Intel has announced the new brand name for its consumer GPU products as well as code names for future GPU microarchitectures. The Xe microarchitecture has a new codename: Alchemist. Following Alchemist, we’ll see the launch of Battlemage, Celestial, and Druid. Intel did not disclose whether future Druid customers will be required to demonstrate neutral alignments as per 5th Edition D&D rules. Future consumer GPU hardware and software will be collectively branded as “Intel Arc.” Intel writes: “The Arc brand will cover hardware, software, and services, and will span multiple hardware generations, with the first generation, based on the Xe HPG microarchitecture, code-named Alchemist (formerly known as DG2).”
There’s an interesting tidbit buried in Intel’s press release: “Alchemist, the first generation of Intel Arc products, will feature hardware-based ray tracing and artificial intelligence-driven super sampling.” This sounds like something akin to Nvidia’s DLSS. A leaked slide earlier this year referred to XeSS as a forthcoming Intel standard that would compete with DLSS:
Slide leaked by Moore’s Law is Dead
DLSS 1.0 was not a tremendous hit but Nvidia’s DLSS 2.0 has been much better received and a similar solution from Intel would be welcome. To date, AMD has not announced anything like DLSS. The company’s Fidelity FX Super Resolution sharpening technique does offer faster performance with a small image quality tradeoff when used at its top quality settings, but it is not backed by the AI model training Nvidia has done or that Intel presumably is doing.
It is not clear if the above is accurate or whether performance characteristics are expected to change between April 2021 when this slide was purportedly made and early 2022. If we assume that these figures are accurate, it means Intel’s top-end GPU would come in between the RTX 3070 and RTX 3080, but more towards the RTX 3070.
It’s been said that there are no bad GPUs, just bad prices, and I suspect that holds true here as well. It isn’t realistic to expect Intel to leapfrog AMD and Nvidia in its very first product. It’d be great if they did, but most companies do not become market leaders of a new segment with a single product launch.
Assuming that the market remains meaningfully constrained, the appearance of any additional GPU competition will be welcome. Both TSMC and Samsung currently fab top-end GPU chips and Intel’s GPUs will also be built at TSMC rather than on its own fabs — but the document above also suggests that Intel was targeting a $200 – $300 MSRP for midrange parts. Currently, Nvidia and AMD have both dragged GPU prices higher and effectively restricted ray tracing to GPUs priced at $400 or above at MSRP. If Intel hits the right price points, it could potentially undercut both companies — if Alchemist’s overall performance is high enough.
Earlier this year, when GPU prices were closer to ~3x MSRP, we wound up recommending users buy 7- to 8-year-old used GPUs if they needed a card urgently because the old AMD Hawaii family was the best deal one could find. At this point, we’d be thrilled to see GPUs from any company that held MSRP and ensured cards were available. One point that remains true in the current climate: The best GPU deal is the GPU you find at MSRP, period. We didn’t feel as though AMD’s 6600 XT was as well-positioned as it could have been, but the 6600 XT’s performance per dollar at $379 is vastly better than paying a 1.4x – 1.6x markup over MSRP for a more powerful card.
If the market cools down, Intel’s Alchemist might have a bit more of a fight on its hands. So long as demand is meaningfully outstripping supply, any GPU that doesn’t start fires and avoids sounding like a Dirt Devil has the potential to be a bona fide hit.
Now Read: