TSMC is posing a serious threat to Intel's own foundries

[Greg S]

In brief: Intel's manufacturing prowess that once ruled the industry may have already been quietly superseded by TSMC. As Intel struggles to keep up with die shrinks and TSMC earns even more business, Intel's own fabs have a lot of catching up to do.

Intel launched its 14nm process back in 2013. As this year comes to an end and Intel's 10nm chips are still figments of imagination, TSMC has shown that its contract manufacturing model could be a threat to Intel's own fabs.

Over the same time span, TSMC started producing 20nm chips and has managed to scale down to 7nm. Arguably Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured, but by the time Intel makes it to that point TSMC will already have produced millions of usable products that are in the hands of consumers.

Perhaps the biggest win for TSMC in recent history is earning all of AMD's business for production of desktop and server processors. Apple will also continue to use TSMC for the foreseeable future for A-series chips found in iPhones and iPads. Nvidia, Qualcomm, Broadcom, and many other big names are all customers as well.

Intel's own production has had its share of problems this year. A shortage of 14nm chips has lead Intel to actually contract out some of its own CPUs to TSMC for manufacturing.

Manufacturing woes aside, Intel has to worry about custom processors being built for cloud computing. Amazon has announced that it will be using homegrown, ARM-based CPUs officially named AWS Graviton Processors. Simply eliminating demand for some of Intel's pricey enterprise products could lead to trouble over the long haul. TSMC may gain even more business as a result of tech giants needing more custom hardware.

Even though Amazon is still going to be buying huge numbers of Intel products, removing the dependency opens the doors to new possibilities. As AMD churns out more server chips on smaller process nodes courtesy of TSMC, there is an even greater threat to Intel's historic dominance of enterprise markets.

Permalink to story.

https://www.techspot.com/news/77653-tsmc-posing-serious-threat-intel-own-foundries.html

 

[JB3543]

It's called "Lack of competition". It makes companies fat, slow, crappy, and only interested in covering their asses.

Just look at the US government, at the DMV, at Comcast, at Microsoft...

 

[wiyosaya]

It's called "Lack of competition". It makes companies fat, slow, crappy, and only interested in covering their asses.

Just look at the US government, at the DMV, at Comcast, at Microsoft...

I was going to say the same thing. sIntel has become arrogant in the advantage they held in the marketplace. All that money that they made, though, is doing little good for them ATM. The smaller players, notably TMSC and AMD, have figured out how to innovate with a minimal budget. sIntel would improve if it figured out how to innovate again.

Personally, I think NASA, the DoD (even with all the cost overruns on various projects), and, perhaps, the CDC all do a reasonable job of innovation, the rest of the US Govt., IMO, not so much so.

 

[mbrowne5061]

"Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured"
You got that backwards.

 

[mbrowne5061]

It's called "Lack of competition". It makes companies fat, slow, crappy, and only interested in covering their asses.

Just look at the US government, at the DMV, at Comcast, at Microsoft...

I was going to say the same thing. sIntel has become arrogant in the advantage they held in the marketplace. All that money that they made, though, is doing little good for them ATM. The smaller players, notably TMSC and AMD, have figured out how to innovate with a minimal budget. sIntel would improve if it figured out how to innovate again.

Personally, I think NASA, the DoD (even with all the cost overruns on various projects), and, perhaps, the CDC all do a reasonable job of innovation, the rest of the US Govt., IMO, not so much so.

I would narrow DOD to DARPA, but yeah. Also, surprisingly, the USGS; they have a pretty advanced, friendly, and free to use map viewer online. Just zoom in on the area you're curious about, and they'll show you every map they have that includes that area, all available for download in geoPDF or KMZ so you can put them into most GPS systems.

 

[wiyosaya]

"Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured"
You got that backwards.

What the statement says is that because of the method of measuring the transistors, Intel's 7nm transistors are comparable in size to TSMC's 10nm transistors.

 

[mxfalcon54]

"Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured"
You got that backwards.

What the statement says is that because of the method of measuring the transistors, Intel's 7nm transistors are comparable in size to TSMC's 10nm transistors.

No, mbrowne5061 is correct, Intel's 10nm process will be comparable to TSMC's 7nm.

 

[Dimitrios]

Sorry, but any bad news for INTEL is good in my book. Karma, INTEL was always slimmy greedy cheating POS monopoly company and once they got caught they had to play fair just like a baseball player on steroids that had to give up the juice. For now it's been fair game but we all know INTEL will do something shady to keep that title in the near future.

 

[wiyosaya]

"Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured"
You got that backwards.

What the statement says is that because of the method of measuring the transistors, Intel's 7nm transistors are comparable in size to TSMC's 10nm transistors.

No, mbrowne5061 is correct, Intel's 10nm process will be comparable to TSMC's 7nm.

Did I say mbrowne5061 was wrong? What I did was reiterate what the statement says. If the statement is wrong, that's a different matter.

From a slide in this article - https://www.extremetech.com/computi...samsung-globalfoundries-intel-gear-7nm-beyond sIntel may be exaggerating their processes. From recent sIntel antics, that would not surprise me in the least. The article does state

As nodes have become more difficult to hit, each foundry has adopted various methods of defining what constitutes a node shrink. This is most apparent with Intel; its 10nm node is expected to be equivalent to 7nm from TSMC.

If that expectation is met, it remains to be seen; IMO, whether that expectation will be met depends on the source. Is the source independent, or is it sIntel's marketing in attempts to build expectations for this node from sIntel?

Everyone exaggerates. However, it is still adversely affecting sIntel. I am not so sure that companies moving to TSMC would be easily fooled by exaggeration.

 

[Evernessince]

Sorry, but any bad news for INTEL is good in my book. Karma, INTEL was always slimmy greedy cheating POS monopoly company and once they got caught they had to play fair just like a baseball player on steroids that had to give up the juice. For now it's been fair game but we all know INTEL will do something shady to keep that title in the near future.

You mean more shady then using a water chiller to advertise product performance or paying for a rigged benchmark while all the real reviewers have to wait a week before they can release their reviews?

Too late, Intel has been engaging in shady business and only god knows what they are doing behind the scenes given what they do out in the open. I would not be the least bit surprised if they restarted their "rebate" program or threatened OEMs and server customers. Ever wonder why there are only a handful of Ryzen laptops despite the Ryzen mobile chip being an excellent product? It likely has something to do with Intel's grip on the laptop market.

 

[Theinsanegamer]

Another industry titan falls in the name of progress. No company, no matter how powerful or advanced, will eventually fail to compete and eventually fall.

Couldnt have happened to a 'nicer' group of industrial *******s. While intel has made many good products, their marketing practices are vile and ingenuous. Forcing them to play the defensive has been amazing to watch this last year, and will only get more entertaining as intel attempts to get back up on its feet after sitting on its hindquarters for so long.

 

[hahahanoobs]

AMD was completely out of the game for 5 years and came back. Albeit they rebranded Polaris 3 times, and cut Ryzen prices almost monthly. Meh.

I'd be a fool to count Intel out after a $17b quarter. 7nm Ryzen 2 will be what determines AMD's future, so I'll wait til at least then before I start throwing shade Intel's way.

 

[Knot Schure]

AMD was completely out of the game for 5 years and came back. Albeit they rebranded Polaris 3 times, and cut Ryzen prices almost monthly. Meh.

I'd be a fool to count Intel out after a $17b quarter. 7nm Ryzen 2 will be what determines AMD's future, so I'll wait til at least then before I start throwing shade Intel's way.

Just five years....? Are you sure?

 

[Danny101]

"Intel's 7nm process will be more comparable TSMC's 10nm technology due to how transistors are being measured"
You got that backwards.

What the statement says is that because of the method of measuring the transistors, Intel's 7nm transistors are comparable in size to TSMC's 10nm transistors.

No, mbrowne5061 is correct, Intel's 10nm process will be comparable to TSMC's 7nm.

Did I say mbrowne5061 was wrong? What I did was reiterate what the statement says. If the statement is wrong, that's a different matter.

From a slide in this article - https://www.extremetech.com/computi...samsung-globalfoundries-intel-gear-7nm-beyond sIntel may be exaggerating their processes. From recent sIntel antics, that would not surprise me in the least. The article does state

As nodes have become more difficult to hit, each foundry has adopted various methods of defining what constitutes a node shrink. This is most apparent with Intel; its 10nm node is expected to be equivalent to 7nm from TSMC.

If that expectation is met, it remains to be seen; IMO, whether that expectation will be met depends on the source. Is the source independent, or is it sIntel's marketing in attempts to build expectations for this node from sIntel?

Everyone exaggerates. However, it is still adversely affecting sIntel. I am not so sure that companies moving to TSMC would be easily fooled by exaggeration.

I would say that Intel's 10nm performance characteristics would be comparable to TSMC's 7nm. That much, I can believe.

 

[Theinsanegamer]

AMD was completely out of the game for 5 years and came back. Albeit they rebranded Polaris 3 times, and cut Ryzen prices almost monthly. Meh.

I'd be a fool to count Intel out after a $17b quarter. 7nm Ryzen 2 will be what determines AMD's future, so I'll wait til at least then before I start throwing shade Intel's way.

Just five years....? Are you sure?

Phenom II wasnt the fastest, but it was a genuine competitor. It was slower then nehalem, sure, just like ryzen is slower then core I. But they were cheap, OCable, and were capable of making the fastest GPUs of the era the bottleneck in any properly set up system. Not to mention, you could OC any black phenom, not just the "K" models, letting you get high end clock rates out of the lowermost black models.

Bulldozer is when AMD began really huffing their own farts.

 

[mbrowne5061]

I would say that Intel's 10nm performance characteristics would be comparable to TSMC's 7nm. That much, I can believe.

It has more to do with how the measurements are being made, and less to do with performance. Intel measure in such a way that it include "more" of the transistor, TSMC seems to measure in such a way that it measures "less", and this difference has more to do with difference in engineering opinion than anything else. The benefit here is TSMC gets to sound like they're smaller, and Intel gets to sound like they're able to do more at larger lithographies. In reality, both are pretty much the same size, and will perform about the same (when Intel finally figures out how to produce 10nm processors, that is)

 

[hahahanoobs]

Phenom II wasnt the fastest, but it was a genuine competitor. It was slower then nehalem, sure, just like ryzen is slower then core I. But they were cheap, OCable, and were capable of making the fastest GPUs of the era the bottleneck in any properly set up system. Not to mention, you could OC any black phenom, not just the "K" models, letting you get high end clock rates out of the lowermost black models.

Bulldozer is when AMD began really huffing their own farts.

AMD's last competitive chip was Athlon64.
And a lower price is not competition. It's desperation.
Don't get it twisted.

 

[Kashim]

About damn time! Intel had their monopoly long enough, time for some actual competition.

 

[p51d007]

It's called "Lack of competition". It makes companies fat, slow, crappy, and only interested in covering their asses.

Just look at the US government, at the DMV, at Comcast, at Microsoft...

Yep...Xerox, Zenith, American Auto companies, Apple, the list goes on and on.

 

[Forebode]

Phenom II wasnt the fastest, but it was a genuine competitor. It was slower then nehalem, sure, just like ryzen is slower then core I. But they were cheap, OCable, and were capable of making the fastest GPUs of the era the bottleneck in any properly set up system. Not to mention, you could OC any black phenom, not just the "K" models, letting you get high end clock rates out of the lowermost black models.

Bulldozer is when AMD began really huffing their own farts.

AMD's last competitive chip was Athlon64.
And a lower price is not competition. It's desperation.
Don't get it twisted.

Desperation or not, it doesn't really matter if it works and they still make a profit right? So you don't think the ryzen or threadripper are competitive chips? Not even sticking with AM4 for a bit longer? What about motherboard pricing? The percentage of gains Intel might have over AMD for performance is one thing.. but look at the costs of building each system with the features you'd like. the Intel will cost more.

Most builders have a budget that wont allow them to get the most powerful single core performer out there. The future of programs and games are leaning towards multicore and gpu acceleration. Say you can save $5-$600 by going to with Amd, that's $5-600 more you can toss at the vcard.

Are you really saying AMD isn't Competitive?

 

[hahahanoobs]

Desperation or not, it doesn't really matter if it works and they still make a profit right? So you don't think the ryzen or threadripper are competitive chips? Not even sticking with AM4 for a bit longer? What about motherboard pricing? The percentage of gains Intel might have over AMD for performance is one thing.. but look at the costs of building each system with the features you'd like. the Intel will cost more.

Most builders have a budget that wont allow them to get the most powerful single core performer out there. The future of programs and games are leaning towards multicore and gpu acceleration. Say you can save $5-$600 by going to with Amd, that's $5-600 more you can toss at the vcard.

Are you really saying AMD isn't Competitive?

Multi-core chips are not what the average consumer needs. Especially multi-core chips with lower clock speeds. Smartphones are a lot of peoples' daily driver when it comes to computing and that says a lot, but is often overlooked.

AMD pushed multi-core, because they suck at clock speeds and desired architectural advancements. You can only root for the underdog for so long before that gets old, so again, 7nm Zen and consumer Navi will determine AMD's future.

I'll root for AMD when their mainstream products are consistently better in everyday computing.

 

[SharkMe]

Did I say mbrowne5061 was wrong? What I did was reiterate what the statement says. If the statement is wrong, that's a different matter.

mbrowne5061 is in contradiction with the statement and with your re-iteration of it. Only one can be true. mbrowne5061 is correct and the article statement is incorrect.
To answer your question, yes you did say that mbrowne5061 is wrong by contradicting what he said.

 

[dj2017]

20nm from TSMC where a joke. TSMC not only managed to move from 20nm to 7nm, but also managed to make a reputation from a company that you could not entirely trust to a company that is the logical choice for any next generation chip.

 

[wiyosaya]

If AMD can produce higher end chips and mid ranger chips cheaper than intel they would control the market.

AMD is going after the server market - and they are making inroads.

 

[wiyosaya]

mbrowne5061 is in contradiction with the statement and with your re-iteration of it. Only one can be true. mbrowne5061 is correct and the article statement is incorrect.
To answer your question, yes you did say that mbrowne5061 is wrong by contradicting what he said.

I left room for mbrowne5061 to provide a source by not outright saying what was posted was wrong. Also, your post had no source. In fact, I provided that source - which both of your posts could have provided easily and there would have been no question about it. Sources matter.

My apologies you found my statement disagreeable. Next time, it would be helpful to have a source of the information.