Recently, news dropped from a Taiwanese publication stating that the next-gen Intel Xe GPUs plan to utilize TMSC’s 6nm and 3nm process. This decision became crucial after the release of their Intel Xe DG1 graphics card. This massive architecture jump will most likely take place in 2021 with the next-gen Intel Xe GPUs.
While Intel has suffered from the production of 10nm and 14nm, this may be a game-changer for the company. This is not the first time Intel and TMSC (Taiwan Semiconductor Manufacturing Company) have worked together. If we rewind back to 2014, both companies worked as a tag team on the Intel’s Sofia series mobile processors.
Before the company takes the jump, Intel’s Chief Financial Officer (CFO), George Davis made a public statement regarding the lack of capacity for the 14nm process. He said that the company will increase its production for the 14nm and make for the lack of capacity issue in 2020. Moving on, 2021 seems to be the year where Intel plans to utilize the 6nm process at first.
However, this news is still not 100% accurate as Intel’s transition to TSMC for graphics products has been a topic of discussion based on rumors before too. Intel’s DG1 graphics card will be holding its first GPU trials this year. The graphics card itself is based on the 10nm process and consumes almost 25W when it comes to power consumption. Further claims highlight DG1 to be performing better when compared to the NVidia MX250 iGPU.
The DG1 falls in the discrete form factor due to its similarities with the Intel TGL (Tiger Lake) processor. With only 96 EUs (execution units), Intel is still in testing and research state before it takes the leap. This approach is healthy for the company as it will gain more insight on GPU performance while studying rival products.
Later on, they can push further working on high variants of the Xe series. The CFO did mention that the 10nm aren’t competitive against the new architecture, therefore, the motive behind increased production capacity is still unknown.
The 10nm problem may prove some difficulty for GPU architects. Though 7nm will make things easier as Intel has plans for shifting to (EUV Extreme ultraviolet) which may help reset the complication chart. The 10nm seems to be difficult in terms of high volume production. This may be one of the main reasons why Intel is ditching TMG and shifting to TSMC.
While the road for Intel still holds challenges, the company confirmed the Ponte Vecchio GPU will utilize the 7nm process. With this said, we are not sure if Intel has flexible arrangements working on both TMG and TSMC or either all of this is just “bark and no bite”.