MediaTek is allegedly working with NVIDIA to integrate its GeForce graphics IP into its next-gen smartphone flagship SoC as early as next year. NVIDIA has established near-complete dominance over the last few years in the discrete graphics card market. Buoyed by the RTX 30 and RTX 40 series GPUs launch, the chipmaker controls over 80% of the PC gaming space, leaving AMD and Intel with scraps.
With every passing generation, NVIDIA has been increasing its focus on machine learning and neural networks, developing the required hardware and software in the gaming and data center segments. MediaTek will leverage the gaming and AI capabilities of NVIDIA’s latest graphics architecture to one-up its rivals in the smartphone sector.
It’s hard to guess what kind of IP NVIDIA will integrate into MediaTek’s flagship smartphones, but don’t expect PC-level feature support. Die space and power efficiency are key to mobile SoCs, which may prevent the inclusion of RT and Tensor (matrix) cores. Given the immense potential of the latter, though, we may see a cut-down variant on MediaTek’s latest processors.
In addition to smartphone SoCs, NVIDIA and MediaTek are also poised to cooperate in developing WOA (Windows on Arm) SoCs. Team Green can leverage the full potential of its next-gen Ada Lovelace graphics architecture on this platform. You’ve got cutting-edge H/W accelerated ray-tracing, advanced upscalers like DLSS 3, and more to show off.
All of NVIDIA’s Ada and Hopper graphics products are fabbed on TSMC’s 5nm process node. The 3nm (N3) capacity is limited, so the two are likely to adopt the 4nm (N4) node to produce these processors.
AMD and Samsung tried something similar with the Xclipse GPU on the Exynos 2200 SoC, but it didn’t work well. Lackluster performance and excessive power consumption doomed the chip to oblivion.