpected findings and new discoveries. Thus, our approach is characterized by regular, thorough discussions with our client and a collaborative spirit.

The deliverables

As a proof of concept, QBayLogic developed a fully interconnected mesh of eight Xilinx FPGA boards, linked via optical fiber. Each board can fine-tune its clocks with Skyworks clock multiplier boards. We implemented cutting-edge clock synchronization control schemes and provided a configurable platform for Google to experiment with.

Google can now use a Rust toolchain targeting a custom RISC-V core, which in turn controls the clock multiplier boards.

All our projects

Our clients

What's next?

QBayLogic will develop a custom compiler targeting these synchronous Bittide systems. Upon successful implementation, we will also guide an ASIC implementation of these systems.

Why QBayLogic?

QBayLogic stands out for its ability to harness the power of both software and digital hardware. Our Clash software-to-hardware compiler, based on the Haskell programming language, is a prime example of our innovative approach.

This unique expertise made us the ideal fit for Google’s ambitious Bittide project. Not only did Google require a hardware implementation of their ideas, but they also needed a sophisticated software stack on top of it.

This project not only demonstrates our expertise but also contributes to a more sustainable future by optimizing resource usage in data centers worldwide. Applications where the applied expertise would also fit: