The Iris Xe MAX is a totally discrete GPU constructed on Intel’s 10 nm SuperFin silicon fabrication course of. It options an LPDDR4X devoted reminiscence interface with Four GB of reminiscence at 68 GB/s of bandwidth, and makes use of PCI-Express 4.zero x4 to speak to the processor, however these are simply the bodily layers. On high of those are what Intel calls Deep Link, an all encompassing {hardware} abstraction layer that not solely allows specific multi-GPU with the Xe LP iGPU of “Tiger Lake” processors, but additionally sure implicit multi-GPU capabilities equivalent to fine-grained division of labor between the dGPU and iGPU to make sure that the correct of workload is cut up between the 2. Intel referred to this as GameDev Boost, and we detailed it in an older article.
Deep Link goes past the 3D graphics rendering area, and in addition gives augmentation of the Xe Media Multi-Format Encoders of the iGPU and dGPU to linearly scale video encoding efficiency. Intel claims {that a} Xe iGPU+dGPU mix provides greater than double the encoding efficiency of NVENC on a GeForce RTX 2080 graphics card. All that is doable as a result of a standard software program framework ties collectively the media encoding capabilities of the “Tiger Lake” CPU and Iris Xe MAX GPU that ensures the answer is greater than the sum of its components. Intel refers to this as Hyper Encode.
Deep Link additionally scales up AI deep-learning efficiency between “Tiger Lake” processors and the Xe MAX dGPU. This is as a result of the chip has a DLBoost DP4a accelerator. As of in the present day, Intel has onboarded main manufacturers within the media encoding software program ecosystem to help Deep Link—Hand Brake, OBS, XSplit, Topaz Gigapixel AI, Huya, Joyy, and many others., and is working with Blender, Cyberlink, Fluendo, and Magix for full help within the coming months.
Under the hood, the Iris Xe MAX, as we talked about earlier, is constructed on the 10 nm SuperFin course of. This is a model new piece of silicon, and never a “Tiger Lake” with its CPU part disabled, as its specs would possibly in any other case counsel. It options 96 Xe execution items (EUs), translating to 768 programmable shaders. It additionally has 96 TMUs and 24 ROPs. It options an LPDDR4X reminiscence interface, which 68 GB/s of reminiscence bandwidth. The GPU is clocked at 1.65 GHz. It talks to “Tiger Lake” processors over a standard PCI-Express 4.zero x4 bus. Notebooks with Iris Xe MAX have their iGPUs and dGPUs enabled to leverage Deep Link.
Media and AI solely paint half the image, the opposite being gaming. Intel is taking a swing on the 1080p mainstream gaming section with the Iris Xe MAX providing over 30 FPS (playable) in AAA video games at 1080p. It trades blows with notebooks that use the NVIDIA GeForce MX450 discrete GPU. We reckon that the majority e-sports titles needs to be playable at over 45 FPS at 1080p. Over the approaching months, one ought to count on Intel and its ISVs to speculate extra in Game Boost, which ought to improve efficiency additional. The Xe LP structure options DirectX 12 help, together with Variable Rate Shading (tier-1).
But what about different cell platforms, and desktop, you ask? The Iris Xe MAX is debuting completely with thin-and-light notebooks based mostly on 11th Gen Core “Tiger Lake” processors, however Intel has plans to develop desktop add-in playing cards with Iris Xe MAX GPUs someday within the first half of 2021. We predict that if priced proper, this card might promote in droves…
