For the standard Invida A100, the domestic calculation GPU Muxi MXC500 only takes 5 hours to light up successfully.

Thanks to CTOnews.com netizen Xiao Zhan cut the clue delivery! CTOnews.com June 15 news, domestic GPU manufacturer Cersei MetaX announced yesterday, the company's self-developed Xiyun MXC500 series GPU successfully lit up, only 5 hours to complete the chip function test.

Cersei has launched MXN series GPUs (Xi Si) for AI reasoning, MXC series GPUs (Xi Yun) for AI training and general purpose computing, and MXG series GPUs (Xi Cai) for graphics rendering.

MXC500 is the latest product of Xiyun series. It is claimed to be the computing chip of Nvidia A100 / A800. The target FP32 computing power is 15 TFLOPS (FP32 performance of Nvidia A100 is 19.5 TFLOPS). It adopts general GPU architecture and is compatible with CUDA. It is expected to ship at the end of the year.

▲ Xiyun MXC series GPU, picture source Xixi In addition, Xixi's first artificial intelligence reasoning GPU Xixi N100 has also been mass-produced, built-in MXN100 heterogeneous GPGPU processor and HBM2E memory, single card computing power up to 160TOPS (INT8) and 80TFLOPS (FP16), can provide up to 128 channels of encoding and 96 channels of decoding Media Processing Service capabilities, compatible with HEVC, H.264, AV1, AVS2 and other video formats, It supports up to 8K resolution.

▲ The first artificial intelligence reasoning GPU of Xi Si N100, according to the co-founder Yang Jian of Xi Xi, by 2025, Xi Xi will launch a complete GPU product integrating graphics rendering, that is, GPU for display and gaming purposes.