Imagination Technologies recently unveiled its chiplet based graphics architecture, offering up to 6 TFLOPs of computing performance in a PCIe form-factor. The core difference between this multi-GPU solution and traditional AFR-based offerings is that it pulls workloads to the various GPU chiplets, instead of pushing them. Furthermore, to the client application, this solution will be seen as a single GPU block instead of multiple, resolving any compatibilty issues that arise with SLI and XFX.
Innosilicon is a China-based provider of ASIC and LSI solutions such as mining farms, small scale servers, neural network accelerators, etc. This deal with Imagination Technologies will basically allow it to build GPUs for both consumer and Data Center markets using the chiplet design and pairing it with GDDR6 memory.
Imagination’s BXT architecture delivers the level of performance and power efficiency we had been looking for. Innosilicon has a solid track record in delivering first-class high-speed and high-bandwidth computing solutions in advanced FinFET process nodes. Building on this success and strong customer demand, we are announcing our upcoming product which is a standalone high-performance 4K/8K PCI-E Gen4 GPU card, set to hit the market very soon, that will power 5G cloud gaming and data center applications. With a solid foundation in Innosilicon’ s GDDR6 high-speed memory, cache-coherent chiplet innovation, and high-performance multimedia processor optimization, a move into a standalone PCI-E form-factor GPU is natural for us. Thanks to BXT’s multi-core scalable architecture, we are able to build a customized GPU solution integrating graphic and intelligence computing to meet the needs of high-end desktop and data center applications.Roger Mao, Vice President of Engineering, Innosilicon
While these discrete graphics solutions won’t really come close to NVIDIA and AMD’s offering in terms of performance, they might just be the right low-cost option for miners and vendors offering cloud-based solutions.
The biggest advantage of Imagination’s IP is the flexible design and highly scalable nature of the architecture. With the same building blocks as AMD’s Zen architecture, they can build highly power-efficient designs as well as powerful ones by mixing a higher number of chiplets or simply running them in parallel in a multi-socket machine.
Imagination is excited to have a new partner in Innosilicon. Innosilicon has a strong and mature design team with a track record for innovative high-performance products built using advanced processes over the years. With the rise of 5G and high-speed Wi-Fi 6, cloud gaming and data center GPU server is a rapidly growing market, and we are looking forward to seeing the impact of our low-power, high-efficiency technology in this space.”Graham Deacon, Vice President of Business Development, Imagination