According to Chris Longstaff, senior director of product and technical marketing for the PowerVR group: “Although it builds on what we had before, it’s completely changed and represents a significant step up.”
The redesign was approached on the basis that performance per mW was ‘everything’, said Longstaff. “It’s a scalable architecture to ensure it can be used in multiple generations of products and for ease of integration into SoCs.”
Rogue, when first introduced, targeted GPU compute. “But the market has moved on,” Longstaff said, “with changes in gaming, VR, 4K and the need for 120frame/s graphics.”
Furian is said to address these needs through what’s called a multi-dimensional approach to performance scalability. Architectural improvements are designed for enhanced power efficiency and to maintain Imagination’s performance/mW lead over competing solutions. In Longstaff’s opinion, competitors are ‘losing efficiency at higher performance’. “They have to use more silicon to get the same performance,” he contended. “That gap will be important in some applications.”
A further architectural improvement addresses layout. “As process nodes have shrunk, larger GPUs can be integrated to get more performance,” he said. However, that brought issues with Rogue. “We’ve changed layout to get shorter paths and less congestion.”
Compared to Series7XT Plus GPUs in the same process technology and at same clock frequency, Furian brings a 35% boost in Gflops/mm2.
Imagination will be announcing cores targeted at specific applications later in 2017.