Inside the lab testing some of the world’s most capable AI chips
Hardware research engineer JohnDavid Lancaster walks us through his lab at IBM Research. Inside, he tests the latest IBM AIU chips, a new kind of computing resource that radically rethinks how to tackle the growing demand for AI processing power.
Hardware research engineer JohnDavid Lancaster walks us through his lab at IBM Research. Inside, he tests the latest IBM AIU chips, a new kind of computing resource that radically rethinks how to tackle the growing demand for AI processing power.
If someone were to say the phrase “electronics lab” to you, what would you envision? Computers hooked up to oscilloscopes, racks of servers in pieces waiting to be inspected, soldering stations with circuit boards waiting to be fixed, spare wires everywhere you look? That’s more or less the archetype of an electronics lab, and pretty much exactly what JohnDavid Lancaster’s space at IBM Research’s headquarters in Yorktown Heights, NY, looks like.
This is the “bring-up lab” for IBM Research’s prototype AIU devices. Over the last few years, IBM Research has been working on a new type of AI processor, which uses a radically different architecture to those found in traditional CPUs and GPUs, allowing it to carry out AI calculations far more efficiently than traditional devices. Recently, IBM Research worked with the IBM infrastructure team to harden this prototype and create an enterprise-grade version called the IBM Spyre Accelerator, which will power the next generation of IBM Z and IBM Power systems.
But none of that would’ve been possible if those AIU chips hadn’t first made a stop at Lancaster’s lab. Inside, he runs a gamut of tests to make sure the prototypes designed by his colleagues at Yorktown and Albany actually work as intended. With the various machines in his lab, he can check whether power is reaching all parts of a device as expected, whether they’re spinning up and shutting down safely, and if they need any modifications before being deployed. After the devices are cleared by Lancaster, they are ready to be used for experiments, whether that means deploying them somewhere like the AIU cluster IBM Research has at Yorktown’s Think Lab, or with research partners, like the University at Albany, or the University of Alabama in Huntsville.
We were recently given a sneak peek into Lancaster’s lab, what his average day entails, and how his work is ensuring future AI chips are deployed without a hitch. Watch the video above and join us on our tour of his lab.