The danger of dabbling in Alchemy…

CAN INTEL FINALLY DO IT and produce a competitive graphics card? The answer should already be in hand. But the release of higher-performing versions of Intel’s Arc graphics cards is likely to be delayed. For how long Alchemist could be held back isn’t clear, but from a company with a dismal record for delivering dedicated graphics products, it hardly bodes well.

We’ve been here before, of course. The most recent Intel graphics debacle was the Larrabee GPU. Intel popped a cap in that project back in 2009. At the time, Intel said it wasn’t dead, merely that the first chip in the new family of Larrabee graphics architectures had fallen behind sufficiently for it to no longer be viable as a retail product.

But we all knew it was curtains. With Alchemist in mind, it’s scary to recall just how close Larrabee got to launch. It had been integrated into fully working prototype graphics boards, several of which inevitably ended up on eBay. Ominously, Larrabee was just a few months away from release, if only it had performed up to expectations.

Mercifully, the similarities between Larrabee and Alchemist only go so far. Larrabee failed in large part thanks to a radical design using a swarm of tiny x86 cores rather than conventional graphics shaders. Alchemist is a much more conventional graphics design, the basic competence of which is already proven in integrated form as found in Intel’s latest 12th Gen processors, even if it’s not yet been shown to scale up to a larger and much higher-performing dedicated GPU.

But from where I’m standing, two closely related issues threaten to bring Alchemist down, and maybe take the whole Arc project with it. Firstly, Intel picked one heck of a time to enter the GPU market. All indications are that both AMD and Nvidia are set to at least double performance with their new graphics products, codenamed RDNA 3 and Lovelace respectively, when they arrive later this year. That kind of huge generational increase hasn’t happened in, well, pretty much a human generation. Meanwhile, the latest leaked info indicates the highest performing version of the first-gen Arc only has the grunt to take on the Nvidia RTX 3060.

Had Alchemist hit its launch window, which may have been intended for around this time last year, taking on the RTX 3060 would have been dandy. The reality, after all, is that the bulk of gamers aren’t buying more expensive boards like the RTX 3070 and 3080, let alone the megabucks 3090. But as the months tick by, the new RTX 4060 and its AMD rival grow ever closer, narrowing Intel’s window of opportunity before the competition is suddenly twice as good.

The other issue is driver quality, which may be a major factor in Arc’s leisurely wander towards retail availability. It’s a tricky balance to strike, setting a timely launch against the PR impact of launching a product with patchy stability and compromised performance. But if Alchemist is delayed so long that it ends up sharing etailer shelf space with RTX 4060 rather than the RTX 3060, it’s positively toast.

Above: Intel’s Larrabee never made it to market as a graphics card.

Speaking of dismal records, on anything outside its core competence of high-performance x86 processors, Intel doesn’t seem to be able to get the job done. Its highest-profile failure has been getting into smartphones. But over the years, I’ve sat through countless demos involving subdivisions at Intel, from LCOS (Liquid Crystal on Silicon) projector tech to medical diagnostics, only for the project to fizzle, fold, or be flogged to the highest bidder.

And what is going on with Optane, the non-volatile memory that was somehow supposed to take over from both flash and RAM? It’s not entirely dead but hasn’t been the massive game-changer we were led to expect. Maybe I’m jumping the gun and Alchemist will be just fine. Maybe its successors Battlemage and Celestial, due out in 2023 and beyond, will take the fight to AMD RDNA 3 and Nvidia Lovelace across a broader swathe of the market. But I’m definitely beginning to get worried.

If Alchemist ends up sharing shelf space with RTX 4060 rather than RTX 3060, it’s toast.

Six raw 4K panels for breakfast, laced with extract of x86... Jeremy Laird eats and breathes PC technology.