Intel plans to produce its next-generation Xe HP DG2 GPU on TSMC's 7nm process node.

Intel is rumored to be producing their next-generation Xe graphics architecture-based DG2 GPU on TSMC's 7nm process node. The DG2 first appeared in test drivers alongside the DG1 GPU, with details limited to the EU count of three variants and the fact that DG2 uses Xe HP (High-Performance) microarchitecture instead of the Xe LP (Low-Power) suited for entry-tier designs such as the DG1 GPU. According to AdoredTV's sources, Intel will be introducing the DG2 GPU in 2022 and will be relying on TSMC's 7nm process node for production rather than Intel's own in-house 7nm EUV technology, despite Intel's plans to cover losses from their 10nm process struggles with their own 7nm technology for CPUs.

Intel's prioritization of their in-house 7nm process for CPUs rather than GPUs stems from cost and supply considerations. While Intel had revealed that their 7nm lead product would be a Xe architecture-based GP-GPU using Foveros packaging for the datacenter AI and HPC market—now known as Ponte Vecchio—the company is prioritizing a large portion of their fabs to CPUs, as the Xe HPC class GPU is a very niche product that doesn't require as much 7nm volume as a mainstream line of GPUs such as Xe LP or Xe HP. TSMC's 7nm+ EUV process hit volume production in the second quarter of 2019, but Intel chose the standard 7nm route due to higher costs associated with more advanced nodes and supply issues already apparent from the 10nm roadmap.

Intel's 7nm process node will deliver approximately 2 times scaling and is expected to provide approximately 20 percent increase in performance per watt with a 4 times reduction in design rule complexity, marking the company's first commercial use of extreme ultraviolet (EUV) lithography. The company plans optimizations of the 7nm process node with 7nm+ in 2022 and 7nm++ in 2023, with both iterations delivering a good list of additions over 10nm technology.

The DG2 will be powered by Intel's 2nd generation Xe graphics architecture with a planned 2022 launch, placing it against next-generation GPUs by NVIDIA and AMD utilizing more advanced TSMC nodes. Both NVIDIA's Ampere and AMD's 2nd Generation Navi would have already been in the market for months or a whole year by the time DG2 launches, suggesting Intel would likely aim the budget and entry-level market segment. Development of the DG2 GPU appears incomplete, as the DG1—the first generation Xe graphics chip—is expected to ship to ISVs this quarter with a year-end launch, though this is the second report citing Intel's 7nm GPU launch for 2022. Intel has confirmed two products based on their Xe graphics architecture using different generations of Xe architecture, with further details expected from the company at GDC 2020 in March.