The long-awaited Intel stand-alone graphics card has finally been released! The 1998 i740 was a flash in the pan, and the 2008 Larrabee, who had not made it out of the division, has finally returned as king. This time Intel has developed a new Xe architecture, divided into Xe LP, Xe HPG, Xe HP, Xe HPC four micro-architecture, covering everything from thin notebooks to high-performance computing, is no longer a simple “display card.”
The previously released Tiger Lake 11Th generation Core Mobile version already integrates a core graphics card with Xe LP’s ultra-low power architecture called Iris Xe.
Today’s release of stand-alone graphics cards is still based on the Xe LP architecture, and is divided into mobile and desktop versions, both called Iris Xe MAX, with the addition of the new Deep Link technology, Acer, Asus, Dell notebook debut, where Acer has been pre-launched.
First of all, we are most concerned about the specifications and parameters. Iris Xe MAX Mobile has a maximum of 96 execution units (768 stream processors), 48 texture units, 24 ROP grating units (with a low distribution but not published), and a core frequency of 1.35GHz.
The desktop version is fixed full of all 96 execution units, the core frequency increased to 1.65GHz, with 4GB LPDDR4X independent memory, bandwidth of 68GB/s – simple calculation can be known bit width of 128-bit, memory frequency 4266MHz.
Technically, the Iris Xe MAX manufacturing process is as 10nm SuperFin as Tiger Lake, with SUPPORT for DX12.1, OpenGL 4.6, OpenCL 2.0, and PCIe for graphics 4.0 bus, VRS variable refresh rate, adaptive synchronization, synchronous computing (testing), DLBoost DP4A AI acceleration instructions, AV1 video encoding, 12-bit end-to-end data, video output support eDP 1.4b, DP 1.4, HDMI 2.0b, up to four screens.
Deep Link is a new breakthrough and a reflection of Intel’s platform capabilities. It brings together multiple processing engines through a common software framework, providing new capabilities to maximize performance and enhance AI AI creation performance and video encoding capabilities.
The most special thing is that it allows applications to work together across core graphics and stand-alone graphics cards in the Xe architecture, i.e. in parallel with the dual GPU.
Additive AI capabilities, for example, support inference and rendering on both GPUs at the same time, speeding up content creation workloads. Official tests show that the combination of the 11th-generation Core i7-1165G and Iris Xe MAX provides up to seven times more AI performance than the combination of the 10th-generation Core i7-1065G7 and MX350 graphics cards.
For example, through the integrated video coding engine in each GPU, the combination of super video encoding, saving rendering time, can be previewed or shared, official tests show that the 11th generation Core i7-1165G7, Iris Xe MAX combination, compared to the 9th generation Core i9-10980HK, RTX 2080 Super Max-Q combination, coding speed increased by up to 1.78 times.
In addition, due to the common hardware and software architecture, Xe core graphics cards and stand-alone graphics cards also support dynamic power sharing (Dynamic Power Share), mainly for the 28W high-end version of the 11th generation Core, one side of the free time can intelligently allocate power space to the other side, official data said can improve performance by an additional 20-40%.
It’s a bit like collaboration between an AMD Ryron processor and a Radeon graphics card, and that’s the benefit of a platform.
Of course, Deep Link’s implementation naturally requires the optimization support of developers, for which Intel offers a cross-architecture one API tool suite, a media software development suite, an OpenVINO tool suite distribution, and a VTune Profiler.
Software vendors that are currently optimized for Deep Link include Little Hulu (the country’s first online celebrity data platform), Tiger Tooth, Huan gathered YY, Hand Brake, XSplit, OpenStation, Topaz Gigapixel AI, and so on, while Blender, CyberLink, Magix, Fluendo, and more are supported.
In terms of game performance, Iris Xe MAX enables thin and thin books to be stable and smooth at 1080p resolution, surpassing the NVIDIA MX350 in comprehensive performance, as well as graphics post-processing technology Game Sharpening and game optimization technology Instant Game Tuning in the Drive Control Center.
The first Iris Xe MAX mobile stand-alone graphics notebook has three, of which Acer’s extraordinary S3X (SF314) is already on sale, i5/16G/512G 4499 yuan, i7/16G/512G 5299 yuan, AsusVivoBook TP470, Dell Inspiron 15 7000, will follow later.
The Iris Xe MAX desktop standalvetop graphics card will be available early next year, with OEMs saying it will be available only initially.
Server GPUs, also based on Xe LP architectures, will be delivered this year, while products for Xe HP and Xe HPG high-performance architectures will be available in 2021.
More official tours
Acer Extraordinary S3X (SF314)
Asus VivoBook Flip TP47
Dell Inspiron 15 7000