This site may earn affiliate commissions from the links on this folio. Terms of utilise.

An Avoton (Silvermont) wafer

For the past few months, details on the Goldmont architecture (that's the Atom microarchitecture refresh that replaced Silvermont this twelvemonth) have been extraordinarily scarce and hard to come past. Later doing full deep dives on past architectures, including the Silvermont update that powered Bay Trail and Ruby Trail devices, Intel refused to talk near Goldmont in whatever pregnant caste at all. While it's true that these chips are being relegated to the lower-cost Pentium and Celeron lines as office of the Apollo Lake platform,  Goldmont hardware volition all the same bulldoze millions of devices over the adjacent few years — and it'south been interesting to watch the development of Intel's small-core chips in relation to their large-cadre counterparts.

At present, thanks to Intel's updated x86-64 programming guides, nosotros've gotten a wait at what the new scrap tin exercise and how it differs from Silvermont, which showtime debuted in 2022. Goldmont incorporates a number of improvements over Silvermont, though some of the diagrams are a chip bare-basic compared with what Intel typically creates for consumer publications:

Goldmont-Diagram

Where Silvermont (and AMD's Kabini / Jaguar / Puma) were all dual-issue decoders, Goldmont has iii decoder units, and a maximum of twenty bytes decoded per cycle. The fetch and education cache pipelines are no longer coupled, big folio support have both been added, and there'southward a minor L2 "precode" cache (16K) that didn't be on prior Atom processors. Goldmont'due south triple-wide decoder is matched by its power to retire up to three instructions per cycle, and the chip is capable of executing one load and shop per clock bicycle (Silvermont could only perform one load or store per clock cycle). Three simple integer operations can exist executed per cycle and accost generation is now out-of-lodge in Goldmont (Silvermont generated and scheduled retention addresses in-society, but could complete them out-of-order.)

Goldmont-Silvermont

Goldmont also has generally improved instruction latencies (how much depends on the instructions in question, but some of the gains are considerable) and tin can decode a maximum of 2 branches per cycle (Silvermont was limited to ane). Overall, Goldmont is a much smaller proceeds over Silvermont than Silvermont is compared with the original Atom core, Bonnell — but it's simply fair to note that significantly less fourth dimension has passed betwixt the debut of Bonnell (2008) and Bay Trail (2013) as compared to Silvermont (2013) and Goldmont (2016). Intel also hasn't gone through most as many procedure node transitions. Bonnell was a 45nm core compared with Silvermont'due south 22nm, whereas Goldmont is a 14nm chip.

Reading the tea leaves

The Goldmont-Silvermont shift is conceptually similar to the upgrades AMD made to its 40nm "Bobcat" CPU when it congenital the 28nm Jaguar follow-upwardly. While the ii companies made unlike changes to their underlying architectures, in both cases, Intel and AMD chose to enhance and upgrade the basic designs they'd previously deployed rather than making a huge prepare of changes or taking a dramatic bound forward.

Earlier this year, I speculated that Intel may have been forced to pull Goldmont CPU clocks downwards compared with Silvermont, because the CPU performed more than work per bicycle and had more difficulty hitting high frequencies within previous TDP ranges as a result. This could still be true — mostly speaking, we'd await Goldmont to be at to the lowest degree modestly faster than Silvermont on a clock-for-clock basis, and higher efficiency CPU architectures often use more power at the same clock speeds than lower-efficiency cores.

How much whatever of these issues specifically impacted Goldmont, or whether the new architecture's performance influenced Intel's conclusion to kill Atom's smartphone and tablet hardware divisions is yet open up to discussion. Checking Intel's diverse Cantlet pages, it's clear that "Atom" is being phased out as a divide brand — new chips haven't been launched in the desktop or conventional mobile markets for years, and there are only a handful of rebranded 28nm Rockchip designs on the Smartphone and Tablet Ark page. That said, Intel is offer a 1.2GHz 28nm flake equally part of its Rockchip understanding, where it previously topped out at 1.1GHz. Presumably it created the college-cease SKU for a reason, but whether or not the company is shipping any volume on these parts at all is an open question. It probably isn't.

As for whether the Goldmont architecture has a time to come past 14nm or not, Intel really hasn't said. On the one hand, having a low-finish Pentium and Celeron-form core bachelor lets Intel position those systems as meaningfully different (and much less expensive) than Core i3/i5/i7, and information technology may proceed to design Goldmont-derived CPUs for that reason alone. On the other, nonetheless, both Intel and AMD have moved away from multi-CPU architectures as they've abandoned the tablet and smartphone market. It may exist that in the time to come, Intel will simply address this space with derivatives from Cadre chips, but equally it used to practise before Cantlet came on the scene dorsum in 2008.