Sign in with
Sign up | Sign in
Your question

Has Intel solved its foundry problem?

Last response: in CPUs
Share
November 3, 2010 12:49:52 AM



Intel Corp. has agreed to make chips for a semiconductor start-up, the first time the Silicon Valley giant has given another company access to its most advanced production processes.



Read more: http://online.wsj.com/article/SB10001424052748704477904...
Its been my belief that with smaller nodes, and Intels abundance of fabs, theyd at some point run into an overabundance of fabs, which Id thought could be a potential problem for them, as theres just not any real players left wanting old fabs

Now, the only question remains, as Intel expands into more and more markets, how far can they can they do this without competing against itself, and , possibly, having a upstart creating shortages on their fabs for orders?
It seems like a GF dilema all over again
a b à CPUs
November 3, 2010 1:28:47 AM

All they need to do now is give Nvidia an x86 licence then they would have another customer to make chips for. :lol: 
m
0
l
November 3, 2010 1:39:37 AM

Exactly heheh
Awhile ago, during the AMD/GF scenario, this wasnt seen at all
What if the green team sells so many chips, they starve Intels capacity?
OK, all kidding aside, this takes mighty Intel down a notch in my book
Many opinions never saw this, a few did

Its a smart move by Intel, as some saw this coming, with too much production vs sales
m
0
l
Related resources
a b à CPUs
November 3, 2010 1:48:14 AM

It should be interesting to see who they make chips for and what those chips and companies go on to become, there must be quite a few areas that Intel haven't gotten into yet for various reasons and it's in those areas that others might be able to capitalise on.
m
0
l
November 3, 2010 2:06:57 AM

Yep
Whats interesting is, without doubt, Intel has the lead in process
Having new entities using those nodes/processes could literally be aimed at others in competing (with Intel) ventures
Another thing to watch is, their margins
m
0
l
a b à CPUs
November 3, 2010 3:14:29 AM

I guess it makes sense for big companies with fabs. In particular to increasing costs to get to new nodes, more and more companies will rely on fabs. Heck, most of the fab companies are working together anyway for the tougher nodes.

Sure isn't cheap to research new nodes.
m
0
l
a b à CPUs
November 3, 2010 12:58:06 PM

From the front page here on Toms: http://www.tomshardware.com/news/10nm-nanometre-flash-n...

Quote:
Intel, Samsung and Toshiba are banding together to tackle the big task of getting chips down to a 10nm semiconductor line width.

Such a task, would be huge for one company alone, and even the trio will be inviting more to the team to help. The three companies will form a consortium and will invite about 10 more companies into the group.


So I would imagine closer cooperation between these companies in the future - maybe Intel fabbing chips for its consortium, if indeed they suffer from too much capacity. However their quarterly report margins remain pretty high - nearly 70% - so either they charge too much per chip or their overhead costs (incl. the fabs) are low. Or both :p . So it seems they don't suffer from too much capacity, at least for the moment. I also recall some articles a year ago about how China buys older fab equipment, so unless that market has dried up, I think Intel won't worry too much about selling their used stuff which has most likely reached 100% ROI within the first couple of years of use anyway.

Anyway, it occurs to me that GF (ATIC, actually) will have to spend lots more $$ to try and take the process lead away from Intel, if indeed Intel starts competing for their customer base. And GF has a lot of capacity coming online in a few years - the NY fab, the announced one in the UAE, the ones they acquired in their foundry buyouts, plus the upgraded ones in Germany. AFAIK, AMD is the only customer they have for SOI - everybody else is on strained silicon - so if played out as above, this could cause GF to re-evalute their priorities there as well..
m
0
l
a b à CPUs
November 3, 2010 1:04:51 PM

Since Intel is in that consortium, i doubt AMD would be invited onto the group! :D 
m
0
l
November 3, 2010 11:30:10 PM

Funny how things change, innit it?
Seemed a short while ago, this just couldnt happen, as it was said that Intel didnt need this, wouldnt need to do this etc
Im glad theyre doing it, myself, much like their igps
m
0
l
a b à CPUs
November 4, 2010 6:13:56 AM

That is true. If its one thing i hate is a company who tries to monopolize the market and also make the competition look like novices which they tried to do with AMD. Once upon a time Intel was strictly against overclocking. But in time they have realized how it has grown into a hobby of PC lovers and gave in.
m
0
l
a b à CPUs
November 5, 2010 11:42:28 AM

Are they still going along with the overclocking restrictions for Sandy?
m
0
l
a b à CPUs
November 5, 2010 4:03:11 PM

^ Depends on what you mean by 'overclocking restrictions'. The "K" versions have an unlocked multiplier.
m
0
l
a b à CPUs
November 5, 2010 10:05:42 PM

I mean for the normal versions.

K versions almost defeat the purpose of overclocking. That is, taking a low-end chip and clocking it to match the performance of a far more expensive chip.
m
0
l
a b à CPUs
November 5, 2010 10:39:07 PM

amdfangirl said:
I mean for the normal versions.

K versions almost defeat the purpose of overclocking. That is, taking a low-end chip and clocking it to match the performance of a far more expensive chip.

We'll just have to wait and see what they do with the K series and "normal" parts. Maybe you'll be able to overclock the lower parts to a decent level and the K series up further than that. It's all about business with Intel, whatever they can make money on they try it, but AMD is the same way, even if it doesn't seem as much so.
m
0
l
a b à CPUs
November 6, 2010 3:31:57 AM

I still remember when all chips were unlocked :D .
m
0
l
a b à CPUs
November 6, 2010 3:09:28 PM

amdfangirl said:
I mean for the normal versions.

K versions almost defeat the purpose of overclocking. That is, taking a low-end chip and clocking it to match the performance of a far more expensive chip.


According to the Anandtech preview, the two Sandy Bridge K versions should be priced around $560 and $210:

Quote:
the Core i7 2600K will take the place of the Core i7 950, currently priced at $562. The 2600 will fit somewhere around the 680 and 875K ($342) and the 2500K will replace the i5 760/655K ($205 - $216).

The cheapest Sandy Bridge at launch will be the Core i3 2100, which will replace the i3 560 at around $138.


IIRC, you were more interested in underclocking instead of overclocking. Just remember - everytime you turn off your computer, you've achieved the max underclocking possible :p ...
m
0
l
a b à CPUs
November 6, 2010 3:36:45 PM

From EE Times :

Quote:
Will Intel be a big foundry player?
11/5/2010 2:33 PM EDT
SAN JOSE, Calif. – As reported, Intel Corp. this week took a step into the foundry business. Intel will lend its semiconductor process technology muscle to build FPGAs for programmable logic startup Achronix Semiconductor Corp. at 22-nm and beyond under the terms of a strategic agreement between the two companies announced Monday (Nov. 1).

Executives from Achronix (San Jose, Calif.) said the deal would not only help the company bring 22-nm FPGAs to market faster than programmable logic market leaders Xilinx Inc. and Altera Corp., but also give the startup a leg up in 15-nm and future technology nodes.

For years, Intel has dabbled in the foundry and ASIC markets. But the chip giant exited the ASIC business some years ago and has never been thought as a foundry player.

Now, there are signs that Intel is interested in the foundry business. It has reportedly hired a vice president in charge of foundry. Companies are approaching Intel about access to its fabs. Perhaps Achronix is a guinea pig and the start of something big. Rumors are running rampant that FPGA startup SiliconBlue has approached Intel.

The question is will Intel become a foundry player or not? Will it challenge GlobalFoundries, Samsung, TSMC, UMC and others? Or will it just dabble in the arena?

Peter Clarke, European news director at EE Times, said: "I think Intel will get involved in the foundry market. I think the days of the self-sufficient IDM are over. If you are a chip maker, you need volume—and there are plenty of fabless companies out there who will buy a piece of your manufacturing output. We see that Samsung has thrown its hat in the ring, and now Intel is dipping its toe. Intel of course will never be a pure-play foundry like TSMC, but it will be a chip maker that does foundry work like IBM, Samsung and others. TSMC will remain the leading pure-play foundry, but Intel can throw a wrench in the works by doing leading-edge processes for companies it see as strategic partners."

Dean Freeman, an analyst with Gartner Inc., said: ''Will Intel become a big foundry player? On the wireless or (FPGA) side, they have the potential of being a foundry player. Does Intel have the capital and manufacturing to be in the foundry business? Yes. But the question is can they compete at the same cost as TSMC, UMC and GlobalFoundries.’’

G. Dan Hutcheson, chief executive and chairman of VLSI Research Inc., said: ''Given the fact that Intel did not formally announce an entry into to foundry business, but instead disclosed as an arrangement with a fabless company, indicates that Intel is testing the waters, not plunging into them. Not formally announcing an entry into the foundry business means they can easily retreat if it doesn’t work. Intel is not a company that burns the ships when it arrives in a new world. So the risk is low and they will learn a lot. At the same time, releasing their 22-nm process to an outsider means Intel is very serious about winning – more serious than they have ever been. Does it make sense? The short answer is yes. Growth outside a core market is best done by utilizing core strengths. For Intel, this is processors and process. There is no reason to constrain themselves to PC processors. But this means they must move towards embedded processors. Pushing them in this direction is the fact that systems in embedded designs continue to look more like PCs, while applications for classical PCs are not outgrowing the market. Processors are integrating more, rather than just getting bigger, making them look more like an embedded device going forward.''

Mark LaPedus, semiconductor editor of EE Times, said: ''Clearly, Intel won’t fab parts for its competitors, such as AMD, Broadcom, Nvidia, Qualcomm and others. But there are some startups or non-competitive companies that may be interested in doing business at Intel, especially in baseband processors and FPGAs-and for good reason. The foundries, namely TSMC, fumbled the ball at the 40-nm node and struggled with the process. Wait until we get to the 28-nm node and beyond. I hear the foundries are struggling to ramp with good yields. And despite the denials, I hear the foundries are struggling with their high-k/metal-gate technologies. Intel has successfully put two generations of high-k in production. Here’s one prediction: If the foundries fail to deliver high-k in time, watch out! Look for Altera, Xilinx and others to call on Intel. But there is more to the foundry business than just churning out parts. There is a service and IP mind-set. It’s really a service business, which is a whole new mind-set. Intel’s focus is processors. Beyond processors, the chip giant has never really found much success. We see that Intel is re-entering the wireless chip space with the acquisition of Infineon’s wireless business. I see Intel flopping in wireless again. I see Intel making a strong push in the foundry business. I see modest success for them. But TSMC CEO Morris Chang must keep a close eye on GlobalFoundries, Samsung, and now Intel.''

Dylan McGrath, editor of EETimes.com, said: "I don't believe Intel has the willingness, desire or intention to become a significant foundry player. Intel's manufacturing technology is the envy at all, but it is dedicated to Intel. The company has the capacity it needs to build parts for its business, and it's not going to sacrifice its position there to dedicate its lines to other companies. Intel may be dabbling in foundry work, but in the face of a sharp upturn in business, foundry customers (including Achronix) will see their orders take a back seat to demand for Intel products, just as we have seen with other IDMs who dabbled in foundry in the past. This deal may be partly about Intel trying its hand in foundry and experimenting with the possibilities, but in my view it has more to do with getting its feet wet with programmable logic technology—specifically Achronix' asynchronous logic technology."


Some other thoughts on the Achronix deal, summarized from http://roborat64.blogspot.com: Intel's fab processes rely on pretty restrictive design rules in order to get the max performance, perhaps more so than what GF, TSMC, etc. use in their processes. Hence it is more time-consuming and expensive to take some other company's design and retool it for Intel's processes. However an FPGA (field-programmable gate array) can be optimized for Intel's RDR process and then afterwards personalized for the end user (or Intel for that matter). The tradeoff of course is some wasted silicon for the unused parts of the FPGA after it is programmed, as it is unlikey that a general purpose design will use 100% of the available resources. IOW the end customer has a particular set of requirements and selects the closest FPGA that will meet those requirements after the programming is done.

So if Intel wants to position Atom into various embedded platforms like smartphones, smart TVs, car electronics (GPS, communications, entertainment, etc), it would make sense to do Atom + FPGA so that the customer can then program it for the end application, instead of developing an SOC for each customer's application and having the time and expense of RDR design. This would explain why Intel chose to partner with Achronix (well that plus the fact that the big players in the FPGA field - Xilinx and Altera - have already started FPGA designs with embedded ARM CPUs).

My own thoughts are that AMD needs to take note of this trend and see about shrinking Bobcat down to the <1 watt range, and then partner with some FPGA design company, if they want to be on the same page as the other players instead of playing it safe and waiting until the industry is about to move in yet another direction before getting in on the already old action..
m
0
l
a c 172 à CPUs
a b å Intel
November 8, 2010 12:46:42 PM

amdfangirl said:
I still remember when all chips were unlocked :D .


Then you also remember why both Intel and AMD started locking the multipliers.
m
0
l
a b à CPUs
November 9, 2010 2:52:45 AM

AMD shouldn't stop overclocking. I still believe that overclocking would be beneficial to a company if only one lets it.
m
0
l
a b à CPUs
December 12, 2010 5:10:23 PM

Usually I don't revive old (and never necro! At least, intentionally :p ) threads, but I found this article on the very same subject, by David Kanter:

Quote:
Intel’s high-volume logic manufacturing prowess is unmatched in the industry. IBM has equally talented research and development, but silicon is just one dimension of their overall system-centric approach (e.g. packaging, interconnects) and they are a low volume and high cost player. As a pure-play foundry, TSMC has great volumes through their fabless customers, but less impressive R&D and lower performance silicon. The nascent Global Foundries has the potential to marry IBM’s superb R&D expertise with high volumes, but it is far too early to tell. The bottom line is that Intel perceives manufacturing as a key differentiator and core competency and has the results to prove it:

•Intel is 6-18 months ahead of competitors to new manufacturing nodes (e.g. 32nm in late 4Q09, Global Foundries and IBM 32nm expected in 2Q11).
•Intel was the first to commercially implement strained silicon.
•Intel was the first to commercially implement high-k gate dielectrics and metal gates and a full generation (3+ years) ahead of everyone else (45nm vs. 32nm).

Last Monday, Intel announced that they had voluntarily entered into a manufacturing partnership with a third party, one of the first times in the company’s history. This broke an implicit assumption held by many observers, and prompted quite a few questions, which we will endeavor to answer.

•How could Intel open up their crown jewels to a third party?
•Does Intel want to get into the foundry business?
•Why is Intel doing this?

To answer these questions, it is essential to first understand exactly what was announced. Intel’s partner, Achronix is an FPGA start up (under $10M in revenue last year). Achronix uses novel asynchronous logic techniques and specializes in producing high speed FPGAs. Their current products are fabbed on TSMC’s 65nm process and reach 1.5GHz, roughly 3X faster than FPGA leaders Altera and Xilinx. Achronix’s second generation Speedster 22i FPGA will be fabbed on Intel’s 22nm process technology, with engineering samples expected by 4Q11 – simultaneous to Intel’s initial ramp of 22nm. The capacity and revenue for Intel will be miniscule (<1%), although they have also said that other deals may follow. Reportedly, Achronix will also have access to Intel’s cache coherent QuickPath Interconnect (QPI). This strongly suggests that Achronix will be using Intel’s high performance 22nm process technology (P1270), since it is unclear if Intel has designed QPI to work on their low power process and SOCs.


Kanter goes on about Intel perhaps dabbing a toe into the foundry business, but concludes this is pretty unlikely because there will be 3 big players in that market - TSMC (who had a high point of 50% gross margins recently, which is chickenfeed compared to Intel's 65% or higher margins), GloFlo and now Samsung. So all that competition means lower prices and profits, unless one of the big 3 can obtain an insurmountable tech lead over the other two. And Kanter points out that a foundry usually has fairly relaxed design rules compared to a tightly-knit one like Intel uses, because a foundry has to service many customers instead of just one.

Kanter has a 2nd theory about Intel looking to eventually acquire Achronix, but his most likely theory is that Intel wants to put a high-performance FPGA into the new architecture 22nm CPU (Haswell IIRC):

Quote:
A third theory is that Intel is partnering with Achronix to specifically complement and integrate with their existing product portfolio and roadmap. There are two obvious areas for such collaboration.

The first opportunity is a coprocessor for Intel’s high-end CPUs (and perhaps the successors to Larrabee). FPGAs have unique compute capabilities that can easily exceed the performance and efficiency of CPUs or GPUs for the right workloads. In HPC, this approach has been advocated by companies such as SGI or Maxeler. There was even a joint presentation from Maxeler and Schlumberger at Hot Chips 2010 about FPGAs for oil and gas exploration. Certainly, the NSA would love to have high performance FPGAs tightly coupled to CPUs – even more than the popcount instruction. In the commercial server world, Netezza has been using FPGAs to improve query performance for data warehousing for 10 years. FPGAs are even more popular in embedded communications, where they can act as DSPs for wireless basestations, networking line cards for wireline applications and many other roles. The fact that Achronix will have access to QPI strongly suggests a future as a coprocessor for high-end applications, whether they are in the server or embedded market.


Kanter also thinks Atom will benefit from using an FPGA, but either will have to get a QPI link or an FPGA that doesn't need it. So here Intel may partner with the FPGA leaders Xilinx, Actel and Altera.

Anyway he concludes:

Quote:
The hypothesis that seems the most likely is that Intel is strategically engaging with companies that complement their current and future product portfolio. The fact that Achronix will have access to QPI seems to imply that it will be paired with Xeon server or embedded processors. There is definitely a niche for FPGA acceleration and Intel already has partners using the older front-side bus. For the future though, QPI is necessary and may be the first step towards on-die integration with a ring (or other topology) interconnect. A tightly coupled FPGA coprocessor is extremely beneficial for high performance embedded applications (e.g. networking and storage) and would also be a good counter to the use of discrete GPUs for HPC. While FPGAs are challenging software targets, they can achieve higher performance in some cases than GPUs and also be used for specialty problems (e.g. deep-packet inspection, cryptography).

The bottom line is that Intel is not likely to enter the foundry business and the most plausible explanation is that they are pursuing complementary technologies. More generally, Intel understands that they cannot do everything for everyone (nor do they desire to do so). This is especially true when it comes to the embedded market, which is nearly impossible to describe because of the variety and breadth of applications. There are already examples of Intel working with third parties to create products that can address different embedded niches. This suggests that Intel may pursue other partnerships in the embedded space to move Atom into new markets, which could actually have a far larger impact than the Achronix deal. In some senses, this would be a second attempt to achieve the same goals of Intel’s unsuccessful partnership with TSMC on Atom – empowering the x86 ecosystem with a variety of hard and soft IP from third parties.


So this may be Intel's answer to AMD's Fusion on the server and desktop market (imagine both a GPU and FPGA on Intel's ring bus that they are introducing with Sandy Bridge next month), and to ARM in the Atom market..
m
0
l
!