Friday, August 3rd 2018
Chances of Intel Going Fabless Higher Than Ever
Intel is one of the few semiconductor companies that manufactures a majority of its products on its own silicon fabrication foundries. The breadwinner for the company continues to be CPUs, and a majority of its revenues continue to come from its client-computing group (CCG). CPUs, like GPUs, are required to be built on the latest silicon fabrication process to keep up (or catch up) with Moore's Law. Intel is plagued with severe technological roadblocks toward advancing its foundry process from 14 nanometer (nm) to its next step, 10 nm. In its latest Q2-2018 earnings call, the company confirmed that the 10 nm node won't put out before Q4-2019, even as rival AMD's CEO announced that its first 7 nm processors will be up for purchase by the end of 2018 (a year ahead with a more advanced process, on paper). Analysts are beginning to paint a very grim future for Intel's foundry business.
The prospects for Intel going fabless, at least for its cutting-edge products, is higher than ever. Analysts, speaking with Taiwan-based industry observer DigiTimes, mentioned that there is speculation of Intel scaling down its foundry business. Something like this, if true, could hint at the company looking for foundry partners with newer silicon-fabrication nodes at a more advanced stage of development (eg: GlobalFoundries 7 nm) to manufacture its processors, while relegating its own foundries to manufacture less complex products such as chipset, NAND flash, 3D XPoint memory, 5G PHYs, etc. Fancy a Core processor made by GloFo in the great state of New York?
Source:
DigiTimes
The prospects for Intel going fabless, at least for its cutting-edge products, is higher than ever. Analysts, speaking with Taiwan-based industry observer DigiTimes, mentioned that there is speculation of Intel scaling down its foundry business. Something like this, if true, could hint at the company looking for foundry partners with newer silicon-fabrication nodes at a more advanced stage of development (eg: GlobalFoundries 7 nm) to manufacture its processors, while relegating its own foundries to manufacture less complex products such as chipset, NAND flash, 3D XPoint memory, 5G PHYs, etc. Fancy a Core processor made by GloFo in the great state of New York?
80 Comments on Chances of Intel Going Fabless Higher Than Ever
What they want, well they're not getting from the looks of it as things are now.
Yeah that's what they say, what we know about, they said they had to delay 10nm for those reasons, which seem pretty reasonable, they're not getting exactly what they wanted, so they have to keep refining until they reach that goal.
Intel is the one that has to learn how to make it work so they can then pass that along to the fabs to even start testing.Can't really argue with this except in this case it's not a matter of refinement since they don't have a working design to refine that we know of ATM.
This is a matter of wait and see - Our own opinions about it don't matter either way because we're not a factor in this at all. :rolleyes:
Getting it to do something is one thing regardless of what the "something" is, getting it to work as intended without all the issues is another.
They get that figured out then they can move things along.
The best process – no matter how oh so advanced it is going to be (on paper) – is worth exactly no·thing if it ain't working.
Then, even a (on paper inferior) node is superior, since it at least meets a pretty crucial condition. It's working (already).Only thing is, even if their 10nm will be working anytime soon™ (see above), they're still stuck with an µArch which ran out of steam already years ago … And which also needs to be patched up while loosing assumingly already even more performance in the process and has virtually no room for any improvement at all whatsoever.
Sure, that's purely speculative already, as of yet! Though, given the most recent disclosures, some of that might be already foreshadowings becoming executed …
Meanwhile AMD can capitalise upon not only at least (pretty much) equal process-technology (and reaches node-parity! again?) but sure as heck can even shift up some gear and drive innovations while improving a brand new µArch (and getting performance-improvements out of that department) too. Considering that they managed to archive pretty solid improvements with Ryzen-Refresh (Ryzen 2xxx) only with some changes on µCode-level (which already if not equaled, already out-paced some of Intel's lower refreshes on some optimised 14nm-process; gaining like what? 2,6% IPC-improvement with only minor changes on Ram-latency, sharper timings at the fabric et cetera?) while they assumingly didn't even touched any Core-design at all, well … It's pretty hard to believe that they won't surpass Intel's IPC or single-thread-performance when they finally beginning to lay their hands upon the very Core-design and the Fabric while fabbing those on a even smaller node (which brings already some solid performance-/efficiency-betterments on its own). Good luck, Intel!
@lexluthermiester It's pretty dry, isn't it?!
On a serious note, I don't know what those product names are supposed to be. About half a year ago, Ice Lake was supposed to be next in the desktop lineup, and then Whiskey Lake happened/is scheduled to happen. Ice Lake was supposed to have fixes for Spectre and Meltdown baked into the silicon (without a performance penalty), rather than relying on patches and updates, but between the time I read that and now, there's been quite a few new developments with these security holes, so I'm not even sure how much that's worth.
But if I had to take a guess, that's may be, just maybe, since TSMCs 7nm node is already working and ships AMD-products?
Sure, those aren't available for anybody except a few selected AMD-partners for evaluation, but still. There are no actual signs indicating any contrary condition.
It's assumingly already that good to deliver a complete working CCX which sums up to a 32-Core processor – while at the same time Intel isn't even able to fab a fully working 2+0 Die-config and has to fuse off the graphics-part completely (and with that half the Die-size of some Dual-Core of a size of merely 50mm²). And even then, such halfway fused-off Dual-Core a) doesn't even reach the same frequencies of Intel's 14nm process and b) consumes even way more power to do not even the same.
So, knowing that, it takes no wonder, why no-one considers it any appropriate to speak the same about TSMCs 7nm. It's already working completely and ships products and all tests ran completely beyond expectations (hence the preponement). That's why it seems completely legit what Lisa Su said already month ago …
There's no proof 7nm is ready and working besides their CEO's words, and i'm pretty sure that even if it is ready, they might be launching products on 7nm only a few months before intel will be able to launch their on 10nm.
Intel has been leading the industry for the last 10 years pretty much, it's not likely that GF/Samsung/TSMC will totally revert that in less than 2 years.
Just to be clear here, Intel have had to bring something on 10nm just for the shareholders (that infamous 2+0 Die being fused off halfly) so that those may be eventually pleased after years of rescheduling on 10nm – and that those finally at least see that Intel actually had a 10nm node running at all.
I think it's rather wishful thinking than actual condition that Intel will bring anything to market just a few month after TSCM being already shipping in volume. The process is broken atm and even if they'll be able to fix it (no given indication for that being actually happening anytime soon though), the last signs didn't even paint the picture their 10nm will bring any higher clocks, having higher energy-efficiency or bringing anything being superior to their latest 14nm process.Sure, they were leading for decades, yes. GloFo/Samsung/TSMC didn't weren't able to take the crown in less than 2 years. The thing is, Intel had a significant node-advantage ever since, even a quite great one, but that's past now. They're struggling not only on their 10nm node but have ever since their 22nm. Even their first 14nm products weren't able to deliver higher clock speeds than their former node and it took them years to fine-tune 14nm to archive the same yields on their 14nm they had on their 22nm.You're picturing it as all this happened overnight after all … I mean „less than 2 years“?!
So, recently?! May I kindly remind you of the initial roadmaps that seem like a silly joke today?
Since, according to those, they should've had (or rather wanted to) already deliver 10nm in 2015 ...
... and next year in '19 would have been already 5nm. Would … have been.
So judging by their own road-maps, they're already 3 nodes in arrears by next year. That surely isn't a time-frame I would dare consider to call „less than 2 years“ you know …
5nm in '19. They aren't even close to shipping anything from two nodes before that node as of yet. 5nm is scheduled to come after 7nm, 7nm after 10nm. Not even the latter one is in a condition to allow shipping anything fully working, apart from the fact that 10nm doesn't even deliver anything superior to their 14nm node.
I mean, let's take it for granted that 10nm is working as they (re-) scheduled it (again) the last time, shall we?
So 10nm is working. What you think they going to fab on it? Their usual bread-and-butter-chip like a quad-core? Do you honestly think you're going to draw any customers from the woodwork with a quad-core in '19? No way! Not after AMD humiliated the market for Quad-cores in mainstream with Octa-Cores with their penny-pinching.
The least you have to bring is a Octa-Core (which Intel already at leat understood, and act accordingly with the 9700K & 9900K). Do you think they're going to be able to fab a Die this big on 10nm by now? No way. If it isn't impossible, we'll see and witness a incredibly epic mircale in the electronic semiconductor industry for the first time. It's nigh impossible to increase the yield-rate this fast (to make a chip this big to happen), even with a completely working and perfect process. No-one is able to archive a stunt like that. It would not only be magic but a pure act of doing wondrous deeds, performing a miracle. Not going to happen.
As bad as it seems, we have to acknowledge the fact that AMD hit Intel with the most incredible potential at Intel's single worst time in both of their respective history. They have them by the balls – and to even shift that case of power-balance any back, let alone to correct it to the former state, will take years. As said, even if 10nm will be working perfectly today, they're still left with a architecture which is pretty much dead (not only on the security part …). Even if 10nm is working perfectly today, it will only by them a few month at best, until AMD can punch back even harder (as they do already by now).
The silly thing is, most of the epic failures Intel did, was not even to sit idling and waiting but to allow AMD to reach a position they have now. The moment AMD hired Jim Keller, alarms should've going off and even the fact that they were working on a 'complete new architecture' should've raised every fucking single red flag possible. They did exactly nothing instead.
It's like Intel watched their murder to sharpen the very knife AMD was going to slit their throat with (and even announced to do so) – and they (Intel) were somewhat stunned and fascinated by doing so. They wasn't even interested about any ongoing but just did not care, at all. They just gave a fuck about it. Cause of … I really don't know.
It actually really seems that Intel was just profoundly convinced that AMD – after their own official capitulation in '11 – would strike sail and that they henceforth, awed before Big Blue™, would content theirselves with just getting the fallen breadcrumbs. In the meantime Intel was supposed (and most probably saw themselves in doing so) in self-aggrandisement indulging to stoop downwards every third quarter to release their newest iteration of their oh so holy IPC-incarnation (which was nothing less than directly casted out of finest gold bullion – and hence pricy as fuck) into the market for the populace – which for sure included the obligatory supererogation within the range of measuring-tolerance, but for sure make prices spiral up for that anyway.
Though, that worked pretty well for Intel for almost a decade, didn't it?. And Intel most certainly would've had driven that game forever and for always even up to Tigerlake, wherewith Intel – on the condition that AMD never would've get up ever again – even would've had gone through with …
To be honest, what shocks me the most is the fact, how incredibly blind (and with that stupid or at least shortsighted) Intel must have been in actually thinking or at least fall victim to the green belief, that this would a) last for some time and most importantly, b) that AMD wouldn't come back (or at least try to do so). There were a greater number of companies in the past you better never ever write off (no matter in what condition or incredibly miserable their position was at that point). AMD always was that kind of company you better watch out for.
I mean, wasn't that this very company who managed to reverse-engineer Intels 386/486 and come out with a copy of it which even performed ways superior to their own (Intel's original)? It honestly drives me nuts how Intel could be so naive in thinking AMD wouldn't try something to come back. Those are these companies you better have a pret·ty damn sharp stare at if they 'surrender', lock theirselves into their basement and hunker down at home (and keep staying scaring silent while doing so).
Sure, it's just my opinion but the biggest fault Intel let happen is allowing AMD to reach the position they managed to have now. Intel should've been at full alert twenty-four hours a day the moment after the news that AMD hired Jim Keller was coming through the ticker – especially after the past and what Jim „The Godfather of Chip-design and the embodied IC“ Keller already did at and with AMD already back then in the times with the Athlon.
The thing is, here we're not talking about opinions of points of view, we're talking about facts. Facts say neither AMD nor Intel have nothing tangible to prove their words, nor that 7nm is ready to ship next year, nor that Intel should be able to launch Q4'19. And there's nothing that says otherwise besides words from both companies. I know very well intel's own roadmap is delayed, and they main reason for that is that they spent too much time with 14nm, which are so good now, that realizing a good 10nm which will actually show noticeable gains, is pretty hard, and that's also what was said around the internet, so this too could be a fuss and they could be having big troubles as you say, but even that is another news that nobody can confirm for certain, so i just don't get it how can you be so sure about stuff nobody except respective companies know. They said they should be able to launch products with 10nm by the end of 2019, what if it's all a bluff? What if they just can't get it right at all? What if they said that but they could be able to ship earlier than that, and what they said is only part of a market tactic to play? We can't know anything for sure, so you better stop acting like the truth is in front of us, because what you're seeing, i'm not, and viceversa.
Facts on the other hand are that Intel's architecture as old as it may be, it still shows what it's capable of versus a much newer architecture that is ryzen, and most of the times it also pulls ahead, even being short of cores/threads compared to ryzen.
Another fact is current process superiority, Intel, stupidly or not, they have a perfect 14nm, which smokes in every case what the other foundries have, and the 7nm TSMC is said to have ready, could even be not that much better than Intel's super refined 14nm, because if they work, like they had in the last processes, they will only make a "decent enough" 7nm, just good enough to secure a decent gain compared to THEIR own 14nm, not to the competition's.