NVIDIA’s GPU IP integrated into MediaTek’s Dimension Auto SoC

MediaTek launched its new Dimensity automotive cockpit chip series this week, covering the entire market from entry-level to high-end. While automotive chip launches aren’t usually the most interesting things to do, this launch will be an exception to that rule, as MediaTek is leveraging its graphics IP: NVIDIA’s chips. This means that the upcoming Dimensity automotive cockpit chip will be the first chip based on NVIDIA GeForce graphics technology released by a third-party (non-NVIDIA) supplier.

NVIDIA’s first attempts to license its GPU IP to third parties date back to 2013, when the company proposed licensing its Kepler GPU IP to compete with Arm and Imagination Technologies. This effort failed at the time. But more than a decade later, NVIDIA is making another effort to license some of its IP, and it looks like NVIDIA has finally succeeded. All in all, MediaTek’s new Dimensity automotive cockpit system-on-chip will rely on NVIDIA’s GPU IP, Drive OS and CUDA, creating historic development for both companies.

MediaTek’s next-generation Dimensity automotive cockpit processor family consists of four different SoCs, including CX-1 for top-tier cars, CY-1, CM-1 and CV-1 for entry-level cars. These are highly integrated SoCs that package a general-purpose Armv9-A-based CPU core as well as NVIDIA’s next-generation graphics processing unit IP. NVIDIA’s GPU IP can run AI workloads for driver assistance and power infotainment systems because it fully supports graphics technologies such as real-time ray tracing and DLSS 3 image upscaling.

According to HardwareLuxx, the Dimensity Auto Cockpit processor is a single-chip SoC with built-in multi-camera HDR ISP. The ISP supports front-facing, in-cabin and bird’s-eye view cameras for a variety of security applications. Additionally, these processors feature audio DSPs that support various voice assistants.

MediaTek’s announcement did not reveal which generation of NVIDIA graphics IP they are using, only that it is a “next generation” design. Automotive SoC releases tend to be quite conservative given the certification requirements involved, so it remains to be seen how the “next generation” of this graphics IP actually compares to the current generation Ada Lovelace architecture.

The new MediaTek SoC will be fully supported by NVIDIA Drive OS, which is already widely used by car manufacturers. This will allow automakers to unify their software stacks and use the same set of software for all MediaTek Dimensity-powered cars. Additionally, since NVIDIA’s Drive OS fully supports CUDA, TensorRT, and Nsight, MediaTek’s Dimensity SoC will be able to take advantage of AI applications developed for the green company platform.

Jerry Yu, senior vice president and general manager of CCM Business Group at MediaTek, said: “Generative AI is changing the automotive industry just as it revolutionized the mobile market with more personalized and intuitive computing experiences.” “Dimensity Automotive Cockpit Product Portfolio “Set a new wave of AI entertainment in cars, and our unified hardware and software platform allows automakers to easily expand AI capabilities across their entire product lines.”

There is no doubt that licensing graphics IP and platform IP to third parties marks a milestone for NVIDIA, especially its efforts in the automotive field. Leveraging DriveOS and CUDA outside of NVIDIA’s own hardware platform is a big deal for a business unit that NVIDIA has long believed was poised for significant growth but has faced stiff competition and slow growth from conservative automakers. Adoption rate. In the meantime, what remains to be seen is how MediaTek’s new Dimensity Auto Cockpit processor will stack up against NVIDIA’s previously announced Thor SoC and related DRIVE Thor platform, which integrates a Blackwell-based GPU and delivers 800 TFLOPS of 8-bit floating point AI performance.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *