Imagination IMG CXM is a power-efficient embedded GPU family with HDR support user interfaces, thus enabling OEMs to provide users the enriching visual experience they have come to expect but at a lower cost and with more design flexibility.
The new CXM GPUs also deliver close to 50% performance density improvement over IMG BXM consumer GPUs, offer support for 10-bit RGBA/YUV to deliver a HDR graphical user interface with less visible banding, and integrate well with Imagination RISC-V CPUs.
Three CXM GPUs are available:
- The CXM-2-64 with 2 pixels per clock (PPC) and 64 FLOPS per clock is the smallest GPU with native support for HDR
- The CXM-4-64 with 4 pixels per clock (PPC) and 64 FLOPS per clock targets mainstream products
- The CXM-4-128 with 4 pixels per clock (PPC) and 128 FLOPS per clock targets is better suited to “premium DTV solutions”
All share the following features:
- 10-bit RGBA/YUV for full HDR user interfaces
- 4x MSAA to smooths the outlines of texts and images.
- TFBCv2 (Tiny Frame Buffer Compression) support with higher quality lossy/lossless compression and an additional compression level (37.5%)
- Various API support including Vulkan 1.3
- DDK optimized to support RISC-V, Arm, and x86 architectures
The main target market for the new CMX CPUs is TV box/STB SoCs with HDR support, but Imagination also lists wearables and Smart Home appliances benefiting from the smaller area and power efficiency. Wearables benefit from the smaller footprint due to space constraints, while Smart Home SoCs can fill the extra space with AI accelerators.
The company did not say when we may expect SoCs with the new IMG CMX GPUs. More details may be found on the product page, and a short video showing what image quality looks like and how the memory bandwidth is affected with TFBCv2 at different compression levels.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.