Intel's next-gen DG2: 7nm node by TSMC, not Intel's in-house 7nm

Intel is still a while away from entering the discrete GPU battle with its DG1 graphics card, with the first-gen Xe GPU architecture demoed back at CES 2020 earlier this month with the Xe Low Power 'DG1 GPU'. I've also got some rumored specs on NVIDIA's new GeForce RTX 3080 and GeForce RTX 3070 graphics card here.

The first Intel Xe based discrete graphics card was always going to be a low-end part, but the second one is going to pack a much more powerful GPU punch. Intel's high-end DG2 GPU is now rumored to be packing the second-gen Xe GPU architecture, but interestingly be made on TSMC's own 7nm node -- not Intel's in-house 7nm.

Intel's new DG2 GPU would arrive in 2022, with the company using TSMC's 7nm process node over Intel's in-house 7nm EUV tech. This would be a move made after the continuous tripping over Intel has experienced with its troubled 10nm node, and Intel does not need a repeat of that mess as it's just getting on its feet in the GPU market.

But things get even more interesting as TSMC has its own 7nm+ EUV process which entered volume production in Q2 2019. Intel, at least according to these new rumors, won't be using that -- and it could be cost associated. TSMC's fresh new 7nm+ EUV node costs more, and then the supply issues with a first-gen node are something Intel would want to avoid.

NVIDIA will have its next-gen Ampere GPU on the market by then, as Ampere launches this year -- while AMD will be launching its Big Navi GPU this year, too. We should expect another GPU architecture from each company before Intel gets to its second-gen Xe GPU, with things really beginning to ramp up this year.

Intel will be detailing its new Xe GPU architecture at the Game Developers Conference in March 2020.