A new GPU startup has just been born in Silicon Valley, one of the first in the last 25 years or so, with the ex-RTG boss announcing his new startup -- Oxmiq Labs -- and they've got world-class architects to make AI GPUs.
Koduri retired from Intel as its Chief GPU architect close to two years ago, previously working for AMD as the Senior VP and Chief GPU Architect of its graphics department -- Radeon Technologies Group (RTG) -- but is now building AI GPUs from scratch with Oxmiq Labs.
Oxmiq Labs will represent 500+ years of collective experience and will use hundreds of patents to design new AI hardware, with the company not just tweaking current GPU designs, but rather will be rebuilding the GPU ecosystem from the ground up with an "Atoms to Atoms" mindset. This will see the development of low-level hardware and bring transistor-level innovation to create high-end AI systems.
The new GPU startup will be taking a software-first approach to constructing its AI chip designs and has designed a tool to allow software programs written for NVIDIA CUDA to work on non-NVIDIA hardware "without code modification or recompilation".
Oxmiq has decided to pursue building IP instead of a complete chip design because it wanted to avoid the high costs associated with building cutting-edge chips, which can cost over $500 million to design.
Jim Keller, CEO of Tenstorrent, said: "We're excited to partner with OXMIQ on their OXPython software stack. OXPython's ability to bring Python® workloads for CUDA to AI platforms like Wormhole™ and Blackhole™ is great for developer portability and ecosystem expansion. It aligns with our goal of letting developers open and own their entire AI stack".
Lawrence Loh, SVP of MediaTek, added: "OXMIQ has an impressive bold vision and world-class team. The company's GPU IP and software innovations will drive a new era of compute flexibility across devices - from mobile to automotive to AI on the edge".




