Imagination Applied sciences, the company these days most identified for its GPUs utilized by Apple’s iPhones and iPads, has simply made a large leap in its AI strategy: it removed its devoted NPUs and added related performance to its GPUs, in response to EE Times. Coincidentally, it simply obtained $100 million in funding from Fortress Funding Group.
Imagination has been round for about 40 years and has performed many issues. Nonetheless, it grew to become identified for its PowerVR Kyro-branded graphics processors for PCs and its PowerVR-badged GPU IP, which was utilized by Apple and Intel (and dozens extra firms) for its GPUs. Imagination can also be a participant in the AI area. It has ‘rebooted’ its strategy to AI by discontinuing its standalone neural community accelerators and concentrating on embedding AI capabilities into its GPU IP merchandise.
Imagination admitted in the EE Instances interview that the strategic shift was revamped the previous 18 months because of complexities with a proprietary software program stack that might not maintain tempo with the fast evolution of AI clients’ various wants, which required customized options.
Given the dominance of Nvidia’s CUDA, the challenges prolonged past {hardware} improvement; creating an AI software program stack that might expose the capabilities of ImgTec’s NPUs to builders and clients proved to be a major hurdle. Clients usually introduced their distinctive AI fashions and anticipated Imagination to optimize them for the company’s {hardware}. This led to intense competitors and stress on Imagination’s improvement workforce, which pressured the workforce with none monetary output.
Software program is essential.
Recognizing these difficulties, Imagination shifted its focus to GPUs, that are naturally multithreaded and are naturally suited to duties requiring environment friendly parallel processing and information motion. With their flexibility, the company believes that GPUs may be enhanced with further AI-specific compute capabilities, making them well-suited for edge AI purposes, particularly in units with present GPUs, similar to smartphones, the house turf for ImgTec.
To help this transition, Imagination is repurposing know-how developed for its now-discontinued accelerators, similar to its graph compiler, and integrating it into its GPU stack. This enables the company to leverage its strengths whereas competitively addressing AI workloads.
For instance, Imagination collaborated with the UXL Basis on SYCL, a framework designed to rival Nvidia’s CUDA. This shift additionally aligns with the wants of Imagination’s clients, who already use on-chip GPUs for AI and graphics processing.
Maybe that openness has enabled Imagination to safe a $100 million convertible time period mortgage from Fortress Funding Group associates.
Whereas Imagination is dedicated to this GPU-focused strategy, the company stays open to revisiting devoted AI accelerators in the long run, relying on how AI software program infrastructure evolves, in response to EE Instances. For now, the company believes that compute-oriented GPUs supply one of the best platform to satisfy their clients’ AI wants.