Intel showed the work of multi-adapter – bundles from the built-in GPU and discrete graphics card

Almost all modern Intel processors are equipped with an integrated graphics core. To date, the capabilities of such GPUs are sufficient for normal operation and even for launching some old or undemanding games.

If the laptop also has a discrete graphics card, the integrated one saves energy when performing simple tasks. But is it possible to get the integrated and discrete GPU to work together?

Intel has published presentation materials that were intended for the canceled GDC 2020 event. One of the presentations is called “Multi-Adapter with Integrated and Discrete GPUs” and is dedicated to the creation of a multi-adapter – an integrated GPU with a discrete one.

For example, Intel brought a bunch of Intel HD 530 and Radeon RX 480, which is a bit strange, because the company already has samples of its own video card DG1. Be that as it may, the question is still very interesting.

As part of the presentation, Intel talked about several scenarios for using a multi-adapter implemented using Direct3D 12. The first is the transfer by a discrete adapter to the built-in tasks for processing computational shaders, which allows you to free up additional discrete power for processing graphics. As an example, Intel showed a simulation of N body, consisting of 4 million particles. There are no comparative tests – Intel’s task was not to compare this approach with the usual one, but to demonstrate the capabilities. Separately, it is worth noting that the combination of Intel HD 530 and Radeon RX 480 was not the only one that was used by Intel. It appears in the document on the site, while on the video you can see another (Intel HD 630 and Radeon RX 580), and on the attached image – the third (Intel HD 630 and GeForce RTX 2080 Ti).

The second scenario is the joint rendering of images, for example, with frame rotation. But this option is pointless for bundles where a discrete GPU is much more productive than an integrated one.

The third is the performance of asynchronous loads. For example, tasks of artificial intelligence, rendering physics of objects, modeling particles and shadows, and so on.

Considering the fact that Intel is preparing this year to enter the market of discrete graphics cards, and the first one will obviously not be particularly productive, we can assume that with the release of such 3D-cards Intel will introduce multi-adapter technology.

We can separately recall that AMD previously had a technology that allowed combining a discrete Radeon graphics card with an integrated GPU in a hybrid processor, but now it is not used. In addition, Intel’s approach implies the possibility of creating multi-adapters regardless of the GPU model.