Among the many announcements in Intel Architecture Day today, Intel is also offering a major update to their GPU roadmap over the next 24 months. The Xe family, already packed with blockchains with Xe-LP, Xe-HP and Xe-HPC parts, is now getting a fourth planned variant: Xe-HPG. Directly aimed at the gamer enthusiast market, this latest Xe variant will be the most focused part of Intel games yet, and the biggest step still in Intel’s plans to diversify into its founding resources.
What is Xe-HPG? At a high level, it is meant to be the missing piece of the puzzle on the shelves of Intel products, offering a high-performance gaming chip and focused graphics. This is in contrast to Xe-HP, which specializes in data features like FP64 and multi-tile scalability, and Xe-HPC which is even more esoteric. In this regard, the Xe-HPG can be thought of as everything in the Xe family, distilled down into a single design to push FLOPs, rays, pixels and anything else that might need a powerful video card.
As with Intel’s upcoming Xe releases, the company is not offering performance projections, features or the like. But we have some small details of what to expect.
First and foremost, beyond going beyond the enthusiastic performance space, Intel has confirmed that this part will support beam tracking. A characteristic feature of high-end video cards, beam tracking will take on even greater importance in the coming years, as next-generation fast-paced consoles will launch the door with this feature as well, transforming eventually it into an initial feature across all gaming platforms. Similarly, beam tracking is a critical component of Microsoft’s DirectX 12 Ultimate standard, which given the timing of this GPU and Intel goals, I would be shocked if Intel did not fully support it.
Theipi will be built on the foundation which is Xe-LP. However, it will attract technologies that Intel is pioneering for Xe-HP, and Xe-HPC. No less important is the raw scalability, which is able to take the Xe-LP foundation and scale it up to hundreds (if not thousands) of GPU execution units. But Intel is pulling what they are calling “frequency calculation improvements” from Xe-HPC, which will likely allow them to maximize overall chip hours. All told, I would not be surprised if it looks a lot like Xe-HP in general, except with server-driven features like fast FP64 support and lots of removable boards.
But Xe-HPG will also bring something new to the table for the entire Xe family: GDDR6 support. Intel is confirming that the chip – or rather, the microarchitecture on which the chip will be based – will be designed to work with GDRR6. This is in contrast to Xe-HP (C), which uses HBM as part of high-end servers, and Xe-LP, which is designed to be used with more common types of memory. GDDR6 compatibility is a unique need that reflects this is a game-focused part: GDDR6 provides the bandwidth of memory needed for high-performance graphics, but without the stratospheric costs of HBM memory (a problem that has affected some GPUs other high-level over the years). In a further twist, Intel apparently licensed the GDDR IP controller from outside the company, rather than developing it inside; so Xe-HPG will have a very visible external IP IP in it.
But perhaps most interesting of all for graphic insiders and Intel investors is also where the Xe-HPG will be built: not Intel. As part of the Architecture Day guide, Intel has confirmed that the part will be done in an external model. It is actually the only Xe part where the GPU (or at least the computing element) is being made entirely on a third party model. Intel obviously won’t reveal which fab this is – whether it’s TSMC or Samsung – but that means we’ll see a full Intel GPU built into another model. If nothing else, this will make comparing Xe-HPG to its rivals AMD and NVIDIA much easier, as Intel will use the same fab resources.
Looking at the same roadmap, it is worth noting that Intel will not use any of their advanced packaging technologies for the part. Since they are not using HBM and are not doing a lot of tiling, there is no need for things like EMIB, never forget Foveros. There is still a lot of unfamiliarity with the cost aspect of Intel’s advanced packaging technologies, so keeping it from Xe-HPG will likely help keep costs in a highly competitive market.
And this is the trough in Xe-HPG. The latest and most focused gaming console for the Intel Xe GPU is set to release in 2021 – and while Intel looks set to explode in the wider GPU market, I have no doubt that it will not last life you will hear about it between then and now.