Imagination Technologies also announced that it's working together with TSMC to optimize its PowerVR chip designs for the foundry's 28nm HPM and future 16nm FinFET processes. The company claims that initial efforts have already achieved a 25 percent overall performance improvement for the PowerVR series 6 "Rogue" GPU core, with key blocks demonstrating up to 30 percent improvement compared to existing design flows.
Imagination R&D teams working with TSMC are developing highly-optimised IP libraries and fully characterised reference system designs for Imagination’s IP cores on TSMC’s advanced processes including 28HPM and 16nm FinFET technologies. This collaboration will enable mutual customers to achieve significant performance gains for next-generation applications. Leveraging extensive analysis of factors affecting performance, Imagination is taping out a highly optimised next-generation multi-core, multi-cluster PowerVR series 6 GPU design to characterise these performance improvements and plans to demonstrate a test chip later this year. The resulting reference flow and optimised libraries will enable SoC designers to significantly accelerate time-to-market and increase performance for similar implementations of PowerVR Series6 GPUs.
“The growing reliance on graphics processors to power the best user experiences means the GPU has become the new silicon process driver for the coming decade. It occupies as much as 50% or more of leading edge SoCs, and demands the highest performance memory bandwidth while always delivering high performance at low power. These characteristics are critical to defining the key device parameters of future generation processes. We’re delighted to be working closely with TSMC to ensure future silicon technologies are fully tuned to the needs of next generation’s most advanced SoCs,” said Tony King-Smith, executive vice president of Imagination.