On the Intel Architecture Day 2021, Intel finally gave the details of its partnership with TSMC, despite their competition in the foundry business. At the event, Intel introduced two new x86 cores, two new data center SoCs and two discrete GPUs. Under the new brand Intel Arc, Intel unveiled its first-generation GPU, codenamed Alchemist, based on a variant of its Xe architecture designed for GPUs.
Built on TSMC’s 6nm process, the Alchemist graphic card represents Intel’s commitment to challenge the duopoly of AMD and Nvidia in the PC gaming sector, and will be launched in Q1 2022. Following Alchemist, the later generations of Arc GPUs will be named Battlemage, Celestial, and Druid.
Commonly known as DG2, the Xe-HPG (High Performance Gaming) architecture exists alongside members of the Xe family used for data centers and high performance computing, including Xe-LP, Xe-HP, and Xe-HPC. Intel has never hidden its ambition to compete against Nvidia and AMD in the GPU market, but it was previously assumed that it would only take place in the server market.
However, Intel did unveil another GPU based on its Xe-HPC architecture. Named “Ponte Vecchio”, the chip leverages advanced packaging and incorporates 47 chiplets, some of which will use TSMC’s 7nm and 5nm process nodes, despite using an Intel-manufactured base.
With AMD as TSMC’s 2nd largest customer, and Nvidia allegedly switching back to TSMC for its GeForce RTX 40 series GPUs, used for gaming, as well as its Hooper series GPUs that target data centers, TSMC is likely the key to win on the GPU battlefield.