Neousys Rugged Edge AI Inference and GPU-aided Platforms

The Artificial Intelligence (AI) technology can be found in many advanced applications such as machine learning, autonomous driving, intelligent video analytics, facial image and speech recognition with deep-learning breakthroughs. As GPUs are naturally great at parallel workloads, they can speed up the AI learning process dramatically.

Neousys provides rugged and scalable computing platforms supporting various GPU graphics cards (shown below) that works as GPU-aided edge computing platform and can benefit the development/implementation of related applications.

Among machine learning applications, there are two different phases: AI Training and Inference. AI training requires intensive computing power to develop a neural network model while AI inference focuses on analysis and prediction in the field (edge computing), based on the trained model.

Tesla GPUs are especially great at delivering real-time inference as it is powered by tensor cores, a revolutionary technology that can accelerate large matrix operations and perform mixed-precision matrix calculations. Tesla T4 GPU's 130TOPS in INT8 can provide up to 40X faster performance. Using only 75W power, it is an ideal fit for Neousys Cassette design and makes Nuvo-7164GC/7166GC an affordable solution for inference decision-making applications.

What sets Nuvo-7164GC/ 7166GC apart from the rest?

Neousys' patented Cassette and dedicated air tunnel design, which guides the intake air to flow through the passive heat sink of Tesla T4/P4 to effectively dissipate the heat generated by the GPU. The design allows the system to operate under 60°C ambient temperature with sustained 100% GPU loading.

Nuvo-7164GC supports one PCIe cassette while Nuvo-7166GC has dual PCIe expansion slots for Tesla T4/ P4 plus one more PCIe slot for add-on card such as the 10G NIC card. As the new bundle unit, Neousys can offer edge devices with high-speed transmission for inference acceleration.

Nuvo-7160GC, 7164GC, and 7166GC are based on the same platform with dedicated Cassette modules as the only difference. You can buy one platform with other Cassette modules to take advantage of the benefits various GPUs offer!