MIT releases processor database to track chip history beyond CPUs
The public resource covers CPUs and GPUs. Its roadmap adds specialized processors and benchmark results to help users compare hardware beyond specifications.
Loading page…
The public resource covers CPUs and GPUs. Its roadmap adds specialized processors and benchmark results to help users compare hardware beyond specifications.
Listen to this story
The MIT FutureTech-led Processor Database gives chip designers and technology buyers a searchable way to compare processor history as hardware choices become less uniform. Its released coverage is mainly CPUs and GPUs; SPEC benchmark results are planned, not yet included, and FPGAs and AI-specific processors remain on the roadmap. The historical record could help inform chip design and purchasing decisions, while giving researchers evidence to study how hardware progress shapes AI capabilities.
MIT introduced the database at Supercomputing 2026 on November 17.
The project combines MIT FutureTech’s work with CPU data Jonathan Koomey and Zachary Schmidt began compiling about five years earlier.
Stanford’s CPU Database, the earlier resource this project builds on, stopped being updated in 2016.
Researchers and chip buyers have a new public resource for tracing processor development, rather than relying on an archive that stopped receiving updates a decade ago. MIT introduced its Processor Database at Supercomputing 2026 on November 17, according to project collaborator Jonathan Koomey. The MIT FutureTech-led effort currently covers mainly CPUs and GPUs.
The project builds on Stanford’s CPU Database, which stopped being updated in 2016. MIT’s approach is to extend that foundation with modern CPUs and other kinds of processors, not simply preserve a historical collection. Research scientist Emanuele Del Sozzo describes the goal as maintaining a resource that stays useful as computing hardware changes.
Koomey’s contribution began with a separate data-gathering effort. In his November 18 account, he says he and Zachary Schmidt started compiling historical CPU characteristics about five years earlier. Initial funding came from his client at Sony PlayStation, Kieren Mayers, who wanted computing-trend data to show engineers and spur ideas.
That work converged with MIT’s after Koomey met FutureTech director Neil Thompson at a March 2024 dinner in Palo Alto. They discussed computing trends and agreed to explore combining their efforts. The resulting database brings that historical collection into a joint research project led by FutureTech.
For chip designers, the intended use is competitive research. Steering committee member Shekhar Borkar, whose career includes Intel and Qualcomm, describes looking up what rivals have built and which features they chose. A searchable history can inform the choices behind a new processor, rather than leaving designers to assemble their own collections.
Technology buyers face a related but different problem. Thompson argues that predictable, broadly shared hardware gains no longer keep technical features moving together. Chief technology officers must make more active trade-offs over which CPUs and GPUs fit their AI strategies. He sees the database as a way to examine both the options and how they are changing.
The team also plans to add results from established CPU benchmarks, including SPEC tests. Those results are meant to let users compare measured performance across processors and track progress over time. MIT describes this as a future addition, not an already completed performance dataset.
Everyone says in their sales press release that their chip is the very best, but of course, once you actually measure them, you get more variation there.
Neil Thompson, MIT FutureTech project director
Despite its relevance to AI hardware decisions, the database’s current focus is CPUs and GPUs. Del Sozzo says future releases will include FPGAs—field-programmable gate arrays—and AI-specific processors. That separates the released resource from the team’s ambition to cover a wider range of computing devices.
For researchers, the historical view serves another purpose: studying the pace of hardware improvement. Thompson says FutureTech needs that evidence, alongside changes in algorithms, to project AI’s future capabilities and impacts. Committee member Charles Leiserson similarly argues that understanding past chip development helps researchers decide which problems to pursue next.
Keeping the collection useful is also part of the plan. The team intends to combine its own work with steering committee guidance and user feedback to keep the database updated. Users can submit suggestions or additional information through the contribute button on each device page, giving them a direct route to help maintain the resource.
Loading discussion...
Join the conversation
Explain what would matter more in your decision.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.