The state of 4nm process node is starting to resemble the problems of Intel's 10nm chips
Intel is now trying to catch up to TSMC and regain the manufacturing process leadership that underpinned its past dominance. So perhaps the biggest question for years to come is whether it has already managed to “fix” its factories, or whether processor delays and missed targets will be a recurring theme in the future. Warning sing are appearing that the first 4nm Meteor Lake processors could possibly end up like the notorious 10nm chips.
Last fall and this January, Intel released Raptor Lake processors which should be the last in the 7nm generation. Intel’s 7nm process is an evolution of its previous 10nm technologies, which are very good today (in Tiger Lake, Alder Lake, Raptor Lake), but that only came after many years of tuning, while the first generations were underwhelming (Ice Lake) to practically broken (Cannon Lake). Coming after Raptor Lake are the Meteor Lake processors, and they just might hit the rocks again, as these will be using the truly new 4nm “Intel 4” process for the first time. In the earlier days when Intel’s notation scheme was ahead in numbering, it was referred to as 7nm.
And it’s already starting to look like bad news. There have been doubts before, and more recently reports that Meteor Lake will be delayed and not released this year as originally planned. According to youtuber Moore’s Law is Dead, there are some problems with Meteor Lake processors.
Earlier, they reported that Intel cancelled the most powerful versions of these processors, which were supposed to have 8+16 and 6+16 configurations (i.e. eight or six big P-Cores and 16 little E-Cores). After this move, there was reportedly still a 6+8 version left in the roadmap, i.e. six big, eight little cores, which would correspond to the current Core i5-13600K or i5-13600/13500 configurations. But that version couldn’t sell as the top gaming model and would probably only have to be marketed as Core i5 or i7, which wouldn’t be ideal for a new generation product.
Now Moore’s Law is Dead reports that Meteor Lake desktop processors may be dead altogether. On the Broken Silicon 192 podcast, they reported that according to one of their sources – an OEM manufacturer – Intel has apparently cancelled the desktop version. Rumor has it that it looks like Meteor Lake will only be for laptops. Or rather, probably only in the mobile U, P and H series. The HX series wouldn’t work out since that’s based on desktop chips, or in a possible 14th Core generation, the HX processors would probably still be 7nm, based on 8+16 Raptor Lake chips.
„At least as far as we (OEM) are concerned recently, Meteor Lake is Laptop Only […] we expect to be using Raptor Lake for at least another year in desktop…and probably even in Extreme Laptop [ed: likely the HX series]“
The mentioned PC manufacturer expects Raptor Lake (and thus the LGA 1700 socket) to have its product life in desktop PCs prolonged by another year, essentially replacing the unreleased Meteor Lake generation. Instead, Intel could release a rebranded Raptor Lake under the 14th Core generation designation, but it likely won’t be based on new chips. So the idea would be to increase the clock speeds and/or power draw of the top Core i9 models, while the lower models could have changed core counts within what today’s dies allow (for example, you could theoretically make a Core i7 with 8+12 cores and a Core i5 with 8+8 cores, or a Core i3 with 6+0 cores). Such a refresh would not bring new features and technologies, but it could have better performance.
According to Moore’s Law, the reason behind the setbacks or outright cancellation of the desktop version of Meteor Lake is that the team developing it is struggling and Intel doesn’t want to waste resources, preferring to prioritize maximum effort on the successful and as little delayed as possible launch of mobile Meteor Lake for laptops, where it is more worried about market share. But at the same time, it is also running into problems with the 4nm process, which may be another reason to shift resources to the mobile version. That segment doesn’t need such high clock speeds, so issues with reaching top clock speeds on the new node are less problematic for it. That’s why the 10th and 11th generation 10nm Core were only launched for laptops, and it’s possible that the 14th generation 4nm Core will now repeat that again.
This does not mean that the 4nm process and its derivatives (to be followed by Intel 3nm) and successors (Intel 20A, Intel 18A) will be completely hopeless. It’s possible that with further development Intel will make them a viable alternative again. But this could again come at the cost of a delay against the original plans, with TSMC releasing their new technologies in the meantime, and thus again leapfrogging Intel by a generation. If this situation, where Intel eventually succeeds in developing its own processes, but only perhaps a year or two after TSMC, becomes a reality, then it is essentially a defeat for Intel as that will not give the leadership competitive positioning the company seeks.
It should be added that youtuber-claimed information may not be fully reliable. Therefore, all of this should be taken with a grain of salt and it is probably not yet time to accept the cancellation of Meteor Lake as an established fact. Moore’s Law, for example, offers the possibility of some Meteor Lake desktop models being released in a limited form for the retail market, for example, where Intel would offer them to DIY PC builders, but not to the OEM market. This would be quite problematic, however, as the retail market is only a small percentage of the overall market. A retail-only product that didn’t target OEM builds would have significant problems, especially since Meteor Lake will require new motherboards.
LGA 1851 for the next generation?
Meteor Lake for desktop was supposed to use the new LGA 1851 socket, which has already appeared in schematics. This platform would lack a CPU counterpart with after the cancellation of desktop Meteor Lake, so the launch of these boards would probably be postponed and they would only come out with the second generation of processors bound for LGA 1851, which were supposed to be Arrow Lake processors. It’s possible that this would then be the last generation for this socket as well, unless some sort of Arrow Lake refresh comes out or another generation is added to the roadmap for this platform.
Sources: Moore’s Law is Dead (1, 2)
English translation and edit by Jozef Dudáš
⠀
