Intel Formally Announces Iris Xe MAX Graphics, Deep Link

Laptop vendors recently disclosed "Xe MAX" graphics as discrete Intel graphics set to appear within laptops in the coming weeks. That announcement was a bit unexpected and Intel did not brief the media in advance while today -- in an unusual announcement for a Saturday (Intel says it's timed for system availability, seemingly first in China) -- the company is formally announcing Iris Xe MAX.

Iris Xe MAX graphics are based on Xe-LP as found in Tiger Lake while it's the first Xe-based discrete GPU to market. Xe MAX will be found in various laptops over Q4. Interestingly, Intel did also disclose they will be bringing Xe-LP-based discrete graphics to "value desktops" in the first half of 2021.

Iris Xe MAX graphics feature 96 execution units, a 1.65GHz clock frequency, and 4GB of video memory while the rest is in common with the Xe-LP Gen12 graphics found currently with Tiger Lake.

Besides the Xe MAX hardware, Intel is also talking up "Deep Link" as a new software framework across their CPUs and GPUs. Deep Link provides a framework that can encompass the multiple processing engines in a unified manner. From the limited information disclosed thus far, basically it allows exploiting the media engines and execution units of both the discrete Xe MAX and integrated Xe hardware in a unified and concurrent manner.

Intel is preparing Deep Link for OBS, HandBrake, and other software while moving forward they expect it coming soon to Blender as well as software from CyberLink, Fluendo, and Magix. When inquiring more about the Linux support for Intel Deep Link given that I hadn't seen "Deep Link" mentioned in their various source repositories to date I was told it will be supported with VTune and OpenVINO. In the coming weeks they will have more news on the Linux software support.

With Xe MAX and Deep Link Intel discloses "proof of concept" data that in H1'2021 they have a "2x protected performance over RTX 2080" for video encode. The RTX 2080 SUPER is with NVIDIA's NVENC proprietary Video Codec SDK while they show with current Xe Graphics to be slightly ahead but once leveraging Deep Link they project it to be around twice the performance of the RTX 2080 video encode capabilities.

In their Windows gaming benchmarks, Intel puts the Iris Xe MAX as competing and sometimes outperforming the NVIDIA GeForce MX350.

The launch models for Xe MAX include the Acer Swift 3x, ASUS VivoBook Flip TP470, and Dell Inspiron 15 7000 2-in-1 models. These laptops with Intel Tiger Lake CPUs and Xe MAX graphics are expected to be available this quarter -- or as soon as today reportedly in some markets.

Given the great performance of Gen12 Xe Graphics on Linux for OpenCL / oneAPI Level Zero / Vulkan, we are certainly excited about the prospects of Xe MAX graphics appearing in more laptops and ultimately in desktops starting next year. From our Tiger Lake testing thus far there is long overdue substantial uplift over Gen9 (and Gen11) graphics. The Linux driver support for Gen12 Xe Graphics have been spot-on and continues seeing new optimizations in Mesa and the Linux kernel as well. It will also be interesting to see what comes of the open-source/Linux support of Deep Link moving forward.