CPUsNews

Intel’s Meteor Lake CPUs to Use GPU Chiplets Fabbed on TSMC’s 3nm Node, SoC Die on N4 [Rumor]

Intel’s Alder Lake processors mark a major shift in the chipmaker’s design philosophy. For the first time, we’re seeing a hybrid core architecture across the board from the world’s largest semiconductor company. The 12th Gen Core CPUs also adopt DDR5 memory and the PCIe Gen 5 interface for the fastest I/O capabilities. While Alder Lake represents a momentous change in Intel’s CPU architecture design, the 14th Gen Meteor Lake lineup will mark the most drastic change in its packaging and manufacturing floorplan.

The company plans to shift from a monolithic to an MCM or chiplet based design consisting of multiple dies or “tiles” on the same substrate. Sapphire Rapids-SP will be the first lineup to adopt this strategy (with four tiles) in the server space, with Meteor Lake following it up in the client segment. According to a report from Commerical Times, Intel will be leveraging multiple foundries and process nodes to power its Meteor Lake processors.

The compute tiles (CPU cores and logic) will be fabbed on the Intel 4 node (7nm), and the GPU tile with up to 192 EUs will be manufactured on TSMC’s 3nm (N3) process. Furthermore, the SoC or the I/O die will also be fabbed by TSMC, albeit on the more mature 4nm (N4) node. Given that TSMC’s 3nm node won’t enter mass production before the second half of 2022, this means that Meteor Lake won’t be launching before H2 2023 (possibly even early 2024).

The recently announced Xe-HPG “Alchemist” graphics cards are the first major family from Intel to rely on a third-party foundry from the get-go. The ARC GPUs will leverage TSMC’s 6nm (N6) node to compete with AMD and NVIDIA’s latest offerings on even ground. They’re slated to launch in the first half of 2022. With both AMD and NVIDIA launching their next-gen graphics cards by late 2022, Intel won’t get much time to prepare its 2nd generation of ARC graphics cards. These, like their competitors, will almost certainly be fabbed on TSMC’s 5nm (N5) process and land sometime in the first half of 2023.

In addition to Meteor Lake and Alchemist, Intel’s Xe-HPC mega-chip Ponte Vecchio will also utilize TSMC’s 5nm and 7nm process nodes to fab the compute and Xe-Link dies, respectively. In addition to a chiplet architecture, the 14th Gen Meteor Lake lineup will also take advantage of the 2nd Gen Foveros 3D stacking technology to minimize die area and power, while maximizing performance. Meteor Lake will feature the Redwood Cove as the performance cores and Crestmont cores as the efficiency cores. A major upgrade to both the core architectures as well as the packaging technologies should allow for a massive uplift in overall performance. It’s worth noting that AMD is already planning to launch its first consumer lineup with 3D stacked V-Cache in the first half of 2022 in the form of Zen 3D/Ryzen 5000 XT.

Via: Commerical Times

Areej

Computer Engineering dropout (3 years), writer, journalist, and amateur poet. I started my first technology blog, Techquila while in college to address my hardware passion. Although largely successful, it was a classic example of too many people trying out multiple different things but getting nothing done. Left in late 2019 and been working on Hardware Times ever since.
Back to top button