Video Analytics – Fine Tuning for Big Gains
AEWIN has been working hard to provide highest performance out of your hardware investments. We’ve been tasked to provide GPU accelerated MEC edge servers for video analytics application. Specifically for accelerated live traffic monitoring, which utilizes NVIDIA T4 to decode live video feeds and perform AI inferencing to detect traffic patterns by different types of vehicles, collisions, traffic violations, license plates, and other useful metrics.
This seems like an easy task for servers to handle. However, not all servers are created equal. Off the shelf servers are typically optimized for various different workloads, making it jack of all trades and master of none. Our MEC servers and engineers have been steadily focused on maximizing GPU acceleration performance. We received customer praise due to our ability to handle 24x full HD live streams compared to their previous system which is only able to process 18 streams utilizing similar setup. Through continued testing and fine tuning, we’ve exceeded the performance of some of the biggest names in the server space and closely matching the benchmark published by NVIDIA.
Table: Resnet-50v1.5 + TensorRT Inference Test
*Test setup: SCB-1921 MEC version + 2x Nvidia T4 vs published numbers by Nvidia and competitors
This is the difference between going with a general purpose server from other manufacturers versus AEWIN. Through personalized collaboration, it enable us to fine tune performance of the servers and GPU for your specific workloads. Our flexible manufacturing, which is located right next to our engineering/headquarters right here in New Taipei City, allows us the flexibility to work with customers from smaller scale projects to large. Our experienced engineers are available and can offer insights and give recommendations for system configs. Make sure to bring your projects to AEWIN before deciding on your next GPU accelerated platform and see how AEWIN can help you get the most performance out of your investment.
- Support Dual 2nd Gen Intel Xeon Scalable Processors
- Support 2933MHz System Memory
- 4x PCIe x8 for Network Expansion Module, plus 4x PCIe x8 or 2x PCIe x16 for GPU or FPGA