A big win for AI hardware startup
Tenstorrent, Toronto based hardware startup developing next-generation computers, announces the addition of industry veteran Jim Keller as President, CTO, and board member.
Tenstorrent has been around just a month short of five years and has been founded by Ljubisa Bajic, ex ATI / AMD and shortly Nvidia engineer, to build processor IP and software for truly efficient deep learning. By hiring high-level engineer such as Keller’s sends an important message about the hardware startup. Tenstorrent is a next-generation computing company in the company’s own words with the mission of addressing the rapidly growing compute demands for software 2.0.
Headquartered in Toronto, Canada, with US offices in Austin, Texas, and Silicon Valley, Tenstorrent brings together experts in the field of computer architecture, ASIC design, advanced systems, and neural network compilers. Tenstorrent is backed by Eclipse Ventures and Real Ventures, among others.
Before Jim joined, Tenstorrent announced a Grayskull High-Performance AI Processor supporting A large 120MB on-chip SRAM, eight channels of LPDDR4 DRAM, and 16 lanes PCI-e Gen 4 host interface.
The targeted performance or a 65W TDP card is set to 368 TOPS, 22,431 in Resnet 50 224×224, 2,830 sentences/sec in BERT base, SQUAD, 10,150 sentences/sec in BERT base, SQUAD + conditional features and 23,345 sentences/sec BERT base, SQUAD + conditional features + low prec FP. For a comparison, latest generation Ampere based A100 80GB scores 624 TOPs in INT8 Tensor at close to 400W TDP.
“Tenstorrent was founded on the belief that the ongoing shift towards ML-centric software necessitates a corresponding transformation in computational capabilities,” said Ljubisa Bajic, Tenstorrent’s CEO. “There is nobody more capable of executing this vision than Jim Keller, a leader who is equally great at designing computers, cultures, and organizations. I am thrilled to be working with Jim and beyond excited about the possibilities our partnership unlocks.”
As CTO, Keller will lead Tenstorrent’s efforts to be the hardware solution needed to address Software 2.0. The exciting industry shift towards using machine learning methods to solve problems previously addressed by traditional software.
“Software 2.0 is the largest opportunity for computing innovation in a long time. Victory requires a comprehensive re-thinking of compute and low level software,” Keller said. “Tenstorrent has made impressive progress, and with the most promising architecture out there, we are poised to become a next gen computing giant.”
Keller’s addition to both the team and the board is a strong vote of confidence in Tenstorrent’s ability to become a major player in a sector currently dominated by behemoths.
“Tenstorrent’s differentiated approach to accelerating AI, along with the team’s pedigree, are what convinced Eclipse to initially invest in 2017. We are committed to supporting their ambitious vision, and Jim’s arrival only further bolsters our faith in that vision,” said Greg Reichow, partner at Eclipse and Tenstorrent board member.
Keller joined Tenstorrent following two years as Senior Vice President of Intel’s Silicon Engineering Group. He has held roles as Tesla’s Vice President of Autopilot and Low Voltage Hardware, Corporate Vice President and Chief Cores Architect at AMD, and Vice President of Engineering and Chief Architect at PA Semi, which was acquired by Apple Inc.
During his long and impactful career, he has helped create some of today’s most iconic technology, including the AMD Zen, K7 (Athlon) and K8 micro-architectures, Apple A4-A7 processors, the x86-64 instruction set, and HyperTransport interconnect.
Hiring Keller puts Tenstorrent on a larger scale map, and given some time, Keller can architect a significant positive impact on the company’s roadmap and help them with execution. The market of dedicated AI accelerators is very attractive to the major data center, and hyperscaler players and Tenstorrent have a chance to finish up what it started or get acquired by one of the major players in case the technology proves robust.