Why Intel Leads the World In Semiconductor Manufacturing

MrSeb writes "When Intel launched Ivy Bridge last week, it didn't just release a new CPU — it set a new record. By launching 22nm parts at a time when its competitors (TSMC and GlobalFoundries) are still ramping their own 32/28nm designs, Intel gave notice that it's now running a full process node ahead of the rest of the semiconductor industry. That's an unprecedented gap and a fairly recent development; the company only began pulling away from the rest of the industry in 2006, when it launched 65nm. With the help of Mark Bohr, Senior Intel Fellow and the Director of Process Architecture and Integration, this article explains how Intel has managed to pull so far ahead."

Comments

What's the mystery?

[Anonymous Coward]

Andy Grove paid billions to get access to Area 51 alien technology back in 1998. What's so hard to understand?

Re:What's the mystery?

[Chrisq]

Andy Grove paid billions to get access to Area 51 alien technology back in 1998. What's so hard to understand?

Ah its that chip from the android that came from the future. What could possibly go wrong.

Re:What's the mystery?

[WrongSizeGlass]

Andy Grove paid billions to get access to Area 51 alien technology back in 1998. What's so hard to understand?

Ah its that chip from the android that came from the future. What could possibly go wrong.

It was the chip used by the mother ship in Independence Day that could run the virus from Goldblum's Powerbook. It already had cross platform virtualization technology and was years ahead of its time.

Re:What's the mystery?

[mcgrew]

They're going to need some powerful microscopes to examine a nobot's chip.

Re:What's the mystery?

[jellomizer]

Company A produces a better product then Company B.
Company A has better marketing then Company B.
Company A prices are nearly the same as Company B.

Company A for the Win.

No Conspiracy, No Evil, Their customers want a good product at a fair price, That is what they provide.
Right before Intel released their CORE processors AMD had a very strong showing. Then Intel released a much better product and they took their #1 spot back and put distance behind their competitor.
Now AMD will need to make a much better product, Market their Product better, and/or Lower their costs.

Re:What's the mystery?

[geekoid]

AMD has never made a better product. Cheaper, yes. The same? briefly, once.

AMD is popular because ti rides on the 'Intel is big and established, therefore evil' wave of culture the exists in people under 30.

AMD is a technical yawn fest.

Re:What's the mystery?

[macromorgan]

Athlon 64 > Pentium 4. AMD has made a clearly better product at least once. Currently I'd say a C60 > Atom, but that's a matter of opinion as the Atom has a smaller power draw to make it compelling depending upon your needs.

Re:What's the mystery?

[davester666]

except the alien's were still running IPv4, I guess for some legacy servers...

Re:What's the mystery?

[dionye]

since the chip from the iphone was destroy with it, when it fail to save steve from the said andriod?

Re:What's the mystery?

[gman003]

Uh, there's a reason many Linux/BSD distros refer to their x86-64 port as "AMD64" - AMD *invented* the 64-bit x86 extension. And they've always been just slightly ahead of the curve on multi-core - for quite some time, AMD had viable dual-core desktop processors while Intel's were nigh-unusable due to heat and performance.

And there's a reason many supercomputers are built around massive piles of Opterons - AMD makes a superior massive-number-crunching processor.

Not to mention that the classic Athlon was flat-out *better* than the Pentium III. Even as early as the K5, AMD had technically-superior designs held back by implementation issues. Just like Bulldozer, come to think of it. It's a very *interesting* design, and I'm not entirely convinced the rather obvious shortcomings are due to faulty design work, rather than faulty production work.

Oh, and the Fusion "APUs" are great low/middle-end laptop chips. Far, far better integrated graphics with comparable CPU performance and power draw, compared to Intel's offerings. If I were to buy a laptop for standard home usage, I'd grab one of those.

Then there's the whole graphics thing. Sure, you could argue that's more ATI than AMD, but they're definitely beating Intel in the graphics market, that's for sure.

Re:What's the mystery?

[triffid_98]

I would agree. If you want a low power CPU/GPU combination AMD is the clear leader with their line of 'APU' chips, assuming you want decent hardware accelerated video.

And of course, price/performance AMD has been ahead of Intel any number of times.

Re:What's the mystery?

[Anonymous Coward]

actually IvyBridge integrated graphic is much better than any of AMD integrated "APUs" so not even this is true anymore, AMD sucks on CPU front (great GPU on the other hand) i only hope they don't disappear from market because without competition we will be using Ivy Bridge for another 100 years at least

also if you check performance/$ (not clock/$) you will see that Intel has lead here also

Re:What's the mystery?

[Tore S B]

Actually, Intel has a better product because they have more R&D because they have higher revenue because they use some fairly iffy pricing strategies with OEMs to always keep competitors, primarily AMD, off their markets even in the periods when they had a clearly inferior product to AMD.

Re:What's the mystery?

[triffid_98]

At what point did IvyBridge get released for the Atom?

Because I'm pretty sure that hasn't happened yet. At the point that it does we can compare benchmarks, but it would be hard to compare them to something that doesn't currently exist.

I also didn't compare performance/dollar currently, I just said it had happened multiple times, which it absolutely has.

Hell, it's happened even fairly recently. (620 = under $100 for a decent quad core CPU in 2009, even the cheapest i5 is still around $200 now, and was obviously more expensive at the time)

Re:What's the mystery?

[jonadab]

Historically, there were other examples as well. I distinctly recall a period of about six months during the 32-bit era (back before x64 existed, when Intel still thought Itanium might eventually take off) when AMD CPUs were absolutely slaughtering Intel ones on the price/performance ratio.

How come Apple

[Anonymous Coward]

doesn't have any of this advanced stuff Intel has, yet selling so well and making so much money?

Re:How come Apple

[Pieroxy]

Let's see. How come GM doesn't have these tires factories that Firestone and Michelin has? I wonder... GM is making a heck of a lot more money than Firestone though...

Re:How come Apple

[Anonymous Coward]

Not this decade they aren't.

Re:How come Apple

[petermgreen]

Apple takes techology designed and manufactured elsewhere (though they may make some minor tweaks), combines it with their own software and turns it into slick products that people are prepared to pay a premium for. That requires design and marketing prowess but not any great technicalcapability.

Re:How come Apple

[Jeff DeMaagd]

This decade or previous decade? The 2000's were pretty sad, but the 2010s GM has shown a pretty robust recovery.

The example doesn't really fly though, GM still makes a lot of its own parts, they haven't farmed out the core product.

Re:How come Apple

[Anonymous Coward]

How come Nestlé doesn't have any of this advanced stuff Intel has, yet selling so well and making so much money?

Re:How come Apple

[drinkypoo]

The example doesn't really fly though, GM still makes a lot of its own parts, they haven't farmed out the core product.

What do they actually make themselves aside from bodies, blocks, and differentials/axles? Do they even make all that themselves any more? Are they still casting in this country?

Re:How come Apple

[afidel]

I'm not sure about GM but I know Ford is no longer casting around here, they recently started tearing down the 60+ year old casting plant in Brooke Park. The reasons for the plants demise is that it's an ironworks and Ford basically doesn't use a cast iron block any longer and they don't view block casting as a core area and so there was no way they were going to invest the massive amount of capital it would have taken to move the plant over the casting aluminium.

Re:How come Apple

[LDAPMAN]

Are you sure you didn't mean Dell? Apple designs mobile processors, batteries, machined aluminum cases, custom chips of various kinds and I'm sure a lot more that I'm missing. Sure they use some off the shelf parts but to say they don't design any of unique parts is just wrong.

Re:How come Apple

[Anonymous Coward]

Apple does have some of it. It's in all of the Macs they sell.

Re:How come Apple

[wavedeform]

Umm... total BS.

They design their own circuit boards, many of the ICs on those circuit boards, the industrial / mechanical design for all the products, the OS and other software, etc. They are at least as technical as any other manufacturer, and more than most.

Hmmm.

[Anonymous Coward]

Industrial spying worked for Intel. That is why.

Re:Hmmm.

[Anonymous Coward]

Spying... on their competitors who are all years behind them?

You must be pretty high up in the CIA to have thought of such a genius spying scheme.

Re:Hmmm.

[Anonymous Coward]

I think the bigger issue is that Intel has large sums of money to spend on it and when they lose out they just use it to pay people not to use the competition's superior chips. Which seems to be a winning strategy for them. Of course AMD doing some really stupid things also helped.

Re:Hmmm.

[Sable Drakon]

AMD's been doing stupid crap for years. Like the original six-core Phenom II chips and the fact MSAA is still a feature that shouldn't be used on their graphics cards. It's going to be a sad day when AMD crumbles, but it is coming.

Re:Hmmm.

[Anonymous Coward]

There was only one stepping of the six-core Phenom II chips so what would be the point of comparison?

Re:Hmmm.

[Pinky's Brain]

Can't really mark up MSAA to anything AMD did, more what they didn't do ... they didn't throw as much money at devrel as NVIDIA.

Because of Tim Sweeney's anti-PC stance and the universal adoption of his engine DX10.1/11 truly portable MSAA implementations have been hugely delayed and most devs have chosen to go with NVIDIA's free on site developers and their DX9 MSAA hack. AMD has been able to get this to work on their hardware, but it's artificially restricted to NVIDIA hardware without hacked binaries (which is NVIDIA's right of course, even if it isn't very nice).

Now this is no excuse for customers ... if you value MSAA in unreal engine TWIMTP games NVIDIA is unbeatable, but I don't think it's right to say AMD was stupid. Devrel as an anti-competitive force came into force rather suddenly and it simply took them time to adapt, they have been able to get some pretty high profile games away from TWIMTP recently.

Apple is not a semiconductor company

[tanveer1979]

Apple is a product company. It designs its products, and then someone else makes it.. Many components like the processor are third party, and companies like APPLE design a system around it.
After that the design goes to samsung, and its manufactured by samsung. I think samsung uses TSMC fab.

So if apple wanted to have a 22nm chip it could
1. Build a Fab(invest many billions)
2. Pay TSMC and partner with them in tech (invest some billions).

Return on investment may not justify the cost.

As you go smaller, you do gain an area and cost advantage, but you also run into lot of issues related to physics. So 28->22nm is not easy, and its really commendable Intel has done it.

Re:Apple is not a semiconductor company

[Intrepid imaginaut]

Indeed, there's a reason fabs are seen as national treasures.

Re:Apple is not a semiconductor company

[Kjella]

As you go smaller, you do gain an area and cost advantage, but you also run into lot of issues related to physics. So 28->22nm is not easy, and its really commendable Intel has done it.

There's also the limited competition to TSMC, since nobody but Intel has access to Intel's plants (partnering with FPGA companies don't count) the rest of the market "has to" go with them even if they're behind Intel. Their competition is GloFo and UMC, none of which are impressing much. So what if Intel has 22nm? AMD still has to buy from TSMC. nVidia still has to buy from TSMC. Apple still has to buy from TSMC. They simply don't feel the pressure that their customers do, they sell and make a profit anyway. Meanwhile Intel is shipping a 160mm^2 IB to compete with a 315mm^2 Bulldozer for a huge cost advantage.

Re:Apple is not a semiconductor company

[Anonymous Coward]

I think samsung uses TSMC fab.

Samsung makes its own chips.

Re:Apple is not a semiconductor company

[moss45]

Apple is not a semiconductor company

They've bought both Anobit and P.A. Semi, doesn't that make them a semiconductor company?

Apple already co-invests with building fabrication plants through extremely large pre-purchases, and some analysts think Apple is already directly funding Samsung's A6 fab. You are right about Apple being a product company, they don't want to fund these factories. But they are forced to, because that is the current price of staying ahead of the competition in the phone industry.

These fabs aren't an optional investment, they are the cost of doing business.

Re:Apple is not a semiconductor company

[rimcrazy]

Let me say a few words here as I worked in the semiconductor industry for over 28 years. So you fully understand just what it means to make a semiconductor foundry these days, here is a thought experiment for you I worked a few years back.

1) You want to build facility for manufacturing wigit.
2) That facility will cost you between 3b to 5b dollars.
3) In order to justify the ROI on that facility you need to take at least 5% total world wide market share for that wigit
4) You get to scrap your factory in 3 years.

My numbers may be a little outdated today but that only means my cost projections are too low as well as the total market share. From simply an accounting standpoint this is nuts. When I got into the business in the early 70's there were hundreds and hundreds of fabrication facilities. Every start-up had it's own fab. Today you can count the premier companies that have fabs on maybe 1 hand and the total number of significant players in the semiconductor market with their own fabs on both hands.

Intel deserves very high kudo's for what they have accomplished. The risk they take is enormous but they demonstrate time and time again what a manufacturing powerhouse they really are.

Re:Apple is not a semiconductor company

[Anonymous Coward]

Anobit and P.A. Semi are fabless companies. They design chips but do not manufacture them.

Re:Apple is not a semiconductor company

[Anonymous Coward]

Intel's really in another league.

There's a reason Intel shows yields on a log scale. Hint: it's not because it's low. The rest of the industry is reasonably happy at 50-70% yields (and everyone knows it since every buyer sees the yields on the chips its buying). That's why Intel dominates. Getting to smaller feature sizes means they can make smaller die, which means more die per wafer, which means cheaper CPUs. People arguing over little performance gain are missing the fact that going from 32nm to 22nm means Intel just cut their costs in half. But did you notice the price getting cut in half? Nope.

There's another huge advantage having your own fab: turnaround time. When I worked at HP and HP had its own fab, turn around times (from tape-out to parts back) was 2-3 weeks. Later, fabbing a chip at an external fab, it was about 2 months, and you had to pay a lot to get that. Time is extremely valuable, and fabless companies are stuck waiting for chips. Because fabs are optimizing their own time, not their customers, because that's how they are paid.

Re:Apple is not a semiconductor company

[Anonymous Coward]

4) You get to scrap your factory in 3 years.

Why would you scrap it? Just because 16nm would come along, doesn't mean 22 stops being useful. I'm sure you can sell capacity to third-parties once something better comes along if they don't care about getting the latest and greatest. There are chips everywhere (cars, washing machines, stoves) that can happily be produced on an "older" technology without issue.

Whether you scrap it or keep it would simply be another accounting decision. Once you've sunk the cost, I would think that OpEx would be fairly predictable, and so extra cash flow from third-parties wouldn't hurt one's bottom line.

Re:Apple is not a semiconductor company

[unixisc]

Doesn't matter. A lot of semiconductor companies design, but avoid the cost of owning and running their own fabs. Apple's ownership of PA Semi and now Anobit was a pretty low profile entry into the semiconductor market.

But even there, they never took the initiative in developing their own architecture. PA Semi was doing a PPC, and Apple could easily have taken that and had them put together a roadmap that the Mac platform needed. Instead, the team was converted to an ARM team. Apple's role in semiconductors, despite being a founding member of AIM, has been mediocre.

Funding certain fabs is just Apple's way of locking capacity, so that in an allocation market, guys like TSMC or Samsung don't tell them that other customers are ahead of them in terms of allocation. But aside from that, when times are bad and demand low, the last thing any semiconductor company wants is ownership of fabs.

Re:Apple is not a semiconductor company

[unixisc]

You don't notice the price getting cut in half, b'cos you're the end user. When Intel sells to motherboard manufacturers, such as Asus, Gigabyte and all those hundreds of Chinese and Taiwanese, it's all the time price negotiations. In order to build in periodic price reductions, these die shrinks need to happen - it's not just Intel's margins that are supported by it.

I agree about the advantages of owning one's own fabs. It's not just the turnaround time, but when a company owns its own fabs, then in order to achieve a solution, both design and process tweaks can be employed to achieve it. For a fabless company, more often than not, it's up to design, and if the fab is on good terms w/ the customer, the process engineers may be willing to co-operate, but if they throw in the towel and tell their management that supporting a different customer instead of this one would mean better yields w/o any process changes, they're more likely to prevail. With a company that owns its own fabs, process guys would have the same goals as the design guys, since they belong to the same company.

Re:Apple is not a semiconductor company

[unixisc]

I think he may have meant doing an accounting write-down, so that the fab is depreciated, and anything produced by it is a windfall. This is a result of depreciating annually the new equipment that gets installed in a fab, and is typically used for 1 or 2 product nodes. Once most production has moved to a new process, the company can either use this node to produce something low tech and cheap, whose cost would be supported by the fact that the fabs are depreciated, or, if no such opportunity is there, they can upgrade some or all of their equipment and restart that cycle.

Usually, you do have 2, maybe 3 nodes of process technology, but beyond that, there does come a point when the node becomes ancient, and unless you're making something like power supply ICs, it does need to get scrapped.

Re:Apple is not a semiconductor company

[afidel]

Saying Samsung does x is always going to be wrong because Samsung is HUGE and so they will inevitably be doing just about everything. Samsung has their own fabs but other divisions are free to source chips from wherever they can get the best deal or product that best meets their need (if I remember correctly the phone division was pissed at the memory division a few years back because they couldn't buy at the same cost as Apple despite the fact that it was an internal sale).

Re:Apple is not a semiconductor company

[geekoid]

The factory isn't scrapped every 3 years. What are you, the Drama Queen of the semiconductor industry?

Re:Apple is not a semiconductor company

[Nethemas the Great]

I think we'll also see it start to become somewhat irrelevant. We're running to the edge of what is possible with traditional silicon both in terms of physics and manufacturing methods. Unless Intel has a proverbial rabbit to pull out of the hat the head start they have will soon be removed by TSMC et al.. when the financials incentivise the creation of a new fab. The roadmap they forecast for the next decade is indeed dependent upon those rabbits and their not being cost prohibitive at that.

Re:Apple is not a semiconductor company

[Anonymous Coward]

Apple is a product company

Right, but who said it was? I don't understand the point of your comment.

Re:Apple is not a semiconductor company

[ChrisMaple]

Anybody with enough money can buy the Cadence software and call themselves a fabless semiconductor company. If you aren't at least part-owner of a fab, you have almost no influence on the direction fab technology takes, which is what this thread is all about.

Re:Apple is not a semiconductor company

[ChrisMaple]

You get to scrap your factory in 3 years.

There is plenty of demand for ICs built on non-leading-edge technology. For instance, On Semiconductor has custom foundry services at 0.18u, 0.25u, 0.35u, 0.6u, and 0.7u. 0.35 micron is about 14 years old now, IIRC.

Re:Apple is not a semiconductor company

[Anonymous Coward]

You do know that the cost of materials for a processor or gpu are a tiny fraction of their cost right? Cost of materials to build a processor is $10. It's the R&D brain work that costs a lot of money.

Re:Apple is not a semiconductor company

[unixisc]

The key thing is whether they design the chips for their own use, or for sale to other manufacturers. Freescale, NVIDIA, AMD, et al design the ICs, and after they are manufactured, these companies sell them to OEMs. Apple, OTOH, designs them purely for in-house use - you don't see them selling A5 and A6 to the likes of HTC, Motorola, LG and so on. That's why they're not a semiconductor company any more than a company that hires developers to write in-house software that they don't plan to sell or distribute is a software company.

Re:Apple is not a semiconductor company

[Anonymous Coward]

The cost per unit may be cut in half (once yields stabilize), but the difference is what paid for the fab in the first place. The prices won't go down until the R&D on the fab is paid for.

dom

Re:Apple is not a semiconductor company

[Anonymous Coward]

$10 ?? There is no way it can be that expensive, the materials is much less than a dollar, and the whole manufacturing cost is well below $10. Research on the other hand can easily be $100 per chip, especially if they end up not selling very well.

Re:Apple is not a semiconductor company

[AdamHaun]

There is plenty of demand for ICs built on non-leading-edge technology. For instance, On Semiconductor has custom foundry services at 0.18u, 0.25u, 0.35u, 0.6u, and 0.7u. 0.35 micron is about 14 years old now, IIRC.

To elaborate on this, older/larger processes have several advantages:

* Power consumption -- smaller transistors are leaky.
* Startup cost -- mask sets on new processes are insanely expensive.
* Yield -- older processes have better yield, which means more good die per wafer.
* Quality -- older processes usually have fewer problems, and the test software has been updated to catch more defects. This is particularly important in safety-critical markets like automotive, aerospace, and medical.
* Some products are pad-limited, which means that the number of external connections forces a minimum die size larger than the space needed for the core transistors. In this case, shrinking the core only costs extra money.

Re:Apple is not a semiconductor company

[Kjella]

You do know that the cost of materials for a processor or gpu are a tiny fraction of their cost right? Cost of materials to build a processor is $10. It's the R&D brain work that costs a lot of money.

The materials themselves perhaps, but each wafer also takes up processing plant time which is an extremely expensive commodity. With half as many chips per wafer you need double the number of production lines to get even. Also smaller chips improve yields because you have to throw away less of the wafer due to defects. This is one of the reasons the original Atom was a smashing commercial success for Intel, it cost them almost nothing to produce yet sold for a good price.

Re:Apple is not a semiconductor company

[Anonymous Coward]

After that the design goes to samsung, and its manufactured by samsung. I think samsung uses TSMC fab.

Samsung owns its own fabs, and is a competitor to TSMC in the foundry business.

It would actually be a pretty big deal if Apple shifted some business from Samsung to TSMC, which is why there was a lot of attention paid a year ago or so when there were rumors that Apple had taped out a chip with TSMC. TSMC is the dominant merchant foundry, and Samsung is (in that market) a relative upstart.

Re:Apple is not a semiconductor company

[Anonymous Coward]

AMD still has to buy from TSMC. nVidia still has to buy from TSMC. Apple still has to buy from TSMC.

Apple (currently) buys from Samsung, not TSMC.

(Parenthetical is there only because as I pointed out in another post, there were rumors that they taped out a chip with TSMC. If true, nothing has shipped yet -- all current production A4, A5, and A5X silicon for iPod/iPhone/iPad/ATV is Samsung, some 45nm, some 32nm.)

Re:Apple is not a semiconductor company

[Anonymous Coward]

The rest of the industry is reasonably happy at 50-70% yields (and everyone knows it since every buyer sees the yields on the chips its buying).

Not always true. I work at a fabless semi company and we don't usually get told what the yield is. Normally, it's an internal responsibility for the fab partner to make sure their yields are good enough that they can (a) turn a profit at the negotiated price per good part delivered to us and (b) meet the negotiated production volume targets in the contract.

So long as things are going well, they'll never tell you what the yield is. Even if things are going badly, they still might not say. It's treated as proprietary information.

why?

[crutchy]

"wintel"

why will it eventually be overtaken by samsung?

"winfail"

Too bad their 22nm 3D failed

[Anonymous Coward]

The shrink from 22 to 32nm is a staggering size change - 33% finer lithography - and it uses their much-hyped 3D transistor technology on top of things. Yet, Ivy Bridge, being just a shrink of the older Sandy Bridge die, shows no improvements over the 32nm version. Traditionally, Intel has always been able to show lower power consumption and more than a tangible performance improvement when just doing a process shrink, but the Ivy Bridge does nothing extra in terms of performance and consumes not lower power than its older 32nm sibling - and let's not mention the inefficient heat packaging causing temperatures hotter than the 32nm Sandy Bridge. There's a problem here, Intel.

Re:Too bad their 22nm 3D failed

[DarkTempes]

Is this true or just trolling?

Every benchmark I've seen so far has shown performance increases and power consumption decreases at around the same price.
If your statement is true then that suggests a lot of review sites out there are spoofing their results and that's very very bad.

Sure, if you have a sandy bridge chip there isn't a whole lot of reason to upgrade because for most users anything made in the last 5 years or so can handle everything you'd want to throw at it.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Bullcrap. You just have no idea what you're talking about.

The 22nm chip has significant power advantages attributed to the shrink. It also has around 10% better CPU performance at that lower power, so it's significantly better perf/watt. GPU performance is far increased. And it is 75% the die size of Sandy Bridge, with 20% more transistors (mostly gone into large increases in GPU performance)

Re:Too bad their 22nm 3D failed

[Big Hairy Ian]

Is this true or just trolling?

From what I understand it's about 30% more efficient I don't know about it being any faster though. I also thought it could power down core's independently.

Re:Too bad their 22nm 3D failed

[SuricouRaven]

That's in part because they put the extra die space freed up to a new purpose: Graphics performance. If you just look at processor performance, Ivy is no better. Benchmark the inbuilt graphics and it's far ahead.

Of course, anyone who actually needs decent graphics wouldn't be using the on-chip graphics anyway, so I question just how useful this really is.

Re:Too bad their 22nm 3D failed

[Kjella]

Traditionally, Intel has always been able to show lower power consumption and more than a tangible performance improvement when just doing a process shrink, but the Ivy Bridge does nothing extra in terms of performance and consumes not lower power than its older 32nm sibling

Is there any reason the parent is at +4, Interesting and not -1, Troll? Are the AMD fanbois really so desperate that they have to mod up blatant lies? Ivy Bridge uses 25-30W lower power at stock speed to deliver marginally better than SB CPU performance and considerably better (but still crappy) GPU performance. The only people that whine are those who want a 4.5+ GHz overclock. Anandtech called it quite possibly the strongest tick [Intel] has ever put forth, but I guess if you don't like reality you can invent your own.

Re:Too bad their 22nm 3D failed

[macraig]

You missed this one, I guess.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Most of the complaints are tiered towards the overclocking space; doesn't go as far as the old sandy bridge.

http://www.overclockers.com/intel-i7-3770k-ivy-bridge-cpu-review /
http://www.overclockers.com/ivy-bridge-temperatures

http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review /
http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge

etc.

Re:Too bad their 22nm 3D failed

[macraig]

It's true, more or less.

Re:Too bad their 22nm 3D failed

[jkflying]

That's just for overclocking. Regular usage is improved.

Re:Too bad their 22nm 3D failed

[Goragoth]

And you seem to have missed the part where "running hotter than SandyBridge" applies only to overclocking. Yes, IB is a worse overclocker than SB, but under normal conditions IvyBridge is faster and uses less power than SandyBridge. Remember that overclockers are a tiny portion of the market. IvyBridge isn't the amazing revolutionary chip some people were expecting but it is a successful, evolutionary step forward. Just like most processor generations.

Re:Too bad their 22nm 3D failed

[nataleesmith480]

how about http://www.tabletdirectsale.com/ this site

Re:Too bad their 22nm 3D failed

[QQBoss]

The shrink from 22 to 32nm is a staggering size change - 33% finer lithography - and it uses their much-hyped 3D transistor technology on top of things. Yet, Ivy Bridge, being just a shrink of the older Sandy Bridge die, shows no improvements over the 32nm version. Traditionally, Intel has always been able to show lower power consumption and more than a tangible performance improvement when just doing a process shrink, but the Ivy Bridge does nothing extra in terms of performance and consumes not lower power than its older 32nm sibling - and let's not mention the inefficient heat packaging causing temperatures hotter than the 32nm Sandy Bridge. There's a problem here, Intel.

While I will accept you reversed some numbers (the shrink was from 32 to 22, not the other way around) and Intel is using tri-gate transistors, most everything else you describe is just flat out wrong. Ivy Bridge DOES show lower power consumption at stock voltages (TDPs of 77W vs 95W are a testament to that), and it is higher performance at the lower power consumption (though not by huge amounts, nor was it intended to be). Since it is lower power than Sandy Bridge at the same frequency, it is not having any issues related to thermals and packaging.

Now, if you want to rant about the fact that it doesn't handle overvoltage well for overclocking purposes, that is fine, but it is a separate discussion compared to stock. What you are seeing now is that Intel (probably extremely wisely for the market they are chasing most heavily) has tuned in their process node for stock voltages, but this is resulting in very leaky transistors at high voltages. Additionally, while the current packaging has the ability to remove heat just fine at stock voltages, when you start leaking too much the heat builds up too quickly- which certainly is a 22nm node issue and not actually a packaging issue. Quite possibly, though how far in the future I can't begin to guess, they will probably tweak the process for the Extreme Edition CPUS to make them handle an overclock without leaking so much, but that will take some time learning how they can play with the various knobs to get what they want without destroying what they need.

This leaves me with the feeling that the only problem here is your expectations of a CPU that was manufactured with the intent of taking the mobile market by storm (and they have tuned the process properly for that) when what you want is an overclocking king. Let's see how they tune the process technology for the Extreme Edition (and hopefully copy into other desktop-bound CPUS) before any decisions are made that they have screwed the pooch on being able to overclock.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

why the hell did this get marked intersting?

1. if it was "just" a shrink over 32nm and had the same die size and power consumption, it would be no improvement. instead the die size is quite smaller than 32nm. past shrinks have held the die size only slightly smaller as they either added more cache or other logic in addition to the shirnk. in this turn things are mostly the same resulting in a smaller die. this means more dies per wafer and, eventually, lower cost (noting a move to 450mm wafers within the next 5 years).

2. ivy bridge does have lower power consumption than sandy bridge. same speed chips doing the same task can have the ivy bridge system using up to 30w less power at peak performance. where it hasn't really improved is in 'idle' conditions. why? ecause most tech to turn of various sections of the cpu were introduced in other process shrinks. as this is gennerally "just" a shrink, the idle power being roughly the same isn't about lack of improvement, but moreover less innovation in existing silicon power saving techniques. i mean the chip already slows it's overall speed down when idling, and will shutdown significant portions of it cache as well.

3. there has only been a performance improvement for various definitions of "performance". in other iterations intel has tweaked the chip. this is based on feedback from the chips being in the field. of course with sandy bridge running so effectively, there hasn't been much call to 'tweak' the design. still at the same speed, ivy bridge has better "power" performance than sandy bridge.

4. i don't think you've seen any benchmarks to be honest. graphics has improved significantly, cpu not so much. but it is still an improvement over snady bridge.

5. the packaging problems have only been encountered when people overclock the chip. that is, people running it out of spec. sure there's 1% of us out there that will push the limits. but for most people, they will never ever see those hotter temperatures. personally i go the opposite direction by undervolting to consumer less power and in turn producing less noise.

Re:Too bad their 22nm 3D failed

[macraig]

I'm getting real sick of dogmatic groupthinkish people being allowed to moderate here. The result is bullshit like this post being punished as Troll, when effectively what he's done is reference another recent Slashdot article . I had already commented here before the real trolls masquerading as moderators leapt in here, so I can't do my part to correct the small injustice; I hope some proper meta-moderation takes place.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Actually, the overclocking would've been no worse if they had used thermal soldier to dissipate heat. This is a case of Intel cutting corners because they can - AMD is far behind, and by their cost-cutting preventing high overclocking on Ivy Bridge they'll sell more Haswells next year.

Re:Too bad their 22nm 3D failed

[QQBoss]

It is being punished as a troll, because it is wrong in annoyingly misleading ways. It probably would not have been punished as a troll if the GP had said "in the area of overclocking, 22nm is a fail because..." By making a sweeping comment that only applies to a very small subset of the market, potentially informative because troll. The mods got it right.

Re:Too bad their 22nm 3D failed

[QQBoss]

grrrr.... because -> becomes

Re:Too bad their 22nm 3D failed

[macraig]

Fine, then mark the previous article - or the editor who approved it - as Troll, and leave this guy the hell alone for being misled by it.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

From what I understand it's about 30% more efficient I don't know about it being any faster though. I also thought it could power down core's independently.

The ability to power down cores (without the apostrophe!) independently is not related to transistor size or transistor design.

Re:Too bad their 22nm 3D failed

[dkf]

Of course, anyone who actually needs decent graphics wouldn't be using the on-chip graphics anyway, so I question just how useful this really is.

There's a whole world of people who would quite like decent graphics, but who don't want to spring another hundred bucks or two to get something fancy. There's also the mobile market (laptops, tablets, etc.) where fitting an extra graphics card looks more like a liability than a good thing. Overall, it looks to me like a smart area for Intel to pitch their transistor budget at.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

1. if it was "just" a shrink over 32nm and had the same die size and power consumption, it would be no improvement. instead the die size is quite smaller than 32nm.

The complete die is quite smaller than a single transistor on it? Not that's an achievement! ;-)

Re:Too bad their 22nm 3D failed

[QQBoss]

Given that the previous article you cite clearly calls out that this is an overclocking issue, "When overclocked, Ivy Bridge runs as much as 20C hotter than its Sandy Bridge predecessor at the same speed, despite the fact that the two chips have comparable power consumption", I believe that the correct troll is self-identified.

Re:Too bad their 22nm 3D failed

[Calos]

Yep, totally not his/her fault for spreading lies and not bothering to read the several articles linked to the previous /. story or the commentary attached to the /. article. S/he deserves being modded up, plus some bonus karma for being such a victim, and not having lies modded down so that they stop sustaining themselves is a necessary sacrifice. The fact that s/he is posting as AC and will not suffer/benefit beyond that one post notwithstanding.

Sorry. I think I need more coffee.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

There is nothing on the Ivy Bridge die that is "far increased" in terms of performance. NONE of the almost dozen of tests have been able to show anything remarkable. Not in CPU perf, not in GPU perf, not in lowered power consumption. We're talking about loads of tests that show only a measly few watts lower TDP, a measly few percentiles extra of CPU perf, and a measly few extra percentiles of GPU perf - nothing even scratches on "large increases". Something went wrong here. This isn't what we're used to from Intel when they do a massive shrink like this.

Re:Too bad their 22nm 3D failed

[Theophany]

I think you've completely missed the point.

1. Ivybridge is a die shrink, nothing more nothing less. Everybody who really thought it would be lightyears ahead of Sandybridge in terms of performance was simply deluded. It's a new die size that anybody has yet to perfect, that will come in Haswell.

2. All this ire is unfairly directed at Intel. On the basis that AMD seems to have no idea what it's doing at the moment, Intel can relax and do as they please. If you want to be pissed at anybody, be pissed at AMD for not being anywhere near competitive and pushing Intel to continuously raise their game.

Since 1998 I've only ever used AMD CPUs in my builds. When I came to build a new rig in February, I simply couldn't justify buying AMD for my CPU again because they were so far behind and there was zero indication that Bulldozer would rectify that. It's sad watching them busily engage in killing themselves off as a serious desktop CPU manufacturer and leaving Intel to potentially become lazy and overpriced, but that isn't Intel's fault.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Is there any reason the parent is at +4, Interesting and not -1, Troll? Are the AMD fanbois really so desperate that they have to mod up blatant lies?

Looks like the AMD fans are faster than the Intel sockpuppets. Although the latter seem to be more persistent, what with getting paid for it and all ;)

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review/

GPU performance is 60% increased, in some cases.

Re:Too bad their 22nm 3D failed

[macraig]

Just because a post might be incorrect doesn't warrant moderating it as Troll. Just leave it unmoderated. Nowhere did I advocate a positive moderation.

Re:Too bad their 22nm 3D failed

[macraig]

Just because a post might be incorrect doesn't warrant moderating it as Troll. Just leave it unmoderated.

Re:Too bad their 22nm 3D failed

[dpilot]

> but that isn't Intel's fault.

Actually it is, to at least some extent. Go back a few years, when Intel was making misstep after misstep, and AMD was coming on gangbusters with K8. At that point, Intel had missed the market so badly that had they been AMD they would have gone under. They weren't AMD, they were Chipzilla. AMD enjoyed a good product cycle with K8, until Intel managed to come back. But they didn't enjoy the great product cycle they should have. Their great product cycle was turned into a merely good product cycle because Chipzilla twisted a few arms and kept K8 out of key opportunities.

The other piece of reality is that Intel combines first-rate process technology with first-rate design capability. (I say "capability" because more than once they've shown themselves to be very capable of letting their eye off the ball, design-wise.)

AMDs biggest problems have always been financing and less-than-best process technology. Bulldozer is a misstep, agreed. But it's not a misstep of the degree of Netburst or IA64. Had K8 gotten the success it deserved, AMD would have been better able to properly fund their design shop. That wouldn't have helped their process problems, however.

The simple fact is that the way things are today, Intel can afford to screw up badly, and can recover. None of their competitors can.

Re:Too bad their 22nm 3D failed

[macraig]

Whose ire exactly are we talking about here? It's not mine. I've used mostly AMD processors myself for the last 15 years, but that will probably change because of the better power and thermal characteristics of the Sandy Bridge series. My next system from scratch will likely host an Intel CPU.

As far as the parent of my original comment, I think he probably read that earlier article BEFORE someone corrected the summary to qualify that it involved overclocking (the comments make clear it wasn't initially so). Someone who posts a comment that is merely incorrect doesn't warrant the post being moderated as Troll. The correct moderation is no moderation at all. Some idiot(s) let his emotional investment in some idea/company distort his response to that post.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Yes, but how many of these people who don't want to spring $100 on fancy graphics want to spring $2-300 on their CPU?

Re:Too bad their 22nm 3D failed

[Jeff DeMaagd]

When you've mistaken overclocking for normal operation, I think you've either missed the boat or you're not being honest with your agenda.

IB is hotter in overclocking, but at recommended settings, it is cooler and a little faster than SB. If you only care about overclocking, that's fine, but to not say that much is not being honest to us.

Or did you not even bother to understand the article you're referencing?

Re:Too bad their 22nm 3D failed

[h4rr4r]

Overrated I would think.
There needs to be a factually wrong mod. The lack of one causes a lot of problems.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Intel has always hidden behind the "We're leading the world in marginal graphics" mantra. The real fact is that Intel Mfg is really the only part of Intel that is worth discussing. Their IAG group is your typical large corporation disaster: they go for quantity over quality in everything, from products to engineers. They spend a fortune in equipment and processes that wouldn't be needed except that their management is pushed to show a large % of overseas labor utilization, but that labor isn't terribly useful unless you hold their hands or spend a fortune in equipment to enable them to work cluelessly but to give the rest of us some oversight.

It's a complete clusterfuck, and although Intel probably has all the elements they require to make a good high end graphics chip, and the software to make it work, they can't create anything new and make it successful because they have completely lost their engineering ability.

(And no, I have never worked for AMD)

Re:Too bad their 22nm 3D failed

[drinkypoo]

When you've mistaken overclocking for clocking differently from how the processor was marked for economic reasons, you've missed the boat, too.

Intel marks perfectly good processors down to lower speeds intentionally because people won't buy them at the top price point. But they still make the top price point available to people who have more money than time to dick with overclocking. And there's enough crap parts in the channel (that won't OC well) to keep some people buying the expensive chips.

How the processor performs when overclocked (or really, not factory-underclocked) provides an extremely revealing look at business practices and yields.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

I know it has come up before in discussions about GPGPU that a lot of smaller tasks aren't worth off-loading to a discrete GPU that could compute them more or less instantly because the transfer time is so expensive. Maybe an on-die GPU could be useful as a math coprocessor even when it is too slow for actual graphics work? And that only makes sense if it is on every CPU as it is now.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

"1. Ivybridge is a die shrink, nothing more nothing less"

Yeah, that tri-gate transistor stuff is just hocus pocus.
Intel has more money and more smart folks.

Re:Too bad their 22nm 3D failed

[Cid Highwind]

"How the processor performs when overclocked (or really, not factory-underclocked) provides an extremely revealing look at business practices and yields."

It's an interesting look to be sure, but considering the IB chips perform fine and use less power than SB at their marked clock speeds, there's nothing you can derive from that look to back up "the Ivy Bridge does nothing extra in terms of performance and consumes not lower power than its older 32nm sibling" in the post you appear to be defending here...

Re:Too bad their 22nm 3D failed

[Cid Highwind]

Not to mention the damned-if-they-do-damned-if-they-don't aspect of that line of reasoning. If the 22nm parts don't overclock as well as the predecessors, their node transition "failed". If they do, Big Bad Chipzilla is gouging us by artificially restricting the supply of high-clock parts.

Re:Too bad their 22nm 3D failed

[geekoid]

misstep after misstep? no. The made the mistake of changing the processor to that stop slot in the Pentium 3, and a small release of chips with a rare rounding issue. I don't want to seem like I'm downplaying the problem, it was a critical error and embarrassment for Intel, but it only happen in rare and specific situation, and the recalled it and replaced like a responsible company should.

"Had K8 gotten the success it deserved, "
it got all the success it deserved.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Hmmm... looks like someone else working in VCG reads /.

Re:Too bad their 22nm 3D failed

[geekoid]

How it's worded will also dictate is something is wrong. All the bold letters in the world won't change that fact.

"You mother is over weight" - Observed fact
"You're mother is a fat ass lard beast" - Observed fact by a troll
"You're mother is thin" Incorrect
"You're mother is a thin whore" Incorrect and a Troll.
"You're mothers children aren't as smart as people would like them to be" - Observed Fact
"You're mothers Children are a bunch of drooling potato heads" - Fact and troll.

Today's lesson was brought to you be the letters F and U.

Re:Too bad their 22nm 3D failed

[geekoid]

It would be interesting if actual experts could have a special account, and they could mod +1 factual -1 wrong in the area of their expertise
IT would be too expensive for /. to do, but it would be interesting.

Hell, I'd like to see an experiment just to see f having a bonafide experts modding would change the conversation to a more productive one?

Re:Too bad their 22nm 3D failed

[timeOday]

Framerates on games are about 50% higher than on Sandy Bridge. I know, hardcore gamers don't care. But now that the bulk of the market is laptops, being able to play most current games acceptably without discrete graphics is a very good thing.

Re:Too bad their 22nm 3D failed

[geekoid]

People should be punished for for spreading lies and commenting in a factual way about things they don't know anything about. This, let people go unpunished for spreading shit need to end. Because it's not only that person, it;s the people who read the comments. The poster is putting shit in the swimming pool.

Re:Too bad their 22nm 3D failed

[geekoid]

That's completly wrong.

Bottom line: the gap from rated to actual use is shrinking, a lot.
The fct that some people want to run the chip outside performance spec is fine,l but don't go crying to intel when it doesn't work the way you want it to.

I can add nitrus to my Dodge, but when the engine gets damaged, IT's not Dodges fault they they don't design the ability to do it into the Caravans engine.

"Intel marks perfectly good processors down to lower speeds intentionally because people won't buy them at the top price point."
The makes no sense at all. Did you even think about the sentence? The cost is in the manufacturing. Lowering it's capabilities only adds more costs to the process.
What really happens is that if more then a certain percentage in a run can' run at speed X reliable, the slow it down and test again. Repeat until percentage is low enough to sell. I know, boring and reasonable business reasons. Sorry to shit are your conspiracy theory, and you feeling that you are entitled to be able to over clock chips with no bed effects.

Re:Too bad their 22nm 3D failed

[dpilot]

I'm suggesting that the original sans-L2-cache Celeron, the entire NetBurst architecture, and IA64 were all pretty bad missteps.

Chipzilla twisted a lot of arms to hinder and delay K8 adoption in the marketplace. It should have done better.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

Not to diminish Intel engineering, but I detect a bit of cheerleading and penis waving (ironically in the diminutive sense) about 22nm pioneer etc.
Whether it's Intel, TSMC, Global Foundry...much of the technology and equipment are from the same general group of manufacturers, outside of the fab per se (Applied Materials, Zeiss ...).
TSMC probably have no real need to be on the bleeding edge. Cheaper to wait for all the kinks to be worked out before committing.
There is always a cost penalty if you desire to penis wave. It also may not be much of an improvement over the older process.

Re:Too bad their 22nm 3D failed

[mcgrew]

Someone who posts a comment that is merely incorrect doesn't warrant the post being moderated as Troll. The correct moderation is no moderation at all.

No, the correct moderation for a factually incorrect comment is "overrated". The correct moderation for this (my) comment is "redundant" because someone else said the exact same thing you did that I quoted above, and I told him the same thing I'm telling you.

You shouldn't have to worry about your karma unless you're a troll, a shill, unpleasant, or an idiot, because an occasional downmod won't hurt otherwise. If your karma is bad, you probably don't belong here.

Re:Too bad their 22nm 3D failed

[mcgrew]

No, a factually incorrect comment should be modded down so I don't have to waste my time reading it. Mod it "overrated" rather than "troll".

Re:Too bad their 22nm 3D failed

[macraig]

You must be new here. Moderation != peer review. You've made it clear that's the delusion you live in, but it's not the observable fact. Moderation is politics and emotion and groupthink most of the time.

Re:Too bad their 22nm 3D failed

[TheRealMindChild]

Try even a low end AMD APU platform. I got a board, "APU", 4GB of ram, for $120 at tigerdirect. I can play Portal 2 at fantastic settings with smooth play, and just about every other game I throw at it

Re:Too bad their 22nm 3D failed

[macraig]

No, the correct moderation for a factually incorrect comment is "overrated".

Agreed. Moderation winds up smelling a lot more like politics than peer review, though. Thus someone moderates him Troll instead.

Re:Too bad their 22nm 3D failed

[Nethemas the Great]

Not exactly true but also not exactly false. They're packing just over 400M additional transistors on it. Most of which are being dedicated to graphics processing. Some minor tweaks to the cores give a very modest improvement over a Sandy Bridge with the same clock. When Intel packed the cores into a much smaller space simply to make room for the larger on-board graphics unit they actually shot themselves in the foot a bit. Very similar thermal loads are having to be displaced through a much smaller area. This is causing these processors to run significantly hotter and one of the main reasons why the overclocking ability of this chip is pretty poor. While this won't mean much to Joe Bloggs today, it does spell out trouble for the traditional speed improvements you'd normally see down the road.

I'm sure some bean counter somewhere thought beefing up the on-board graphics would boost revenue somehow but as a performance enthusiast all I see is a huge waste of transistors at the expense of future performance increases. The $40 entry-level discrete graphics card still easily outclasses what's on-board. This chip was built with intentions of cheap all-in-one PCs and small form factor devices not the performance desktop.

Re:Too bad their 22nm 3D failed

[macraig]

Agreed. Moderation winds up smelling a lot more like politics than peer review, though. Thus someone moderates him Troll instead.

Re:Too bad their 22nm 3D failed

[ChrisMaple]

Netburst, deep pipelines, rambus. Not misstep after misstep, but an explosion of missteps all at once. They were bad strategic errors. The floating point error was of a different nature; it was a design error and a verification error.

Re:Too bad their 22nm 3D failed

[RightSaidFred99]

It by more or less you mean not at all. Graphics are upwards of 50% faster, CPU performance is a fair amount better, and it uses a fair amount less power.

This release is half about cheaper/faster/lower power CPUs and half about getting on their feet with an entirely new process. They will better utilize the new 22nm trigate process in the CPUs upcoming next year.

Re:Too bad their 22nm 3D failed

[RightSaidFred99]

You're flat out lying. GPU performance is _considerably_ increased. CPU performance is somewhat increased and it uses less power.

You just don't know anything about processors and seem to think a die shrink release is going to be some magical, super-advanced leap from the processor which was the un-shrunk predecessor.

Re:Too bad their 22nm 3D failed

[RightSaidFred99]

You mean the low end AMD APU platform that Ivy Bridge outperforms graphically? That one? No thanks, I won't try it.

Re:Too bad their 22nm 3D failed

[gstrickler]

Apparently, you didn't read the comments. No, it's not true, not even close.

Re:Too bad their 22nm 3D failed

[TheRealMindChild]

You are wrong. A fast google brought up this bit (though sandy bridge, same idea) -> http://www.youtube.com/watch?v=XqBk0uHrxII

Also, you would spend $200 more on your ivy bridge board/cpu + you still need to buy ram. If your 5fps is worth that much money, my heart goes out to your family

Re:Too bad their 22nm 3D failed

[RightSaidFred99]

Sandy bridge is not Ivy Bridge. Ivy bridge does well compared to most AMD APUs and is suitable for casual gaming on a laptop or reasonably sized monitor.

Re:Too bad their 22nm 3D failed

[StikyPad]

I don't know about modding the Intel CPU, but I removed the heat spreader on my nVidia GPU (GTX460) and replaced the shite putty they used with some quality paste which lowered temps by 20C, and it wasn't terribly difficult. Tedious and delicate, yes, but not difficult. (Though to be fair, it appeared that the putty had contracted as it dried/cured/whatever, and was not making good contact between surfaces, which apparently was a relatively common problem.)

Re:Too bad their 22nm 3D failed

[TheRealMindChild]

Fair enough. All I see other than pretty graphs is everyone saying ivy bridge is comparable in performance to sandy bridge.

Re:Too bad their 22nm 3D failed

[SecurityTheatre]

"Had K8 gotten the success it deserved, "
it got all the success it deserved.

K8 sales would have doubled from the agreement they made with Dell, alone.

Intel strong-armed Dell into cancelling that contract AFTER it had already been signed.

They've since lost court battles and settled with AMD for a lot of money, but it almost doesn't matter at this point.

Re:Too bad their 22nm 3D failed

[gstrickler]

You seem to have forgotten a few missteps.

The Pentium IV "NetBurst" architecture. Dead end.

The P3 1.13GHz disaster. Unstable, and unable to keep up with AMDs chips.

The Itanic, er, Itanium, repeated delays, failure to perform to expectations, complete recall of first model, marketing failure, etc.

Larabee, or any Intel graphics prior to Sandy Bridge.

Re:Too bad their 22nm 3D failed

[Calos]

I didn't say mod it troll, either. Though it is sometimes difficult to distinguish between someone who is adequately (mis|dis)-informed and someone who is feigning it. Mod it Overrated straight to -1 for all I care. + gives bullshit too much credibility.

Re:Too bad their 22nm 3D failed

[Calos]

So, wait - you're complaining that the GP was wrongly down-modded due to "groupthink;" geekoid and myself opine that it's not necessarily bad to punish factually incorrect posts for being factually incorrect; and your response is that moderation should not be peer review either, and that because the system is not perfect, no one should try?

You don't want posts modded by merit or basis in reality, and you don't want posts modded by popular perception. What do you want? No moderation at all? A new moderation for "+1 Precious Special Snowflake?"

Re:Too bad their 22nm 3D failed

[Calos]

Sigh, /. ate part of my comment as html markup.

That was supposed to read +{integer} {adjective != "funny"} gives bullshit too much credibility.

Mea culpa.

Re:Too bad their 22nm 3D failed

[gstrickler]

80% of users don't need a high performance GPU, and in fact won't benefit from one at all. I'm an example of such a user. I can always use more CPU speed, but I don't even own any software than can get my laptop's discrete GPU anywhere near it's potential. And, since I'm on a 4.5yr old machine, even the HD 3000 graphics in Sandy Bridge would be a significant GPU performance increase, that still wouldn't be used.

But most importantly, if you look at the reviews, IB can maintain playable framerates at medium or high quality settings in most games. It's really only hard core gamers, or people doing work that can benefit from GPU acceleration that need a discrete GPU with SB or IB.

So, thank you for your opinion that is irrelevant to 90% of the market.

Re:Too bad their 22nm 3D failed

[Anonymous Coward]

What you are seeing now is that Intel (probably extremely wisely for the market they are chasing most heavily) has tuned in their process node for stock voltages, but this is resulting in very leaky transistors at high voltages.

Actually, that's not quite what's going on. It's more like the new tech in Intel 22nm dramatically reduces the performance penalty of running at stock and lower voltages. Look at the graphs in figure #2 here:

http://www.realworldtech.com/page.cfm?ArticleID=RWT050511195446

The right hand graph is what I'm talking about. Less gate delay leads to higher clock frequencies, so the 2 curves show that for both 32nm and 22nm, as voltage increases, so does performance. (From seeing this in other contexts, apparently the gray curve between 32nm and 22nm shows a hypothetical 22nm with the same basic transistor design as 32nm rather than the new 3D "trigate" transistors.)

Note that while 22nm is always faster, the slope of the curve is closer to horizontal at every point. This means Intel 22nm performance is relatively insensitive to changes in voltage. That's actually a good thing -- it means you can reduce power (by reducing voltage) without hurting performance nearly as much. But it also means that a favorite overclocker knob is no longer very effective.

Additionally, while the current packaging has the ability to remove heat just fine at stock voltages, when you start leaking too much the heat builds up too quickly- which certainly is a 22nm node issue

Increasing the voltage does not reduce the ability of the packaging to conduct heat away. It just generates more heat to conduct away. And there is nothing inherent to Intel 22nm which causes heat to be trapped or "build up too quickly". The heat flow from active circuits to the heatsink is pretty much the same as it ever was -- circuits to die body to IHS to heatsink.

IMO, the main reason for higher temperatures is the increase in power density (power per unit area). Even though the TDP has gone down, die area has gone down more, and higher W/mm^2 = higher temperature. That's a truism which is independent of process (you can come up with ways to consume lots of watts in very little area in any process), so it's greatly influenced by a particular chip's design tradeoffs (how much power do you try to pack in per area?). But when you shrink an existing design (and IVB is mostly a shrink of SB, plus an improved GPU) you can pretty much count on increased power density by definition -- you're fitting the same design into a smaller area.

Also, overclockers are too accustomed to using the voltage knob to improve performance, and they're probably trying to push IVB voltages far above what they really should aim for. It's never going to give them as much benefit as before, and like all process node advances, 22nm shifts maximum safe voltages down somewhat.

Re:Too bad their 22nm 3D failed

[BootysnapChristAlive]

Greetings, my friend. Why have you not returned to Gamemakerdom yet?

Use Gamemaker. Use it right now. Why not use Gamemaker. There is simply no reason to not use Gamemaker. Time to switch to Gamemaker. Why not make the switch to Gamemaker? There's no reason for you to not use Gamemaker. None at all.

Return... to Gamemakerdom right this minuteness!

Intel makes for awesome Linux boxes.

[MnemonicMan]

Intel, with their open-source graphics stack, makes for some of the easiest-to-maintain Linux boxes around. I'm typing this right now on Arch with Intel graphics. Sure, they don't have a lot of "gaming punch" but they are darn stable and just work with Linux.

My desktop right now has Windows and is running a first-generation Core i5 with an AMD Radeon 6870 added in. When that machine get's replaced with another gaming Windows machine in a year or two I'll be pulling the AMD graphics out of it and running on the i5 integrated Intel graphics. It will be super-low-maintenance in Linux. None of this rebuilding fglrx or nVidia modules every time you upgrade the kernel.

When I go looking for a Linux machine the very first thing I look to check-off is "Intel graphics"? Yup, then it's a buy.

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

Typing this on my Arch laptop, my desktop is Windows..

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

Do you work for Intel's marketing department?

Re:Intel makes for awesome Linux boxes.

[Malvineous]

When I go looking for a Linux machine the very first thing I look to check-off is "Intel graphics"? Yup, then it's a buy.

I've decided to do the same with my next PC, after growing tired of the lacking multimonitor support under nVidia (sure it's usable, but still very buggy.) But what are Intel graphics like with multiple monitors? I currently have four screens connected via two nVidia discrete cards (no onboard graphics) so how could I achieve this with Intel? The fabled DisplayPort daisy-chaining hasn't seem to have materialised, and the one-to-many DP adapters are still stuck at very low resolutions.

Re:Intel makes for awesome Linux boxes.

[spyked]

That wouldn't make sense. I mean what marketing person would use "Linux machine" and "graphics" in the same sentence?

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

Ha, nope. Anyone who's been around the Linux block a few times should confirm that Intel graphics drivers (xf86-video-intel) are just the shiznit. Try getting fglrx to run with xorg-server 1.12.1 - what Arch ships - and you'll find that that proprietary driver hasn't been updated to support that version yet. You could always use xf86-video-ati or xf86-video-nouveau but, honestly, both those drivers lack the polish of xf86-video-intel. Intel just works.

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

I'm running a dual-head using XRandR. It works fine. Honestly however I've never used anything more complicated than a dual-screen so I don't want to go off and promise you the moon. Xinerama may be an option for you: before using XRandR I set up with that and even though it's not supposed to compiz did work when in that mode. XRandR, Xinerama, TwinView, you'd think between all that something would work for you.

Re:Intel makes for awesome Linux boxes.

[Issarlk]

Okayyyy... So you'll use intel integrated graphics in your next windows gaming machine that'll run on Linux? That sounds like masochism ; do you actually play some games or is Mine sweeper and Tetris all you need to run on your rig?

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

No, when I buy my next Windows machine it'll have either AMD or nVidia graphics. My current desktop machine, which has AMD graphics, will have those graphics pulled at that time. Which means the machine will then be using its Intel graphics - what is integrated right now on the processor. That machine, the old one when I get my new one, will be Linux. With Intel graphics. And I don't play games under Linux, that's what Windows is for, which will be the new machine - in a year or two.

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

One of the touted new features of Ivy Bridge is support for triple head displays, so four screens is probably going to require an additional graphics card.

Re:Intel makes for awesome Linux boxes.

[Bert64]

There are open source drivers for radeon too, they might not perform as well as the closed drivers but they still outperform most intel cards while being just as convenient, plus you have the option of using closed drivers if you want the extra performance.

Re:Intel makes for awesome Linux boxes.

[pr0nbot]

Just for the record... I've used an Ubuntu box for a 5+ years of WoW and now SWTOR, under wine. The proprietary nvidia drivers are easy to install and work well under Linux, and have coped with my moves from 7600 to 9600GT to 460.

Is it masochism? Yes, a bit, but I've not really got much use for a Windows box aside from gaming, so I thought I'd give it a go.

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

OMG....GaMiNg!!!!!! I'm off to raid a troll with all my friends!!!!! ("friends")

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

The last time I tried the open-source radeon driver it was with an Ati Xpress 1100 chipset. The driver would randomly garble the screen and there was nothing I could do to fix it. Things have perhaps changed but I've never had any issues with Intel.

On Arch, at this time, the X.Org server - as already stated, because the X server is too new - isn't supported yet by the latest AMD proprietary driver.

Re:Intel makes for awesome Linux boxes.

[Ed Avis]

It's a shame there are no add-in PCI Express cards with Intel's graphics hardware (perhaps a couple of low-end Core processors with the CPU part disabled and just the GPU running). I would love to use Intel graphics instead of Nvidia but if you have several monitors it's not possible - unless you can now get a motherboard with two or four video outputs.

Re:Intel makes for awesome Linux boxes.

[Malvineous]

XRandR, Xinerama, TwinView, you'd think between all that something would work for you.

Thanks for the reply. I'm currently using Xinerama (without TwinView, as some of my screens are in portrait mode) with the "Awesome" window manager which works fine (apart from the nVidia bugs like OpenGL freezing on one monitor until you switch virtual desktop on another monitor), but I was thinking more of the physical connection for four screens. AFAIK Intel don't make discrete graphics cards so it sounds like I'd only be able to go triple-head before I have to pick from one of the closed source camps ("closed source" referring to officially supported manufacturer's drivers.)

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

I have a laptop, it has its built in screen and a VGA port on the side. Here is it's xorg.conf file:

XRandR xorg.conf.

I did briefly use a Xinerama configuration, and here that is:

Xinerama xorg.conf.

I'm using Xfce 4.10, with the Xinerama config compiz worked fine. Later on someone told me it wasn't supposed to work with Xinerama. Huh, it did.. But, anyway I went with the XRandR config anyway because it is much shorter. However, on my login manager screen - SLiM - the XRandR config has both screens as clones of each other where with the Xinerama config they are independent displays.

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

I have a MythTV box with integrated AMD HD3200 graphics running Arch, the last time I updated it the opensource drivers which I was using just wouldn't work. I had to switch to the proprietary ones, which are also buggy, but they do at least work.

As much as I hate to say it, AMDs linux drivers are buggy pieces of shit, both the proprietary and open source ones. Which is a shame because I would like to support AMD by buying another AMD-based machine, but I don't want the headaches trying to get working graphics when I update.

And on the other hand my netbook with Intel graphics which I have been using for nearly 4 years hasn't given my the slightest trouble with its graphics.

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

If that miniscule amount of convenience is the dominating requirement when chosing a GPU, you either don't actually need a GPU or you get paid for your opinin :D

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

Then you end up buying 50x as many machines when you build your render farm. Good luck with that. Both ATI and nVidia have had Linux support for several years and consistently many times faster performance. I suppose when all you want is bash, and vi, Intel runs great.

Re:Intel makes for awesome Linux boxes.

[Ed Avis]

Thanks - yes a decent laptop can drive an external display as well as the builtin screen. I am thinking more of desktops with two, four or even eight DVI outputs.

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

You're positing a straw-man. I am not, and never mentioned, "running a render farm." I'm using it for desktop use and programming. I have a full GUI desktop, Xfce 4.10, with Compiz enabled and doing just fine. Flash video plays perfect in my Chromium browser. Intel is fine.

Re:Intel makes for awesome Linux boxes.

[jawtheshark]

I still use the Ati Xpress 1100 on one of my laptops. Given that there are no more proprietary drivers for it and the open source drivers are, ehm, "interesting", using Ubuntu with anything but Unity-2D is impossible. Intel graphics in comparison are lighting speed. The garbling on the Xpress 1100 seems to be gone now, though (Ubuntu 12.04). Keep in mind, the Ati Xpress 1100 can run Half Life 2 in low settings on Windows XP, so really to display a desktop.....

However, Intel doesn't get a free pass either. I have an Atom based machine with Intel GMA 3150 and I think it has problems with some OpenGL things. For example, Cheese doesn't work at all, while it does just fine with my work laptop (NVidia-based) and the same webcam. (again Ubuntu 12.04)

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

Yup, most laptops come with an extra display hook-up of some kind. My laptop isn't used in a portable manner, it is plugged all the time and doesn't even have its battery in it. Since it is immobile that makes it convenient to also have the other monitor plugged into it constantly. Now, if it was mobile then the extra screen would be a burden more than something useful. If that was the case then I probably wouldn't go with the xorg.conf file and would instead have two shell scripts that used the "xrandr" command directly: one to enable dual-screen and one to disable dual-screen, which would be the default. I wouldn't have the xorg.conf file because on the go I might not have an extra monitor that that file makes X expect.

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

The "garbling" I experienced was also on Ubuntu, I don't remember exactly which version but 10.04 wouldn't be far out of the park. I remember trying everything to get it to work, like enabling the X-Org Edgers PPA even. Just wouldn't work. That's the thing with proprietary drivers: when the manufacturer stops supporting it then it dies on Linux because other pieces of the system, like X, continue to change. Eventually you get an ABI breakage and you have to either stick with older software all-around or junk the affected hardware. When that hardware is your video that may not be an option - so older software it is, bugs, security issues, and all.

Re:Intel makes for awesome Linux boxes.

[fa2k]

Intel, with their open-source graphics stack, makes for some of the easiest-to-maintain Linux boxes around. I'm typing this right now on Arch with Intel graphics. Sure, they don't have a lot of "gaming punch" but they are darn stable and just work with Linux.

If you don't need gaming performance, you can go with anything on the market. I know that the AMD open source drivers are very stable and support Compiz-like effects, and the same is probably true for NVidia.

Re:Intel makes for awesome Linux boxes.

[jawtheshark]

I'm not disagreeing. If I want to continue this machine, I might have to switch back to Windows... Wouldn't that be ironic? ;-)

Re:Intel makes for awesome Linux boxes.

[drinkypoo]

Don't work at all on R690M. Never buying anything with ATI in it again unless it's dirt cheap used, and NO MORE LAPTOPS WITH ATI GRAPHICS EVAR. It's only fairly recently that the open source drivers don't actually choke on all the various Rage, Rage Pro, etc chips. I know because I have three old laptops with 'em and they always gave me lots of grief. Don't get me started on the radeon driver.

I hope one day that the OSS video drivers are worth a damn, but until then, there's nVidia. Well, and intel, but seriously, they're slow at 2d.

Re:Intel makes for awesome Linux boxes.

[MnemonicMan]

Sadly that is exactly what I did with my machine that had the Xpress 1100. On the bottom of the computer was a Genuine Vista Product Key Sticker. So, I pulled out some Vista medium I had lying around and installed it. AMD's legacy drivers do support Vista, just not the latest Catalyst. So, like Catalyst 10.2 installed on it and it works fine. I gave it to my nephew. ;)

Re:Intel makes for awesome Linux boxes.

[hobarrera]

This is quite true; Intel is the way to go if you don't want to game, no doubt about that. You even do get some pretty good 3D and graphics acceleration for movies and stuff as well.
Nvidia+binary blob is the way to go if you *DO* want to game, and you'll probably want a distro that includes the binary blob to avoid manual installation issues. (like arch).

Re:Intel makes for awesome Linux boxes.

[petermgreen]

hmm, I have an older macbook with intel graphics and i've had quite a bit of trouble with xorg on it. I've found a version that mostly works and i'm scared to upgrade further for fear of breaking it. Maybe i'm just unlucky.

Not that ATI and NVIDIA are great either, back when I last worked with an ATI card in a linux box FGLRX had a nasty habbit of crashing the system when it couldn't find it's module and the nvidia linux binary driver wouldn't work with my latest card (an EVGA GT 430 dual DVI, interestingly the EVGA GT 430 DVI+VGA my brother has works fine) meaning I had to fall back to using nouveau (I play my 3D games under windows so it's not critical but it certainly irks me).

Having been burnt by all three vendors i'm not sure what i'll do for my next machine.

Re:Intel makes for awesome Linux boxes.

[timeOday]

I had to chuck a pretty good laptop (IBM T60) because ATI stopped making drivers for the video card entirely - windows or linux.

Re:Intel makes for awesome Linux boxes.

[geekoid]

Just so you know, the new on chip video stuff is pretty damn good.
Soon it will be like the sound chip. Good for almost everything.
This is the same thing we have gone through many times.
Networking:
Card
on board but sucks, use card
Chip price falls
on board pretty good, use card for network gaming
Corporation stop using the card.
Mobo manufacture highers network chip expert.
On board good for everything except some extreme situation.

Same this for sound,
Same thing will happen to graphics.
There will be the really high end stuff, but for the vast number of users on board will become good enough.

Yes, I recognize that I could go out and drop 25 bucks on a better network card and drop my gaming latency 15% But my latency for most games is about 30.
So making it 26 wont really improve my game much. If I go back into hard crore tournamant play, I would spend the money.

Re:Intel makes for awesome Linux boxes.

[jawtheshark]

Ouch. At least, mine came with XP Media Center Edition. It was cheap too, because it was on sale just before Vista was released. Before anybody thought Vista could be a dud. ;-)

Re:Intel makes for awesome Linux boxes.

[RalphTheWonderLlama]

The fabled DisplayPort daisy-chaining hasn't seem to have materialised, and the one-to-many DP adapters are still stuck at very low resolutions.

AMD Eyefinity does this, right?

Re:Intel makes for awesome Linux boxes.

[Mad Merlin]

FWIW, the GTX 680 supports up to four displays per (single GPU) card. I haven't tried it yet, but I would expect it to work much better than the 4 monitors across 2 cards support that one was limited to in the past.

Multimonitor across multiple cards has never been stellar in any OS, probably because it's a very niche use case, but multimonitor on a single card is a very common use case.

Re:Intel makes for awesome Linux boxes.

[LDAPMAN]

Have you ever seen a Mac Pro with multiple cards? It's flawless. Ive built advertising/media displays with up to 16 monitors. Never an issue.

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

AMD dropped support for the integrated GPU on my CentOS box a couple of years ago, and I'm finally so fed up with the problems of keeping the old drivers working that I'm thinking of just sticking the cheapest Nvidia card I can find in the machine to replace it.

Re:Intel makes for awesome Linux boxes.

[Malvineous]

Multicard multimonitors work fine (probably on most platforms) for simple things like advertising displays that only run one or two programs, but as soon as you try to do anything complex - full screen single-screen OpenGL, multimonitor OpenGL, changing screen resolution, etc. it quickly becomes apparent there's not a huge amount of testing going on for that kind of set up. Application support is lacking too. Only a few of the more popular apps let you select which screen to use when you want to go full screen for example - you'd think it'd stick to whichever monitor the window was on, but usually it just uses whichever screen has pixel (0,0) on it. And good luck if you want a full screen program spread across only a subset of your monitors...

Re:Intel makes for awesome Linux boxes.

[Malvineous]

AMD Eyefinity does this, right?

No, last I checked the cards have six DisplayPort sockets on the back. But there's a big price premium for them, and it is much cheaper to buy multiple dualhead cards. It's a moot point though, according to the FAQ Linux support is only in the planning stage so it's not really an option for Linux PCs.

However if DP daisychaining was to work, presumably you'd need two sockets on each monitor - DP in and DP out/passthrough. I have two screens with DisplayPort inputs on them, but neither of them have an out/passthrough socket so I don't think I could daisychain them anyway.

Re:Intel makes for awesome Linux boxes.

[RalphTheWonderLlama]

I think if your monitors had the passthrough ports then all you would need would be 2 DisplayPorts on the video card.

I did the How To at http://www.amd.com/us/products/technologies/amd-eyefinity-technology/Pages/eyefinity.aspx
and it recommends a 6850 or 6950 (I have) for a 4 monitor setup.

Were you thinking of the 5870 Eyefinity Edition which actually has 6 DisplayPort outputs? I believe the 6000 series and up have the newer DisplayPort for daisychaining.

This guy used a 6950 with 4 odd monitors - http://www.uberreview.com/2012/03/amd-6950-eyefinity-review-and-how-to.htm
But he used all the ports instead of daisychaining DisplayPort only.

Here they say 6 monitors are good with 2 DisplayPorts if you have daisychaining monitors.
http://www.tomshardware.com/forum/309846-33-eyefinity-monitors
"DP 1.2 monitors have 2+ ports, with which you can connect up to 3 monitors in serial. Two ports on the card, so 6 monitors that way. You can do this only with native DisplayPort 1.2 supportive monitors. No DP-DVI dongles work. "

Re:Intel makes for awesome Linux boxes.

[Anonymous Coward]

Multicard multimonitors work fine (probably on most platforms) for simple things like advertising displays that only run one or two programs, but as soon as you try to do anything complex - full screen single-screen OpenGL, multimonitor OpenGL, changing screen resolution, etc. it quickly becomes apparent there's not a huge amount of testing going on for that kind of set up.

Have you actually tried this with a Mac? You'll find there are usually fewer issues.

My understanding it's about a basic choice in system design. The Mac GL driver stack is always structured as a common GL API layer (written by Apple), with backend plugins for HW acceleration and SW rendering. By contrast, it's common on Linux and Windows for the HW driver to be a monolithic unit responsible for implementing all of GL. (IIRC, this is true on Linux if you use the proprietary vendor blob drivers.)

By introducing extra layering to their API/driver stack, Apple sacrifices some performance, but gains a cleaner system design which they have more control over. Which in turn means they can do a better job of supporting things like an arbitrary number of displays, GL-accelerated windows spanning different displays possibly on different cards possibly made by different manufacturers, etc. They don't violate the layered model needed to cleanly handle this kind of thing in a general way.

(Not saying it's perfect by any means, but generally you expect multi-display problems on the Mac to be things like "application does a stupid thing because it was never tested that way", not "drivers prevent it from being possible at all".)

Process the new Speed

[Anonymous Coward]

Ok, so, they happen to be ahead on the lithography. And it's yielding nice kit, to be sure. But in a sense it's the new megahertz race: It means the design itself isn't as important as it should be. And that's something intel actually has trouble with, qv itanic.

TI has their own fabs

[Anonymous Coward]

The article is wrong, TI has their on fabs, they don't outsource manufacturing as the article states.

Re:TI has their own fabs

[AdamHaun]

I work at TI. We do have our own fabs, but we also outsource manufacturing to foundries too. The new 65nm flash process I work on was developed at TSMC, and all manufacturing will be done there. I know that other processes run in TSMC as well, but I'm not sure which ones (we have a lot).

Re:TI has their own fabs

[Anonymous Coward]

I work at TI. We do have our own fabs, but we also outsource manufacturing to foundries too. The new 65nm flash process I work on was developed at TSMC, and all manufacturing will be done there. I know that other processes run in TSMC as well, but I'm not sure which ones (we have a lot).

IIRC a few years ago TI got off the train of developing new logic process nodes (at 65nm I think?) and committed to a future roadmap of outsourcing logic to the likes of TSMC, UMC, Samsung, etc. (Basically, like many other companies, it had become too expensive for TI to stay in that particular race.) So, in the long run, for high performance and/or dense logic TI is fabless -- you guys can still build things in a relatively advanced fab which you own, but every year it gets more out of date. (and is therefore less and less able to be competitive in certain market segments, e.g. you probably wouldn't want to fab a cellphone SoC on TI's logic process these days.)

Maybe there's an Area52, in Tel Aviv

[Taco Cowboy]

Remember Pentium M?

Intel had to rely on Pentium M to pull itself out of that big sink hole back then

Re:Maybe there's an Area52, in Tel Aviv

[Anonymous Coward]

There is actually an Area 52, but it's close to 51. I think it's an artillery test range or somesuch.

Re:Maybe there's an Area52, in Tel Aviv

[Hamsterdan]

And use its monopoly so OEMs would not use AMD processors

Where are the cores?

[Anonymous Coward]

While a 22nm fabrication process is all well and good, why haven't Intel bothered to release their 80-core, teraflop chip that they touted some 6-7 years ago and said that we would have in our computers by 2011? http://techresearch.intel.com/ProjectDetails.aspx?Id=151

Re:Where are the cores?

[garrettg84]

One word: necessity. They simply haven't had to. Besides render/compute farms, we don't really need that kind of power. Most of what may have been pushed towards those massively parallel processors is now being pushed towards the GFX cards - where they DO have more than 80 cores (think CUDA, greater than 512 cores now). Most of the games out there still struggle to use more than a few actual processor threads. Some problems are linear and simply can't have more cores thrown at them for faster work.

Re:Where are the cores?

[zaft]

It's under development... http://www.intel.com/content/www/us/en/architecture-and-technology/many-integrated-core/intel-many-integrated-core-architecture.html

Re:Where are the cores?

[Anonymous Coward]

While a 22nm fabrication process is all well and good, why haven't Intel bothered to release their 80-core, teraflop chip that they touted some 6-7 years ago and said that we would have in our computers by 2011?

http://techresearch.intel.com/ProjectDetails.aspx?Id=151

Because they never said the teraflop chip was going to be a commercial product at all? Learn to read, it was basically the equivalent of an academic research project. I have no idea where you get this 'in the PC by 2011' thing from, certainly not from anything Intel ever said about that chip.

(Nonetheless, you're probably getting some of the fruits of that research whenever you use a modern Intel chip -- the research project was mainly for the purposes of investigating the best, most power efficient methods of interconnecting many cores on a chip. I'd note that shipping Intel chips don't work much like the teraflop chip. Probably one outcome of the research project was that the approach tried wasn't actually all that great. Sometimes research tells you what you shouldn't do...)

Because Intel has the best astroturfers?

[Anonymous Coward]

Gotta love the absolutely objective slashvertisements...

Shill here
shill there
everywhere a shill
lalalala...

Competing with ARM...

[Bert64]

A while ago Intel stated they were intending to keep at least one process shrink ahead of everyone else because it was the only way they could compete with ARM...

Personally i find this despicable and extremely arrogant, pushing their own inferior architecture and holding everyone back when they could be making ARM chips that were superior to everyone else's. Recent benchmarks show their latest low power atom chips are barely competitive with last year's ARM designs (and wouldnt be competitive at all if built on the same process)...

Re:Competing with ARM...

[msgmonkey]

This is called business, using whatever advantage you have to compete against a competitor. Last time I checked Intel was a business.

Re:Competing with ARM...

[Anonymous Coward]

That's not new: Intel was able to beat RISCs thanks to its superior manufacturing process.
The benchmarks of Intel "lndian phone" has shown that Intel is now in the same category as ARM: I expect that Intel's next generation will beat ARM in performance/power, thanks to its better fab process.
Competition is about earning money not "fighting fair"..

Re:Competing with ARM...

[serviscope_minor]

pushing their own inferior architecture and holding everyone back

Oh brother, not this again.

On the low end, Intel will never beat ARM because of the large, expensive instruction decoder. That applies to the deep embedded stuff.

Cellphone chips aren't low-end any more. They're getting bigger and bigger and bigger. For big processors, intel does very well, as they have the best OoO scheduling and branch prediction which keeps the large, expensive ALUs busy, giving a very high IPC.

In 5 years time, ARM will need to have similar guts if they want anything but lots of weak cores. At that point, the main advantage that they have will be more or less down in the noise.

Re:Competing with ARM...

[drinkypoo]

Personally i find this despicable and extremely arrogant, pushing their own inferior architecture and holding everyone back when they could be making ARM chips that were superior to everyone else's.

That's because you're not very smart. Intel tried that already. They weren't very good at it. Went back to making AMD64 chips, which they are pretty good at. Now, if you made AMD and intel into one company, besides that we'd all be fucked, they'd also make some truly awesome processors.

Re:Competing with ARM...

[Bert64]

Having superior technology and then hobbling it so that its no better (or even inferior) to the competition is not good business...

If they built ARM chips on the smaller fab process they would be able to easily lead the market.
With Atom they are barely competitive, while also being incompatible with everyone else.

Only partially wrong

[tanveer1979]

TI closed a lot of FABS. All TI designed stuff does not exclusively get manufactured in TI fabs. Much of it actually goes to TSMC

Re:Only partially wrong

[Anonymous Coward]

Not true.

In order to move a device to a 3rd party fab, you must prove that the technology does not exist in a TI fab or that TI fabs are at capacity.

Re:Only partially wrong

[Anonymous Coward]

TI has been closing down fabs, but they still have 9 200mm and 2 300mm fabs. They have 2 150mm fabs which are being closed down. TI-beats-estimates--plans-to-close-two-fabs

good thing Intel also does fab for ARM

[ChipMonk]

ARM-based CPU's are out-selling x86 by a fairly hefty margin, thanks to the mobile/embedded market, while the desktop x86 kingdom has been nearly saturated for, well, forever as these things go. And until Intel gets a clue and makes a chipset that renders on-screen for less than 5 times the mA-h required by a comparable ARM, it's going to stay that way.

Based on that, it's only good business sense that Intel brings in ARM business for their fabs.

Re:good thing Intel also does fab for ARM

[Anonymous Coward]

Eh, who does intel fab ARMs for? AFAIK, they only fab their own chips; they had an ARM license (I think they still have it) and made their own ARM lines (StrongARM from DEC, then XScale) back in the day, but they quit in 2006, when ARM's upward progress was seen as threatening x86's planned downward progress to smartphones, MIDs, and UMPCs.

Re:good thing Intel also does fab for ARM

[Anonymous Coward]

But note that you might have to sell a thousand ARM chips to make as much profit as a single 'Extreme' x86 chip. If I remember correctly the ARM-based chips the company I used to work for sold had a profit margin of about a dollar each, whereas high-end Intel CPUs must be hundreds of dollars..

....Money

[malignant_minded]

http://en.wikipedia.org/wiki/AMD_v._Intel
When you get all the sales you get to build facilities that can make you even better toys.

cutting edge is not the market

[JasterBobaMereel]

Note this is manufacturing not design, many other companies design chips

Note this is only cutting edge chips, how many PC's were bought with these cutting edge chips... ?

What are people actually buying, Moderate PC's, a few servers, and a lot of Phones and Tablets .... all of which use FAB methods less advanced than this ...

Yes Intel are ahead, but the rest of the world are not necessarily buying their products

Re:cutting edge is not the market

[the eric conspiracy]

That's ridiculous. When Intel rolls out a new processor arc it puts a whole range of products on it, and adoption of the new cpus is very rapid. The Ivy Bridge cpus are absolutely mainstream designs with integrated GPUs and low power consumption rather than high end designs that sell in small volume.

Intel also introduced a new mobile design that looks extremely interesting - the first processor with vertical transistors. And the smaller feature size and power consumption is a huge win in this market.

Re:cutting edge is not the market

[RightSaidFred99]

Lol. Divorced from reality much? PC and server shipments are up, and continue to go up.

Mark Bohr?

[seven of five]

Any relation to Neils and Aagie? No wonder Intel's headed for the atomic realm.

commendable maybe

[__aaltlg1547]

It's notable. It'll be commendable if the huge investment they made to pull it off gives them an advantage with more than the investment. You can't know that until you've gone to mass production and seen what the problems are.

Fab has become a service

[tanveer1979]

You are right, and you are also right that your numbers are on lower side.
Consider the Auto analogy.
If you start an Auto company, manufacturing common rails, you will mostly buy tech from either Bosch(likely) or Delphi.
you won't start from scratch.

Same deal here. To design a chip does not require more investment.
As you go further down the line, you need more investment

For example
1. Algo development - Chip arcitechture - Basically an algo or an idea
2. RTL (Actual Behavioural model) - Now you need Simulators and verification engineers to make sure your RTL works
3. You want to create a netlist too? - Welcome to synthesis tools ($$++)
4. Place and route - GDS II - Even more!
5. Manufacturing - FAB - Really big cost

Since TSMC has fabs, financial prudence dictates companies use those. Its not an ideal situation. With so few fabs in the market, its a sort of monopoly. Probably a handful of semiconductor companies have enough cash to invest in a fab, even then results won't show for many years(profits)

Re:Fab has become a service

[geekoid]

"Consider the Auto analogy."
No.

Re:Fab has become a service

[Anonymous Coward]

Parent should be modded funny, not insightful.

(This post should be modded insightful... but, well, with a score of 1, so yeah, nvmd.)

Intel is half a node ahead.

[Anonymous Coward]

Hmm, let's readjust this article.

So ... (1) 28nm products were available in early January, 22nm is available now, in May. (2) 28nm is a half-node behind 22nm. (3) TSMC's transistor density at a given node is actually very high - I believe the density of TSMC 40nm was comparable to Intel's 32nm according to diagrams posted at RWT or B3D. (4) 32nm at GlobalFoundries is a year old, and given a typical two year cadence we are again talking about half a node behind.

However ... TSMC aren't making a lot of 28nm still, whereas Intel will be making a lot more 22nm.

Re:Intel is half a node ahead.

[Locutus]

and the reason Intel is pushing more than their server CPUs onto the latest/smallest process is because their CPU design needs it to compete with the ARM designs out there on the "older" process sizes. IIRC, Intel used to run their server and gaming CPUs on the latest/smallest process sizes for over a year before moving their desktop and other CPUs to it. They'd charge a premium for those server and gaming CPUs, a big premium. But now, they are being forced to put their low budget CPUs on the latest process.

So it's nice they spun this as a big plus for Intel but as you mentioned, they are only a little bit ahead of the others and it's costing them by doing this.

LoB

Apple does invest in fabs

[geoffrobinson]

From what I've read they do the following:

They'll give a company a bunch of cash to start building a new chip. They then get that chip exclusively for a period of time at a discount.

They may design some chips but they aren't building any of their own.

Re:Apple does invest in fabs

[Anonymous Coward]

From what I've read they do the following:

They'll give a company a bunch of cash to start building a new chip. They then get that chip exclusively for a period of time at a discount.

No. Apple designs the chips for iDevices in house. These chips do contain licensed IP cores (such as the ARM processors) which get used elsewhere (because the owners of said IP license it to many companies), but the chip design as a whole is Apple's. Only Apple can call up the fab and say "we want you to make 23 million of these, delivered on this schedule, here is a big bag of money".

That's pretty much the standard business model for SoCs like Apple's. You design it, you own it. You don't have to give away the design after some period of time.

The only thing along the lines of what you're saying is that Apple might buy a guarantee of a certain amount of wafer volume from Samsung (which is, so far, Apple's only fab partner). Chip manufacturing is a pipelined (assembly line) process with a long lead time. The industry looks at capacity in terms of "wafer starts" -- each fab has the capacity to start processing only so many wafers per month, and that sets the maximum throughput possible. A customer of Apple's size can use its size and negotiating power to lock up a relatively large percentage of their fab partner's wafer starts.

(But it's not pure might makes right either. While there are advantages to doing business with a goliath, foundries do try to keep smaller customers happy too. That way, they can't be screwed by any single customer deciding to go elsewhere -- the opportunity cost of not running fabs at nearly 100% capacity is frighteningly huge.)

They may design some chips but they aren't building any of their own.

They aren't likely to ever do that either. Apple doesn't need and doesn't want to be in the fab owners' club. Nobody wants to be if their business model doesn't require it.

Answer: Taiwan

[retroworks]

Taiwan is to circuitry what Japan was to radio technology. http://www.intel.com/jobs/china/students/

Intel Development Model

[Bacon Bits]

Two words: Tick Tock.

Intel's development model is Tick (die shrink), Tock (new features). It's been this way for many years. Honestly, I'm not sure why you expected a Tick to have any new features. They did call Ivy Bridge a "tick plus," but even then I wouldn't expect any major overhaul in features or performance. Tick is a manufacturing process improvement, not an architecture improvement.

As far as heat packaging, I believe others have covered that sufficiently.

A full process node ahead of the competition...

[Anonymous Coward]

...and they still suck. The only reason intel is king of performance is they have a vast majority of the market share. I remember reading an article when ivy bridge first debuted and performance tuning gcc for both bulldozer and ivy bridge respectively bulldozer averaged 30 to 40 percent slower. Now let's consider that bulldozer is near EOL and trinity is about to come in with piledriver, which is supposed to be 25% faster in cpu performance narrowing the gap to 5 to 15 percent. The cpu performance coupled with the outstanding gpu performance makes me wonder what's up with intel?

Explained in "Great by Choice"

[seifried]

Go read "Great by choice", and Intels strategy (aka "Intel delivers") is explained, but in a nutshell they realized early on (like in the 70's) it wasn't enough to make good chips, you have to make lots of them, perfectly. So they are heavy into the manufacturing side and making sure it works really really well.

And AMD got kind of an unexpected break

[Sycraft-fu]

So the original Athlon was a shot out of the blue, it was the first AMD chip that really competed with Intel chips. Intel had to stop sadbagging and release faster P3 chips (it was capable of making them just wasn't because it didn't need to). AMD legitimately brought some serious competition. It was badly hamstrung by having horrible, horrible motherboard chipsets, but there you go.

Now the Athlon maintained competitiveness the next generation... Because Intel fucked up. Their Netburst architecture wasn't very good. I don't fault Intel on this, their research showed it would scale really well MHz wise, possibly up to 10GHz, so the slower IPC wouldn't matter. However it didn't, so they had a slower architecture compared to AMD. The problem? AMD wasn't updating. They just kept doing minor rehashed on the same thing.

Then, as you say, Intel dropped Core. They hadn't been standing still, they never do. They corrected the mistakes of Netburst and made a chip that was very fast per clock. AMD was still playing with old tech and Intel pulled way ahead. Then even worse as it continued, Intel kept revising their chip, AMD kept playing with the same basic thing. Their Bulldozer launch got pushed back and back. When it finally did happen recently, it was not at all competitive to Sandy Bridge, and of course Intel now just launched Ivy Bridge.

So AMD's initial competitiveness was no fluke, they dropped a good product. But the length it went on was kinda a fluke, since Intel screwed up, and AMD didn't do anything to work on improving their tech in a big way.

Re:And AMD got kind of an unexpected break

[SecurityTheatre]

It might be worth pointing out that Core wasn't on the roadmap. It was a happy accident.

The design came from the Pentium-M, which was just a rehashed Pentium 3. The P4 "Netburst" was on the roadmap for a decade when it came out, followed by IA-64.

The Pentium M was intended to be a "mobile" chip to put into mid-range laptops where the P4 was too big and hot. The rather unknown Israel design team was put on it and produced a really remarkable product that scaled far better than they expected. As a result, after release, they were put to the task of improving it and re-working it to be a real desktop chip (the Core) and then, because it was still so tiny, the CoreDuo and later the Core2Duo.

Talking about flukes, anyway. Sometimes engineering stems from them. It's not because they did it wrong, it's just how it is.

As far as I know, they are still benefiting from some of the amazing hand-layouts that were done on the Pentium M and early Core chips. Nobody else would even consider doing a manual layout on a modern chip. They had a few people who did just that and it made all the difference.

Re:And AMD got kind of an unexpected break

[Anonymous Coward]

Just FYI the Israel group was unknown for a particularly good reason. They'd spent years working on a SoC implementation based off the P3 and were dumb enough to use an RDRAM controller assuming it'd be ubiquitous by the time the processor came out. As it turned out rambus wasn't that competitive, the memory manufacturers colluded against them over their submarining, and cheap computing history got set back by almost ten years (SoC would've had embedded 810 era gfx, cpu, and memory controller all in one chip.)

Captcha: Postpone.

Re:And AMD got kind of an unexpected break

[Anonymous Coward]

So, to paraphrase you, you mean they created a very small chip and ended up with a very good chip, besides small. Then they added the missing functionality and ended up with something small, good and fully operative.

It sounds like all the agile methodologies evangelization manual.

good silicon, bad architecture

[OrangeTide]

If Intel would have kept their perpetual ARM license, they could rule the world. But even with cutting edge fabs their are going to be overrun by a more ubiquitous CPU architecture.

Intel should stop making x86s. And especially stop the nonsense of trying to use x86s to compete against stream processors in GPUs and HPC. But it is really too late for them, they gave their ARM license to Marvell (who have basically pissed away a good opportunity as well by not aggressively pursuing new designed based on the license)

Re:good silicon, bad architecture

[RightSaidFred99]

The more transistors a processor ends up having, the bigger Intel's advantage will be. This is because they have the manufacturing capacity and the design resources to scale at that level - they have already with their current high end CPU lines.

So tell me, are mobile and tablet CPUs coming out with _more and more_ transistors, or _less and less_? What does that tell you?

It tells me that ARM shouldn't count their eggs before they've hatched.

Re:good silicon, bad architecture

[OrangeTide]

every transistor draws current. if a transistor doesn't provide useful functionality to the user, it is wasted die area.

Competitors?

[Catas]

Intel is not TSMC's nor Globalfoundries' competitor. They are not a foundry business. Also both companies are already establishing 20nm. It is also significantly more difficult to design a fab around hundreds of customer designs and production, rather than the IDM model which shows less versatility in it's processes. I'm not entirely sure what the article is trying to accomplish by skewing facts.

Is this milking?

[thoughtsatthemoment]

There was a high jump champion who was able to break the world record by a few centimeters. But to claim multiple record breakers, he improved his performance by one centimeter at a time. I wonder if Intel is in this position as well. Based on the performance gain for each generation (especially if you consider the ever increasingly bloated software), it does seem like Intel is enjoying "hardware updating" as much as Microsoft is with software updating.