At Least One of the Recent Intel Roadmap Rumors Is Likely Wrong
A new Intel roadmap has leaked, predicting a launch for a 10nm Intel “Alder Lake” as soon as late 2020. Normally, when I write up a rumor, it’s because I’ve already thought it over and decided it has a decent chance of being true. In this case, I’m doing something a little different. I want to talk about this rumor in conjunction with several others that have surfaced over the last six months.
The new rumor, according to THG, surfaced on Sina Weibo and consists of two slides:
Here’s how Tom’s Hardware describes the contents of the first slide: “the slide claims improved yields, significant increase in production capacity and a series of 10nm products launching in 2020. As for the 7nm lineup, the leak says the leading product will land in 2021, while the complete product portfolio can be expected in 2022. The information also suggests performance enhancements every year.” For comparison, here’s the original version of that slide that Intel released to the public:
The THG description of this slide could indicate that it has been updated since we saw it back in 2019, which is definitely something companies sometimes do — or that someone started from a known Intel slide to give their creation a veneer of authenticity and then created false content to overwrite what had previously been presented.
Other Recent Rumors
In the past six months, major Intel roadmap rumors have included the following:
- Intel has canceled all 10nm desktop CPUs. Desktop remains on 14nm through 2021.
- Intel will introduce a new 14nm desktop CPU architecture named Rocket Lake in late 2020, with backported 10nm features.
- Intel will launch 10nm desktop Alder Lake in an 8+8 big.Little configuration as an upcoming follow-up to LGA 1200 CPUs. Implied 2021 launch date.
Now, let’s add in a few things that we actually can say for certain. We know that Intel has a follow-up architecture to Ice Lake, dubbed Tiger Lake, coming on 10nm. We know that after Ice Lake’s Sunny Cove comes Willow Cove (assumed for Tiger Lake) and then Golden Cove. Alder Lake could be based on either Willow or Golden in theory and use low-power Atom cores (Tremont or Gracemont) for the “little” chips.
If the first scenario is true, the other two obviously aren’t. If the second scenario (Intel backporting 10nm features) is true, the third probably isn’t. Why would Intel launch Rocket Lake after taking the enormous trouble to backport its capabilities, only to turn around and launch a “true” 10nm desktop CPU at exactly the same time? It wouldn’t.
Intel told the press at its 2018 Technology Day that it had already backported certain AVX-512 features to 14nm, but it never said this was the only time it would ever do so. It also has said nothing about 10nm desktop CPUs — not last fall, not this spring. When you look at Intel’s presentations, the discussions of 10nm that mention platforms talk about mobile and server. Does this mean Intel isn’t launching 10nm desktop chips? No. But it means Intel isn’t talking about launching desktop 10nm chips, at least not yet, and it normally would be by now. And yet, we also know Intel recently canceled most of Cooper Lake, which implies the company is seeing better 10nm CPU yields than expected — exactly as this slide reportedly states.
After Broadwell, Intel said it wouldn’t skip a platform with a node again, but that was several CEOs ago and a different set of circumstances. When you look at these three rumors, I’d argue we have the most support for #1 and #2. These rumors also do not contradict each other. This third rumor refers to Intel’s Snow Ridge 5G base station on the first slide, a device that’s already launched, and it makes some significant promises around future Intel products. It’s possible that the slide is genuine, but was created at an earlier point in time when Intel had a different idea of what its roadmap would look like. Sometimes, the reason rumors appear to be false is because they were true when the leaks occurred, but were later canceled.
Some of you may remember rumors that AMD would develop a super-APU with a high-clocked Piledriver-derived CPU and a wide desktop GPU for graphics, all connected to a unified pool of HBM, in a socket capable of withstanding the power draw. Sources inside AMD confirmed that this was a real product the company was considering building, but ultimately abandoned. The rumors about Tarnhelm weren’t false; they just didn’t come true. This rumor could certainly still be true, but given that it contradicts earlier rumors, we know that some of the information in circulation about Intel’s upcoming plans is incorrect. Salt accordingly.