Glenfly Tech, a Shanghai-based graphics processor designer, says its first Arm-based system-on-chip, the GF800, has finished chip debugging and is now sampling to end customers. The company showed the chip and its boards at the 26th China International Industry Fair in Shanghai, where its executives described the product as a shift from supplying graphics cards to offering a complete system-level platform, according to an October 16 interview by Xinhua Finance.
The GF800 pairs an all-big-core CPU of eight Arm Cortex-A78 cores with a Glenfly-designed GPU, a dedicated neural processing unit and a unified memory architecture shared by the CPU, GPU, NPU and video units. Glenfly's official announcement, first posted on its website in August, says this is the first time the company has integrated its own GPU IP into an SoC. It lists Arm Mali and Immortalis, Qualcomm Adreno, Samsung Xclipse and Imagination PowerVR as the main GPU options in today's Arm SoC market and presents its design as an additional, independently developed choice.
The GPU supports DirectX 12, OpenGL 4.6, OpenGL ES 3.2, OpenCL 3.0 and Vulkan 1.3, and Glenfly says it has passed the conformance tests for each standard. It also includes a dedicated 2D engine. The chip is built on the company's Glenfly HMUA hardware architecture and Glenfly HUCA software platform, which Glenfly says are meant to handle model migration, mainstream deep learning frameworks, large-model inference and multimodal deployment.
For displays, the GF800 offers HDMI 2.1, DP 1.4, eDP 1.4b and DSI, drives up to four independent displays and supports output up to 8K, with HDCP content protection. A new image signal processor accepts inputs of up to 48 megapixels. Interfaces include PCIe 4.0, USB 3.2, SATA, eMMC and Gigabit Ethernet. On software, it is optimized for domestic operating systems including UnionTech's UOS and Kylin, supports Linux and Android, and is compatible with Windows.
Target uses include AI terminals, edge servers, industrial computing equipment and robots. Vice general manager Wu Wenjie said the aim is not to simply move cloud computing onto devices, but to bring computing closer to data, devices and industrial applications within real power and cost limits. Chief strategy officer Wang Yuanfeng said the chip is now going through adaptation and solution validation for graphics, display, system compatibility and industry scenarios.
Glenfly has not disclosed the NPU's performance rating, the manufacturing process or a timeline for volume production.
SEE ALSO: Axera Unveils AX8860 Edge AI Chip Series With 256 TOPS and Up to 128 GB Memory · Langdao Silex Shows an EUV Light Source Prototype