NVIDIA Secures Multiple COMPUTEX Best Choice Awards for Technical Achievements in Vera Rubin NVL72 Supercomputing Jetson Thor Edge Robotics and Alpamayo Self Driving Software Platforms
NVIDIA secured several honors at the COMPUTEX Best Choice Awards for technical achievements in supercomputing, edge robotics, and self driving software. An independent selection committee chose winners based on innovation, marketability, and their significance in today's technology. The three hardware and software platforms will support massive intelligence solutions across many industries:
The Vera Rubin NVL72 rack scale supercomputer received both the Golden Award and Sustainable Tech Special Award for its engineering design. This massive system combines 36 Vera CPU's and 72 Rubin GPU's through NVLink switches, which uses the 6th generation version of the technology. ConnectX 9 SuperNICs and Spectrum X Ethernet switches move data across the system, while BlueField 4 DPU's accelerate security and storage systems. In comparison, performance has increased up to 10x inference performance per watt while reducing the per token cost by 10x. Performance when paired with Groq 3 LPX is a 35x higher throughput per watt for large models.
To reduce operational costs, the platform includes a cable free, hose free, fanless, modular tray design, reducing compute tray assembly time from two hours to five minutes. The power shelves store six times more power, mitigating the impact of power swings on the grid. The system is 100% liquid cooled to 45 degrees C, so the system can be added to existing data centers and used with ambient air solutions where the heat from power is redirected for token creation.
Jetson Thor's Golden Award for its powerful high performance compute, autonomous robotics applications is the latest testament to NVIDIA's leading technology in this space. Equipped with the Blackwell graphics architecture, this module can achieve up to 2070 FP4 teraflops for artificial intelligence computations. This provides a 7.5x speed increase and a 3.5x energy efficiency increase over the Jetson Orin line of products, can be configured for 40 watts up to 130 watts and is already powering a multitude of medical, automated and industrial applications across the world.
The Alpamayo platform received an award for Vehicle Technology and Smart Cockpit for its ability to aid self driving software in addressing long tail situations where rare and difficult driving decisions must be made. This technology helps self driving cars respond correctly in even the most unpredictable situations, such as contradictory traffic signals or unanticipated pedestrian actions. Alpamayo 1.5, which has 10 billion parameters, and Alpamayo 1 incorporate Vision Language Action models, along with an open source simulator named AlpaSim. It also includes over 1700 hours of real world driving footage taken in multiple cities around the globe.
Source: nvidia
