Intel’s Next-Generation Xe-HPG Gaming GPUs To Utilize TSMC’s 6nm Process Node, Launching in 2021’s Discrete Graphics Lineup
Intel is launching its first gaming discrete graphics cards based on its next-generation Xe-HPG GPUs next year. The gaming lineup will feature a range of products but unlike the Xe-LP & Xe-HP GPU line which will be fabricated on Intel's own 10nm SuperFin process node, the Xe-HPG GPUs are going to be produced at an external foundry with TSMC being the most likely candidate.
TSMC To Win Intel's Xe-HPG GPU Orders, First Gaming Discrete Lineup To Be Produced on 6nm Process Node
Reports coming out of Taiwanese based outlets seem to indicate that TSMC is confident in winning orders for Intel's next-generation Xe-HPG GPUs. As reported by IThome (via @harukaze5719), Intel's Xe-HPG GPUs will be produced at TSMC's fabs, making use of their bleeding-edge fabrication technologies. The Taiwan-based sources claim that Intel is likely to utilize TSMC's 6nm process node for its gaming GPUs that will be competing against AMD's RDNA 2 and NVIDIA's Ampere GPU lineup by the time they hit the store shelves.
The TSMC 6nm process node was unveiled last year in the company's latest roadmap. Codenamed as 'N6', the TSMC 6nm process will make use of an advanced version of the EUV lithography tech. It will offer an 18% increased in logic density over TSMC's 7nm (N7) process and is compatible with 7nm design which is stated to shorten the time its takes to market consumer products.
The Xe-HPG class gaming GPUs will also be utilizing a standard packaging design that should be relatively easy to produce while datacenter and HPC chips are going to make use of a mix of advanced packaging technologies that Intel has innovated over the years such as EMIB, CO-EMIB, and FOVEROS.
Here's Everything We Know About Intel's Discrete Gaming Graphics Cards Powered by Xe-HPG GPUs
Intel's gaming graphics card lineup will be powered by the Xe-HPG GPU. This specific GPU is another category within the Xe micro-architecture family. It falls between the Xe-LP and Xe-HP and is targeted primarily at the gaming audiences. The Xe-HPG GPU is expected to use a single tile and assuming that a single tile consists of 512 EUs, we are looking at up to 4096 cores on the flagship chip.
Based on our exclusive report and from what Intel has talked about in regards to its Ponte Vecchio chip, it looks like Intel is all onboard the MCM train with each chip consisting of several Xe GPU tiles that will be interconnected together to form a monster of a GPU. Here are the actual EU counts of Intel's various MCM-based Xe HP GPUs along with estimated core counts and TFLOPs (You can read more about these impressive performance figures which were presented by Intel over here) :
- Xe HP (12.5) 1-Tile GPU: 512 EU [Est: 4096 Cores, 10.6 TFLOPs 1.3 GHz, 150W]
- Xe HP (12.5) 2-Tile GPU: 1024 EUs [Est: 8192 Cores, 21.2 1.3 GHz, TFLOPs, 300W]
- Xe HP (12.5) 4-Tile GPU: 2048 EUs [Est: 16,384 Cores, 42.3 TFLOPs 1.3 GHz, 400W/500W]
Of course, Intel can opt for a higher EU count on its Xe-HPG gaming-specific GPUs but that remains to be seen until the final specifications are disclosed. As for the gaming-specific features, the Intel Xe-HPG powered graphics cards will feature hardware-accelerated ray tracing and feature GDDR6 memory to optimize performance and value whereas the Xe-HP lineup which is aimed at the data center market will opt for HBM memory.
Intel recently gave us the first demo of its Xe LP GPU inside its upcoming Tiger Lake CPUs which is proving to be a major leap in integrated graphics performance for Intel as seen in the demonstration. Expect more information on the Xe HP & Xe HPC GPUs in the coming months.