Intel’s client roadmap may face yet another slip-up as multiple sources claim a major delay in the deployment of the 20A node. A rebranding of the 4nm node, 20A is a client-exclusive node said to power the 15th Gen Arrow Lake family in 2024. It will be the first step in Intel’s multi-year plan to regain process leadership from TSMC. Unfortunately, “unquestioned leadership” might have to wait a few more years.
According to Golden Pig Upgrade, Intel’s 20A node has been delayed, leaving Arrow Lake with either the Intel 3 or TSMC’s 3nm node. Officially, Team Blue is already slated to leverage TSMC’s N3 node for the iGPU tile, while the CPU compute tile is supposed to be a 20A product.
I find it difficult to believe that Intel would dump both the CPU compute and iGPU tiles over to TSMC. A downgrade to the Intel 3 node (for the compute tile) is much more realistic. The Intel 3 process was originally planned exclusively for the Xeon Scalable Granite Rapids and Sierra Forest lineups.
In other news related to Arrow Lake, a rumor claims that the 15th Gen Core processors will ditch hyperthreading (SMT). This is quite unexpected as SMT has been a part of nearly every Core processor released in the last two decades. One of the primary disadvantages of hyperthreading is reduced single-threaded performance.
When running two logical threads off one physical core, the resources available to each (cache, registers, buffers) thread are reduced, resulting in lower single-threaded performance in certain workloads, including gaming. With the E-core count increased to 32, much of the multi-threaded grunt can be offloaded to the efficiency cores, leaving the P-cores with purely single-threaded work.
The 15th Gen Arrow Lake processors are slated to land sometime in the second half of 2024. They’ll replace the 14th Gen Raptor Lake-S Refresh on the desktop while forming the Core Ultra lineup in the mobility segment.
