NVIDIA teams up with MediaTek on the development of GB10 Grace Blackwell Superchip
MediaTek and NVIDIA Collaborate on AI Superchip
MediaTek, the world's leading chip supplier for smartphones, smart TVs, Arm-based Chromebooks, Android tablets, and voice assistant devices (VAD), has teamed up with NVIDIA to design the NVIDIA GB10 Grace Blackwell Superchip. This collaboration is aimed at ushering in a new era of innovation and making AI ubiquitous.
The GB10 Superchip, part of NVIDIA Project DIGITS, combines a Blackwell GPU from NVIDIA with a 20-core Arm-based CPU from MediaTek. This results in an ultra-compact AI computing platform that fits in the palm of a hand yet remains scalable and powerful.
Applications
The DGX Spark personal AI supercomputer (Project DIGITS) enables AI developers, researchers, and data scientists to run large-scale AI workloads and inference tasks on a small desktop device. It supports running advanced models, including those with up to 200 billion parameters, making it suitable for cutting-edge AI research and development. It is anticipated to be deployed as an AI appliance at the edge, providing users with compute capabilities without the need for larger rack-mounted or cloud-based infrastructure.
Benefits
The compact form factor of the DGX Spark offers a powerful "desktop supercomputer" that facilitates AI development outside traditional data centers. It is scalable by connecting multiple DGX Spark machines to increase performance. High bandwidth unified memory of 128GB enhances performance for AI workloads. The collaboration with MediaTek also enables wider accessibility to NVIDIA's AI ecosystem since large and expensive infrastructure is no longer required to start using NVIDIA’s advanced AI hardware.
Future Implications
This collaboration empowers developers and organizations of all sizes, including smaller businesses and research institutions, to harness advanced AI, accelerating AI democratization. It supports edge AI applications by offering high-performance AI inference capabilities in compact devices. The collaboration strengthens NVIDIA’s AI ecosystem by enabling scalable solutions from a desktop level (DGX Spark) to data center-grade machines (DGX Workstations with GB300 Superchip).
MediaTek’s collaboration with NVIDIA signals a critical fabless semiconductor partnership, integrating efficient Arm-based CPUs tuned for AI with NVIDIA’s GPU technologies, potentially influencing future AI hardware designs and collaborations.
For more information about the NVIDIA GB10 Superchip and Project DIGITS personal AI supercomputer, visit NVIDIA's official website.
[1] MediaTek and NVIDIA Collaborate on NVIDIA GB10 Grace Blackwell Superchip. (2022, March 16). Retrieved April 20, 2022, from https://www.mediatek.com/media/news-details/mediatek-and-nvidia-collaborate-on-nvidia-gb10-grace-blackwell-superchip
[2] NVIDIA DGX Spark. (n.d.). Retrieved April 20, 2022, from https://developer.nvidia.com/dgx-spark
[3] NVIDIA and MediaTek Collaborate to Deliver AI Supercomputing to the Edge. (2022, March 16). Retrieved April 20, 2022, from https://www.nvidia.com/en-us/about-nvidia/press-room/news-releases/nvidia-and-mediatek-collaborate-to-deliver-ai-supercomputing-to-the-edge/
[4] MediaTek and NVIDIA Collaborate to Bring AI Experiences Inside the Car. (2021, August 10). Retrieved April 20, 2022, from https://www.mediatek.com/media/news-details/mediatek-and-nvidia-collaborate-to-bring-ai-experiences-inside-the-car
[5] MediaTek and NVIDIA Collaborate to Deliver Advanced Edge AI Capabilities to IoT Applications. (2022, March 16). Retrieved April 20, 2022, from https://www.mediatek.com/media/news-details/mediatek-and-nvidia-collaborate-to-deliver-advanced-edge-ai-capabilities-to-iot-applications
- The collaboration between MediaTek and NVIDIA aims to make artificial-intelligence ubiquitous by designing the NVIDIA GB10 Grace Blackwell Superchip, which combines a Blackwell GPU from NVIDIA and a 20-core Arm-based CPU from MediaTek.
- The compact DGX Spark, part of NVIDIA Project DIGITS, is enabled by this collaboration, providing AI developers and researchers with a desktop device capable of running large-scale AI workloads using advanced models that incorporate artificial-intelligence technology.
