In 2018, Intel had promised that it would release its first 3D card for PC in 2020. It’s done, here is the Xe MAX, the first official consumer graphics chip from the founder of Core processors. It is tailor-made to fit under the keys of our thin, ultra-thin and hybrid laptop keyboards.
It is therefore no coincidence that the first machines to take advantage of it are the Acer Swift 3X presented a few weeks ago; the Asus VivoBook TP470 and the Inspiron 15 Series 7000 2-in-1 from Dell.
All will soon be commercially available and are, for the first, a 14-inch ultraportable PC and, for the other two, convertible hybrid laptops with 14 and 15 inch screens.
A variation of the Intel Xe MAX will also be available in desktop PCs, but Intel was not very talkative on the subject during our interview with its engineers. On the other hand, where all are inexhaustible is when you ask them what you can really expect from the Xe MAX.
To summarize : this chip has enough strengths in its game to allow you to create excessively and of play in full hd. And it is able, in addition, to evolve in duo with the graphics part integrated into the processor to further increase its efficiency in many scenarios of use.
Intel Xe MAX is just the first
We are not going to go into too technical considerations here, but some foundations must nevertheless be laid to fully understand how Intel Xe MAX works. We will specify in the preamble that it is the least powerful of the dedicated 3D chips that the founder has decided to launch on the market.
The Intel Xe MAX is made to work with the latest Intel Tiger Lake processors, or Core of 11e laptop generation. And only those and future generations. Not earlier. Why ? Simply because the DNA enters the graphics controller of Core 11e generation (Intel Iris Xe) and the 3d part is the same.
These two circuits share many points in common and can therefore evolve in the same “waters” (Common Software Frameworks Intel) without shorting out. So :
- they have the same type of graphics units (the Execution Units)
- the same “hardware” encoders / decoders (the Xe Media Encoders)
- and the same component responsible for processing information relating to the AI (the DP4a)
This duo communicates through a technology (and PCIe 4.0 lanes) called Deep Link. Under this name are in fact hidden three operating mechanisms: Dynamic Power Share, Additive Ai and Hyper Encoding.
The two components can dynamically transmit power. The CPU can thus benefit from an increase in electric current (and therefore be more efficient) when applications do not need the power of the GPU which allows its partner to benefit. This did not fail to remind us of the Kaby Lake-G chips created in partnership with AMD.
Intel assured us that a PC with an Intel Xe MAX platform and a Tiger Lake processor would be 40% more efficient than a combination 10th Generation Intel Core (1065G7) and Nvidia MX350 chip in Adobe Lightroom (using the Image Detail Enhancement filter) for example.
There, the two circuits can act together to manage information or tasks and speed up their processing provided that the software used embeds the right pieces of Intel code.
Intel gave us a really convincing demonstration under Topaz Labs Gigapixel, in which the Intel platform was up to 7x faster than its competitor when it came to enlarging a set of 20 JPEG photos, of different sizes and dimensions, to 4x on the fly.
As its name suggests, it involves using the 4 hardware encoders that make up the two circuits to process video streams more quickly.
The test consists of converting a 4K (60 / AVC) video to 1080p (60 / AVC) format using Handbrake installed on a platform with a Tiger Lake processor, 16 GB of memory and an Intel Xe MAX. The Intel chip combo is 1.7x faster than a machine with a Core i9-10980HK, 32 GB of memory and an RTX 2080 Max-Q from Nvidia.
Create, yes, but also play
On the gaming side, Intel remains very measured in its graphics and does not seem to promise the impossible. Depending on the title, the frame rate per second would vary between 30 and 45, in Full HD with details set to Low or Medium. Scores that allow our duo to be – most of the time – more efficient than a platform equipped with GeForce MX350.
What is much more interesting is that depending on the game, the dedicated MAX graphics part is not always the one that will be used by the laptop. Sometimes the controller integrated into the Tiger Lake processor will do the trick better. You don’t have to worry about anything. The choice of the right chip will be done automatically, according to preprogrammed profiles by Intel within the graphics drivers.
As the visual above shows, when the game DoTA 2 running on a given platform, Intel obtains better results if the integrated Intel Iris Xe controller of the Tiger Lake processor is responsible for generating the graphics. MAX is even less swift than the MX350 in this game (which itself is less swift than the classic Xe).
On the other hand, to turn Metro Exodus, a title full of effects and complex polygons, it is the Xe MAX which is essential (42 fps on average against 29 for the MX350 and 34 for the Xe).
An Intel Xe MAX full of promises
We can’t wait to get our hands on one of the three machines mentioned above in order to be able to realize, for ourselves, the possibilities offered by the Intel Xe MAX, both in the contexts of digital creation and gaming. This will also allow us to assess the consumption of the Intel Xe MAX, a subject on which its designer was very silent during the presentation. There is however very likely this GPU is inspired by its 11th generation CPU cousin in this area … and this is not necessarily good news.
Even more powerful models are coming next year, specializing in gaming for some, in pure creation for others. GPUs that we can’t wait to come see and face the solutions RTX from Nvidia and the last AMD Radeon RX 6000.