Intel's 7nm is busted, chips delayed, may have to use rival foundries to get GPUs out for US govt exascale super Intel on Thursday reported second quarter sales of $19.7bn, up 20 per cent year-over-year, and still investors distanced themselves from its stock, which was down around 10 per cent in after-hours trading following the manufacturer's earnings report. The US giant's results [PDF] exceeded what analysts had anticipated – and …
At what point does a company with big bucks decide to invest in AMD and seriously take on Intel in the server and laptop market? AMD seems to have the products now, they just need the capital to buy more silicon capacity and give the big OEM's and hyperscalers the assurances that they can scale up production. That would put Intel in a serious bind.
AMD's capacity problem is due to their reliance on TSMC. They can't afford to outbid Apple, who gets first access to TSMC's newest processes because they prepay for billions of dollars worth of wafers every year since they know what their demand will be. If you are paying ahead of time for wafers and TSMC is using that money to help deploy their new process, you are going to get first dibs on the first wafers made with that new process.
Perhaps they could move ahead of Qualcomm, though Qualcomm's wafer volume exceeds AMD's wafer volume several times over - it probably even approaches Intel's PC/server wafer volume so good luck cutting ahead of them without throwing around a lot of cash.
I'm not sure they have to outbid anyone. I suspect with the right financial backer for AMD, TSMC would be willing to invest in the additional fab capacity needed. But that would likely require some guarantees concerning future purchases or even direct investment in TSMC, which AMD currently can't provide.
I was thinking about IBM as a possibility - they have the connection with AMD (check the AMD CEO's background), and they should be looking for new opportunities given the somewhat sorry state of their current business.
Just a thought - I have no special knowledge.
Not only that but TSMC adding a bunch of leading edge capacity for a single client who won't need it once something newer comes along is difficult. TSMC plans their capacity based on years of usage, to do otherwise would make the leading edge capacity more expensive.
Reportedly this is why Apple uses 'old' SoCs in some of their products like Apple TV and HomePod - to meet commitments they made to buy x wafers once a process becomes trailing edge (they also use them in the 'last year's iPhones' but they never know for sure how many of those they will sell versus the newer ones)
This post has been deleted by its author
It would have to be either a company with fabbing capacity or the money to pay for it – Samsung maybe. Because the bottleneck isn't chip design but the ability to make the chips. Might have been possible for Softbank to bankroll a couple of fabs before the "Vision Fund" bubble burst.
Otherwise most eyes will be on SMIC, Samsung and the other couple of companies that might want to get involved. And feedback on Apple's own notebbook chips, because if these and the emulation work well there could well be a stampede from x86 to ARM.
"It would have to be either a company with fabbing capacity or the money to pay for it – Samsung maybe"
How portable are the chips at moving to different fabs that use different processes? I don't know, but would suspect they are designed to use the processes that the target fab uses.
"How portable are the chips at moving to different fabs that use different processes? I don't know, but would suspect they are designed to use the processes that the target fab uses."
Certainly Intel used to like people to believe that close architecture/process co-operation worked in their favour.
"Tick tock" advances used to be a key marketing message from Intel: alternate process improvements and architecture improvements:
"A Tick Advances Manufacturing Technology
In every "tick" cycle, Intel introduced advanced manufacturing process technology to help deliver the expected benefits of Moore’s Law to users.
A Tock Delivers New Microarchitecture
In every “tock” cycle, Intel introduced manufacturing process technologies to introduce the next big innovation in processor microarchitecture. Intel® microarchitecture advancements seek to improve energy efficiency and performance as well as functionality and density of features such as hardware-supported video transcoding, encryption/decryption, and other integrated capabilities."
from https://www.intel.com/content/www/us/en/silicon-innovations/intel-tick-tock-model-general.html
Tick tock was publically abandoned around 2016 when it became apparent that Intel's 10nm process wasn't what it ought to have been, see e.g.
https://www.anandtech.com/show/9447/intel-10nm-and-kaby-lake
Intel tried to continue the link by describing their new approach as "Process - Architecture - Optimisation" but sooner or later people will notice that the new products aren't available as planned.
Are not direct replacements for B450 motherboards - they are a step up in quality and specifications (to the point where the cheaper X570s look poor in comparison) . The motherboard manufacturers have finally accepted that AMD chips are not the cheap and cheerful option (and therefore only require cheap and cheerful motherboards) but quality chips in their own right (I'd swap my skylake-x for a ryzen 9 3950 at the drop of a hat - even on a B550 board) .
Competition isn't just about prices but also about driving innovation and quality.
> Are not direct replacements for B450 motherboards - they are a step up in quality and specifications
Indeed they are, but for people that don't need the bells and whistles that B550 provides (perhaps they just want to run Zen 3 when it comes out), there are few alternatives. Which is why you cannot buy decent B450 motherboards for love nor money right now.
There is massive demand out there for more budget boards.
I got a B550 a few weeks ago. First time ever I've built an AMD CPU based machine. But Intel has so dropped the ball with those comet lakes, which is what I originally was looking at. So now I got a B550 board with a midrange Ryzen 5 3600 and going to drop in a beast when Zen 3 comes out and be settled for the next four to five years. We'll see if Intel recovers.
I agree. We dont want the reverese of the past decade with AMD struggling so much Intel could price how they wanted.
Its been great having AMD challenge Intel and their Ryzen range is doing really well but eventually AMD will hike prices if they are outperforming Intel by so much.
Already shown with the odd 3000XT release.
Compared to previous quarter revenue in data centre and consumer down 5%ish and income 25%ish...even more for consumer so they may be having to give big discounts
Intel has used the same Auditor E&Y (of Wirecard infamy) since inception 50+ years ago. I expect another auditor would make them write down the investments in their 10nm fabs by a few billion.
Maybe TSMC might find room for Intel if Intel agreed to transfer some of their ASML ordered machines to TSMC at cost :-)
The analyst Charlie Demerjian who was one of the first to call out Intels failings several years back was tweeting a lot yesterday during the earnings call.
the only options for 7nm grade processes are either TSMC or SAM
TSMC are apparently commited to build a fab state side, but it might take a while
I can see samsung ramping up its low nm fab production and posibly opening some in vietnam tha will eat the capacity, but they are probably looking at ramping up exynos production first
Intell might have a problem getting anyone else to co-operate due to their previous actions, but SAM are not adverse to supplying rivals if the price is right (iPhone screen anyone)
Yes, indeed they could.
And every company using TSMC will therefore be screwed thanks to the Chinese government's policy of requiring all technical innovations be made 'public owned'.
Remember that Intel Management Engine security question that's been bouncing around for years now? "Security through obscurity"? No longer really "obscure" if the Chinese government forces TSMC to share all the internal designs of the Intel chips they'll be building, is it? Design companies need to work closely with fabbers during the design process in order to synchronize abilities and output - I'm sure TSMC will end up with more Intel (and AMD for that matter) secrets than anyone else in the world besides Intel and AMD itself.
So those suspected Chinese government- sponsored attacks vectors? Nothing like having early and best design specs before the chip even hits the streets!
Be interesting to find out what the gate oxide thickness is.
Usual rule of thumb is it's 1/10 line width.
Anyone seeing where there's going to be an issue here?
Personally not bothered. Why should I be? With developer productivity stuck at 10-20 lines of code a day for decades.
The process nodes have not been about the geometries for a while now, it is probably better treated as a marketing term to represent a new foundry process with certain characteristics.
Reference: Work for a company that is a TSMC customer at these new nodes.
https://en.wikichip.org/wiki/technology_node
There are some difference in the way Intel defines a specific node, perhaps to stay true to Moore they do it the hard way? I'm not sure.
Ultimately if Intel's 7nm can match/beat AMD 5nm in performance and power at the same manufacturing costs, does it really matter, beyond bragging rights?
Intel has always had their own process for fabicating its parts -- 'process' being the actual methodology of building the wafers. They've been enormously successful but the downside is that this breeds a sort of NIH mindset where its difficult to muster sufficient resources to keep at the front of the pack and difficult to get the company to start leveraging other companies' technologies. They've been in this hole before although most people didn't notice it -- back in the 2000s there was the original processor group churning out incremental improvements to the Pentium 4, faster and hotter and also completly unsuitable for mobile use. They got bailed out that time by a completely different group (based in Israel) that rearchitected the processor ostensibly for mobile use that enabled a big jump performance without running at a dull red heat.
It would be wrenching for the company and its culture to have to admit that it needs to cooperate to survive but its probably a wise decision. Intel has numerous production plants so hopefully they won't just shut them down and outsource as has been fashionable with corporate America but will switch the plants over and maybe become a contract manufacturer for fabless design houses (a TSMC competitor?).
Process, and processors too maybe?
"It would be wrenching for the company and its culture to have to admit that it needs to cooperate to survive"
Didn't a major part of Intel already go through that when their so-called "industry standard 64bit" (y'know, IA64, Itanic, whatever) failed repeatedly to impress the markets and eventually got blown out of existence by something that Intel HQ had repeatedly said wasn't even possible, let alone going to happen Inside Intel?
From where I sit, Intel is starting to have a history of botching its fabrication process.
It failed on 14nm, now I read it failed on 10nm.
I'm sorry, but how long is Intel going to be able to continue to fail before falling flat on its face ?
Intel is a CPU maker, for Pete's sake. If it can't make its own CPUs, then what can it do ?
"Intel is a CPU maker, for Pete's sake. If it can't make its own CPUs, then what can it do ?"
Intel has historically had a captive market, a huge cash pile, and an ability to do brand management.
Not unlike Boeing, in some ways related to business acumen and competence..
So yes, you're right to ask, what can Intel do?
We may find out soon enough.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
