According to a report by a Taiwanese outlet, Intel might outsource its 2nd generation of Xe graphics cards to TSMC’s 5nm foundries due to insufficient capacity. Intel is already facing 14nm shortages pertaining to existing consumer and server lineups. At present, nearly all the company’s products on the 14nm process node which makes it even worse. The first batch of 10nm chips was shipped in limited volume last year under the Ice Lake lineup, and the succeeding Tiger Lake parts to set to land later this year.
The generation of Xe GPUs will mainly be limited to LP integrated products in the consumer markets while the HPC offerings (if any) will be based on the 10nm+ node. The second generation of Xe HPC processors is expected to leverage Intel’s own 7nm node, with a 2021 launch. However, there’s no word on the consumer LP and HP counterparts. As per reports, the Xe graphics cards for gamers and mainstream users will be based on TSMC’s 5nm process, the same as NVIDIA’s Hopper and AMD’s Navi 3x.
Since TSMC’s 5nm node is similar to Intel’s 7nm in terms of density and power efficiency, this shouldn’t be too hard. It’ll relieve some stress from Intel’s foundries so they can focus on bringing the 7nm to the mainstream CPU market, all the while having a competitive Xe lineup.