Among the many announcements on today’s Intel Architecture Day, Intel is also offering a major update to its GPU routing card over the next 24 months. The Xe family, already packaged with Xe-LP, Xe-HP, and Xe-HPC components, now gets a fourth planned variant: Xe-HPG. Targeted directly at the enthusiastic gamer market, this latest Xe variant will be the most gaming-oriented part of Intel yet, and the biggest step so far in Intel’s plans to diversify more into its urban resources.
So what is Xe-HPG? At a high level, it is intended as the missing piece of the puzzle in Intel’s product stack, offering high performance gaming and graphics-oriented chips. This is in contrast to Xe-HP, which specializes in data center features such as FP64 and multi-tile scalability, and Xe-HPC which is even more esoteric. In that regard, Xe-HPG can be thought of as anything in the Xe family, distilled into one design to print FLOPs, rays, pixels, and everything else a powerful video card needs.
As with the rest of Intel’s advancing Xe announcements, the company does not offer performance projections, features like that. But we have some small details about what to expect.
First and foremost, beyond tracking the space for the enthusiastic performance, Intel has confirmed that this part will support ray traces. Ray tracking will take over even more importantly in the coming years, as next-generation consoles will also be launching the door with the feature, eventually turning it into a basic feature across all gaming platforms. Similarly, raytracing is a critical component of Microsoft’s DirectX 12 Ultimate standard, given the timing of these GPUs and Intel’s intentions, I would be shocked if Intel did not fully support them.
The chip will be built on the foundation that is Xe-LP. However, it will also attract technologies that Intel pioneered for Xe-HP, and Xe-HPC. Not the least of which is raw scalability, which can take the foundation of Xe-LP and scale it to hundreds (if not thousands) of GPU execution units. But Intel is also pulling what they call “computer frequency enhancements” from Xe-HPC, which will likely allow them to maximize the chip’s total clock speed. All told, I wouldn’t be surprised if it looks a lot like Xe-HP in general, except with server-powered features like fast FP64 support and multi-tiling stripped.
But Xe-HPG will also bring something new to the table for the entire Xe family: GDDR6 support. Intel confirms that the chip – or rather, the microarchitecture will be based on the chip – is designed to work with GDRR6. This is in contrast to Xe-HP (C), which uses HBM as high-end server components, and Xe-LP, which is designed for use with more conventional memory types. GDDR6 compatibility is a unique need that reflects that this is a gaming-oriented part: GDDR6 provides the memory bandwidth needed for high-performance graphics, but without the stratospheric cost of HBM memory (a problem that some other high-end GPUs face has affected over the years). In a further twist, Intel apparently dropped the GDDR controller IP license from outside the company, instead of developing it in-house; so that Xe-HPG will have a very notable external IP.
But perhaps most interesting of all for graphics insiders and Intel investors is where Xe-HPG will be built: not at Intel. As part of its roadmap on Architecture Day, Intel has confirmed that the part will be made on an external fab. In fact, it’s the only Xe part where the GPU (or at least the computer element) is fully made at a third party factory. Intel obviously won’t reveal what fab this is – if it’s TSMC or Samsung – but it does mean we’ll see a full Intel GPU built at another fab. If nothing else, this will make comparing Xe-HPG with its AMD and NVIDIA rivals a lot easier, as Intel will be using the same fab resources.
Looking at the same roadmap, it is worth noting that Intel will not use any of its advanced packaging technologies for the part. Since they do not use HBM and they do not multi-tile, there is no need for things like EMIB, never think of Foveros. There are still many unfamiliar with the cost aspects of Intel’s advanced packaging technologies, so keeping Xe-HPG will probably help keep an eye on the costs in a highly competitive market.
And that’s the scoop on Xe-HPG. The latest and most gaming-focused member of Intel’s Xe GPU product stack is set to launch in 2021 – and since Intel looks set to break into the broader GPU market, I do not doubt for a second that this is not the last will be I hear there between then and now.
