GPU-to-AI startup Vastai Technologies (Shanghai, China) has launched its 7nm GPU for cloud AI applications, the SG100.
Jun Qian, co-founder and CEO, demonstrated the 7nm cloud GPU chip SG100 claiming it offers industry-leading graphics rendering performance and world-leading encoding capabilities of ultra-high throughput, ultra-high quality and low latency. It integrates rendering, AI, video integration and optimization for several metaverse applications such as cloud games, cloud mobile phones, cloud desktops, and cloud computing.
In the same presentation Jun described the Vastai Unified Computing Architecture, the data center inference card Zaitian VA10, the edge inference card Zaitian VE1 and the VastStream software platform.
Cards for center and edge
The Zaitian VA10 consumes 150 watts of power while achieving twice the performance of mainstream GPUs with the same power consumption and 6 percent of the latency, Vastai claimed. The Zaitian VA10 is suitable for cloud AI application deployments such as live video enhancement, intelligent traffic management and real-time semantic understanding, the company said.
The Zaitian VE1 series has a power consumption of 40 to 65 watts, and the peak computing power of INT8 data reaches 100 TOPS. This accelerator card supports real-time decoding of 60-channel 1080P video and supports mainstream AI frameworks. Zaitian VE1 is suitable for edge computing scenarios such as smart transportation, vehicle-road collaboration, and industrial quality inspection, Vastai said.
Qian has 25 years of silicon design experience including 13 years with Cisco Systems designing networking ASICs and 9 years at AMD designing GPUs. Qian led the AMD China team to design the first 7nm graphics processor and the first 7nm GPU architecture AI chip. Co-founder and CTO Louis Zhang was formerly AMD Fellow with expertise in video-related technology and has 23 years of SoC and IP architecture development.
Related links and articles: