A couple of alleged Intel CPU roadmaps have leaked, revealing the supposed plans for Intel CPU releases all the way through to Q4 2021. The leak comes from Tweakers.net, who claim these slides were shown in a Dell presentation earlier in 2019.
As ever, we wouldn’t just go assuming these roadmaps are true, or correct, as they’re the sort of thing that’s pretty easy to fake. However, based on the information in them, if true, then Intel’s rocky patch could be set to continue.
We’ll start with Intel’s commercial desktop CPU roadmap. These commercial roadmaps generally lag slightly behind traditional CPU SKUs so we should be able to get our hands on the new chips earlier than the roadmap suggests.
The top couple of rows on the Intel CPU Roadmap are probably of most interest, with the other rows consisting of low-power chips, entry-level server parts, etc. Up top, we see Comet Lake H/G and Comet Lake U CPUs are due to arrive in Q2 2020. Comet Lake H will offer 8 & 10-core variants, while Comet Lake U will cover more budget offerings with 2/4/6/8 cores. However, the big sticking point here is that all of these Comet Lake parts are due to be manufactured on the 14nm process, the same process Intel has been stuck on since Broadwell debuted all the way back in 2014.
We know Intel has been having some serious troubles with 10nm fabrication, particularly in terms of yield, and according to this roadmap, we may have to wait until mid-2021 before we get the first commercial 10nm parts from Intel. These aren’t even the enthusiast parts either, with Tiger Lake U & Y offering quad-core variants that have the whiff of budget parts to them. Their dates are also ‘TBD’, suggesting they aren’t locked in stone yet. Judging from the absence of enthusiast desktop chips on the roadmap, we may even have to wait until 2022 until Intel can bring 10nm to market.
Lastly, Rocket Lake is also pegged for release in Q2 2021, offering 4/6 Core CPUs still using an optimised 14nm++ process. However, they will also allegedly feature in-built 14nm graphics chips.
For Intel, this all sounds a little problematic. Team Blue could be relying on 14nm optimisation all the way through to the tail end of 14nm for their desktop processors, which isn’t a great look when their main competitor is mighty close to bringing 7nm to market with Zen 2.
Intel’s official word on 10nm parts has been a late 2019 release, after numerous delays. You can forget about desktop parts though. Intel is still on track to launch 10nm chips in 2019, but it’s the mobile CPUs which stand to benefit first.
The first roll-out of Ice Lake U-series 10nm chips are going to be in limited supply but are on course to arrive later this year, as is the ultra-low power Ice Lake Y dual-core. From Q1 2020 the 10nm rollout begins in earned though, with Tiger Lake Y offering four-cores with around 5W TDP, Tiger Lake U providing 4C @ 15-28W, and the first 10nm graphics chip from Intel packaged with Rocket Lake U in Q2 2020.
There's still a big question mark over whether these roadmaps are actually legit or not, but they certainly make a great deal of sense. Intel is supply constrained on 14nm as it is, let alone 10nm with its low yields. Sticking to 14nm for the foreseeable future seems to be the safer bet, at the risk of some performance stagnation.
Making the mobile parts the priority is an obvious choice as well. Like it or not, Intel probably does serious volumes of its low-end parts and they're also the CPUs that stand to benefit most obviously from the improved power efficiency of 10nm.
For AMD, this represents a golden opportunity to wrestle back control from Intel, and these sorts of opportunities don't come around often. If AMD can make 7nm Zen 2 stick, and then continue to optimise it efficiently, they could deliver quite the blow to Intel's dominance.