Intel Reported To Be Looking At Acquiring GlobalFoundries

[Keats]

Intel 10nm is very similar efficiency/density as TSMC 7nm+. Simply "nm" aren't there real measurement but marketing term.

Why AMD is winning is because of multi chip design and that over the years they fixed mostly problems of memory controller, cache layout, chip-to-chip latencies, and that Intel on desktop is still using 14nm instead of 10nm. In laptops newest intel 10nm chips are doing actually fine vs newest Ryzens, both from power efficiency perspective as well single and multi core performance.

You mean 10nm+++, right?

[mdedetrich]

They should spend the money fixing their nodes. They have nothing which is even remotely close to TSMC's 7nm in terms of power efficiency and TSMC has long been churning chips using their 5nm node. Who cares that TGL reaches 5.3GHz when it's eating that much power in the process.

No they shouldn't, they would have gone bankrupt if they tried to chase for the top node.

Its a dog eat dog world in the semiconductor business if you are trying to compete in the top, it requires massive amounts of billions in capex expenditure and if you don't delver with good yields you are basically fuked. You also have to do transitions to different lithography which requires replacing a huge amount of equipment.

Also not everyone needs 7nm chips, what GlobalFoundaries did was the right decision. There are a large amount of customers still wanting 14nm or above, because its cheaper.

[drakonas777]

Some comments here are really misrepresenting things. ZEN3 is not TSMC 7N+ (EUV). Actually, specific lithography for ZEN3 is not officially confirmed by AMD to my knowledge, but from what analysts and tech people speak it's most likely not even 7NP, just same 7N used for ZEN2 with updated tools. Intel, on the other hand, is basically at it's 3rd 10N generation now, first being Cannon Lake, second Ice Lake and now "SperFin" for Tiger lake. So it's 10N++ by Intel terms, which is not really comparable to "TSMC pluses". You have to be specific which one you are talking about.

Roughly speaking, from the density point of view at least, Intel's nomenclature resembles one tier lower TSMC nomenclature: Intel 10N ~ TSMC 7N and future Intel 7N ~ TSMC 5N. But this kind of definition is not very accurate, there are a lot of there.

[ddriver]

Who cares if intel 10nm is roughly equivalent to tsmc 7nm, when the former still struggles to ramp up and the latter is already "old tech". Intel 10nm is nowhere near tsmc 7nm in terms of neither scale nor yield.

[hyperchaotic]

I have no idea what they are up to. GlobalFoundaries gave up on small nodes. So you will end up combining 2 foundries neither competent with small nodes? Or maybe if they think if they combine two struggling foundries that it will make a strong foundry? Maybe they finally know what they are doing, or maybe not.

Maybe Intel is trying to diversify business by adding contract manufacturing? There is a strong demand for mature nodes and it is only projected to grow. Over half of the global market for ICs is 68nm/40nm/28nm. GloFlo's 14nm and 12nm nodes could live on for a long time.

[ddriver]

Maybe Intel is trying to diversify business by adding contract manufacturing? There is a strong demand for mature nodes and it is only projected to grow. Over half of the global market for ICs is 68nm/40nm/28nm. GloFlo's 14nm and 12nm nodes could live on for a long time.

Maybe:

1 - intel is after some process IP it deems worth it acquiring an outdated fab
2 - and more likely - intel is in full desperation, spray-and-pray mode, hoping to hit something that will get it back to its glory days, or the good old "swing at every ball"

Cuz it is not like intel is short on what glofo has - dated 14nm process. If anything, their 14++++++++++ is marginally better.

There is healthy demand for older process nodes but it is not exactly high margin products, and intel has proven time and time again completely incompatible with low profit margin products. I mean if they are that desperate to diversify, they got a mountain of money and there is plenty of business prospect outside of semiconductors.

[Brane215]

GloFo has a reputation for being bottom-barrel when it comes to making logic. I got first hand experience of this when working for a while at one of *those* two top-level tooling companies.

So, you are saying GloFo is "bottom of the barrel" of the TOP THREE to FIVE PLAYERS IN THE WORLD ?
That's still pretty damn high.
They have had misfortune with SOI. But then, no one knew how it is going to pan out and (then) AMD wasn't exactly botomless money source.
And they had to compete against 10x bigger Intel.
Ofcourse they had to take some shortcuts and take their chances.

So they went with SOI because they felt they have greatest chances there.
It didn't pan out. That doesn't make them morons.

WRT to that engineerign beer stories, they have been known to say whatever for a beer.
Who cares about the shape of a transistor. Give me something I can work with and I don't care if it looks like a potato.

[piotrj3]

Who cares if intel 10nm is roughly equivalent to tsmc 7nm, when the former still struggles to ramp up and the latter is already "old tech". Intel 10nm is nowhere near tsmc 7nm in terms of neither scale nor yield.

Wrong. Density Intel actually has higher or similar (only TSMC EUV+ 7nm is better). Also yield seems fine too, since Intel newest mobile chips can turbo up to 4.8GHz, what in 45W package is rather promising. And Intel actually according to mobile CPU tests (recent ones) can actually compete and wins in single threaded workloads and in multithreaded wins in one loses in another. Just seems intel put all their 10nm capacity on server and mobile chips.

Again, where Intel is losing is lacking multi-chip designs and not having 10nm capacity on desktop. Those are real deal breakers as it allows producing 64 core CPUs, not by having extraordinary luck that in one large piece of wafer every key part is without errors. Production for AMD is simply cheaper and just by giving more cores or better binned 8 core chips they can compete with Intel very well everywhere, except total budget desktop CPUs or laptop CPUs, where i would say they are equal.

[Citan]

Am I the only one considering as a serious hypothesis the fact Intel wants to buy GloFo just to pursue the same illegal anti-concurrential manoeuvers they did 15 years ago (and probably have continued doing), but on the "root" level?

Like, "hey, sorry, we would REALLY love to produce your chips AMD, unfortunately we had another customer that reserved prod lines for about no more than 5 years ahead... Wait, let me get its name... Oh wait, it's me! No worries though, I can probably make some space for you, but I'll have to change my plans, so I need to overcharge you. Nothing more than 50% overhead, fair enough no?"

Even if AMD could technically fall back to another manufacturer, it would make a mess for them (negociation, organization, etc).

EDIT: ok, my bad, I missed quite a few memos, TSMC being apparently the prime manufacturer for AMD's latest CPUs since at least 3 years... XD

[Developer12]

Who cares about the shape of a transistor. Give me something I can work with and I don't care if it looks like a potato.

Shape is everything. We're forming nano-structures in silicon a few strands of DNA across which carefully manipulate quantum electrodynamics.

Higher precision in a process node means higher repeatability which means pushing the envelope further with more complex, more densely packed structures. It's the difference between being able to sometimes make a 3nm feature without accidentally carving a chunk out of it (and throwing off the electrical specs) and being dead nuts on and relying on it every time.

Unless you like having transistors with terrible characteristics that very from chip to chip, causing low yield issues and forcing you to dial back your designs so even the wonkiest transistors can keep up.