The video shows the performance of China’s Fenghua GPU in the GFXBench benchmark
Recently, Innosilicon released the performance indicators of the “Fenghua 1” GPU on its official website, so that everyone can learn more about this GPU using Imagination Technologies’ PowerVR architecture. “Fenghua 1” GPU will theoretically be closer to NVIDIA GeForce RTX 2060, and the dual GPU version can even catch up with GeForce RTX 3060 on paper, but the actual performance still has to wait for further verification.
Recently, Twitter user @Stewrandall shared a video of using the Fenghua 1 GPU to start and run, showing that the Fenghua 1 GPU is undergoing the GFXBench benchmark. Unfortunately, there is no display of the number of frames on the screen, nor the final score. Although the GFXBench benchmark cannot be fully used as a standard for judging the performance of a discrete graphics card, at least it seems that the picture is still very smooth. In addition, in the video, it is not possible to determine whether the single-core type A card or the dual-core type B card is used. The blogger then released a picture of a Type B card, but TomsHardware prefers that the video runs on a Type A card.
Fenghua 1 GPU is manufactured with 12nm process, FP32 floating point performance is 5T FLOPS, AI performance is 25 TOPS (INT8), the interface is PCIe 4.0 x16, the video memory uses GDDR6/GDDR6X, the maximum capacity is 16GB, the video memory bandwidth is 128 bits, the rate is 19 Gbps, and the bandwidth is 304 GB/s, support HDMI 2.1 and DisplayPort 1.4 output. In addition, the power consumption of the Fenghua 1 GPU does not seem to be high, and the performance per watt has also aroused many people’s interest.
At present, Innosilicon is providing samples to interested manufacturers, but the actual retail and product pricing is still unclear for the time being.