Slashdot Mirror
Moore's Law Blowout Sale Is Ending, Says Broadcom CTO
itwbennett writes "Broadcom Chairman and CTO Henry Samueli has some bad news for you: Moore's Law isn't making chips cheaper anymore because it now requires complicated manufacturing techniques that are so expensive they cancel out the cost savings. Instead of getting more speed, less power consumption and lower cost with each generation, chip makers now have to choose two out of three, Samueli said. He pointed to new techniques such as High-K Metal Gate and FinFET, which have been used in recent years to achieve new so-called process nodes. The most advanced process node on the market, defined by the size of the features on a chip, is due to reach 14 nanometers next year. At levels like that, chip makers need more than traditional manufacturing techniques to achieve the high density, Samueli said. The more dense chips get, the more expensive it will be to make them, he said."
I thought Intel, Samsung and TSMC claim that the upcoming 350mm wafer going to bring along another round of cost saving.
Are they telling the truth, or are they blowing smoke ?
Muchas Gracias, Señor Edward Snowden !
There is no free brunch.
er, LUNCH. Lunch.
The idea that there's going to be a smooth curve between performance, cost and process size/voltage at all points is pretty stupid really.
Used to be you used to have to upgrade every 2 years. Now you really have to upgrade every 5 or 7 years. Once every 10 years sounds pretty good to me. As the pace of computer innovation slows, less money has to go towards upgrades. Computers are now more like appliances, you run them down until they physically break.
Of course if you manufacture computers or work in IT, then such a proposition is horrible as a long product lifecyle means less money coming to you. As a consumer, I like it because I no longer have to shell out hundreds of dollars every other year to keep my computers usable.
Well, we had a good run. 99% of the computing needs of 99% of the people can be met by the existing chips electronics. For most people network and bandwidth limits their ability to do things, not raw computing power or memory. So Moore's observation (it ain't no law) running out of steam is no big deal. Of course the tech companies need to transition from selling shiny new things every two years to a more sedate pace of growth.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
About ten years ago, I went to a talk at Stanford where someone showed that the increasing costs of wafer fabs would make this happen around 2013. We're right on schedule.
Storage can still get cheaper. We can look forward to a few more generations of flash devices. Those don't have to go faster.
The most advanced process node on the market, defined by the size of the features on a chip, is due to reach 14 nanometers next year.
Actually, the "process node" hasn't meant anything for years now.
If that's true, we can only hope that the exponential bloating of software stops as well. Software has been eating the free lunch Moore was providing before it got to the users; the sad reality is that the typical end-user hasn't seen much in the way of performance improvements - in some cases, common tasks are even slower now than 10 years ago.
Oh sure, we defend it by claiming that the software is "good enough" (or will be on tomorrow's computers, anyway), and we justify the bloat by claiming that the software is better in so many other areas like maintainability (it's not), re-usability (it's not), adherence to "design patterns" (regardless of whether they help or hurt), or just "newer software technologies" (I'm looking at you, XAML&WPF), as if the old ones were rusting away.
Yeah, well, we have been in somewhat a flat stage, and will be for a while. Here is a good graphic.
http://www.economist.com/news/21589080-golden-rule-microchips-appears-be-coming-end-no-moore
ummmm.... Moore's law mentions cost and speed explicitly The rule is, every 2 years, twice as many transistors at 1/2 the cost. If transistor density doubles the speed doubles - more or less - more true in the 70s then today.
Great. Thanks to you, now I'm listening to Daft Punk on youtube.com.
God spoke to me
Without adjusting for inflation Intel's processors cost about as much as they did 20+ years ago.
http://www.krsaborio.net/intel/research/1991/0422.htm
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116492
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116899
http://www.nytimes.com/1992/01/09/business/company-news-intel-moves-to-cut-price-of-386-chip.html
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116775
Almost every other component (except maybe the GPU) has dropped tremendously in price over the past couple decades, but CPUs have stayed almost flat. Hopefully the newly competitive ARM processors will finally drive prices down (iSuppli estimates a measly $18 for Apple's new A7 CPU+GPU) but I'm not holding my breath.
How can I believe you when you tell me what I don't want to hear?
To run MS Office and watch cat videos on Youtube? Not very? THen I guess "Good Enough" computing will will moderate the situation....
Shoes for Industry. Shoes for the Dead.
I think a cartel exam is in order. If someone tries to explain a price hike in a field that is allegedly contested, especially when the reason given is threadbare at best, it's time to watch for price fixing.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
If Intel's investor day meeting is to be believed this is not true at least for their next 2 process nodes
http://files.shareholder.com/downloads/INTC/2827417808x0x709360/2D44DBF8-58B8-403F-B0E8-16E114CFF0E8/2013_IM_Smith.pdf .
Look at Slide 36.
--whacky
Yes, this is new.
I got my first computer about 30 years ago.
It is approximately a million times slower than the desktop I bought for less money this year. Admittedly, it uses about a quarter of the power.
Yes, each successive fab has been costly - with newer ones costing over a billion dollars.
But, until relatively recently, you could go from 2x to x feature size, and get lower power consumption, and higher performance.
This simple geometrical scaling has broken down - narrower features may mean you need more power, not less, and though they may be able to run at a higher frequency, cooling becomes a major problem.
As the summary stated, you now can't make - for any amount of money spent on a new fab, a denser process simply by virtue of geometry that gets you better power and clock speed.
I feel I'm really unlikely to see another factor of a million increase in computation, barring mature nanotech.
Oh look the 100th executive to predict the end of moore's law in the last month.
wrong. your reading comprehension is abysmal. The number of transistors doubling says NOTHING about density, only the number of transistors without any mention of area taken. Speed is not mentioned at all.
But an end to Moore's Law has been predicted before multiple times, and it hasn't happened yet. (Things have slowed down, yes, but they're far from stopping.) A few years back hard drives were predicted to reach a storage density limit, but this was solved by turning the magnetic cells vertical. So Moore's Law may finally be coming to an end, but don't be surprised if something new comes along and blows silicon transistors away.
Moore's Law is an expression of exponential growth. All we are seeing is the logical conclusion of applying exponential growth expectations to a real world finite resource (i.e. the fact that atoms have an essential finite size). See Wheat and Chessboard problem for reference.
He hasn't lost his confidence, it's just that his inner fan boy has died.
People are falling out of the "Cult of Tech" left and right. They swallowed the marketing message hook line and sinker and are starting to get bitter than their life hasn't become amazing and fullfulling after buying their 4th iPhone.
Personnally I see this as nothing more than a lul that happens every so many years. Specifically in gaming. Once 1900x1080(1200) became the pinical of screen resolution incentive to improve raw hardware died. We've been running laterally for some time now. Multiple monitors, energy conservation only can take you so far inovatively. The big killer is lack of fast ISP service for internet gaming. Now though we are starting to see the roll out of truly fast fiber sevice and there is a noticeable uptic in much higher resolution monitors being available which is going to start putting alot of pressure on the PC hardware to feed it.
As a non-American, that one doesn't bother me much :)
Encourage inventors rather than patent troll them into oblivion.
Just a thought, I know it would destroy much of the current economic model, but maybe - just maybe - those expensive techniques are merely the product of insufficient brains. Does the semiconductor world forget so soon that "cutting edge" in the 1970s was to melt silicon and scrape off the scum on top? Does it eve r occur to anyone that, just as we use reduction techniques to obtain silicon today because older methods were crap, there exists the potential that the expensive, low-quality techniques of today could be the rejects of tomorrow?
There are no inventors any more because silicon is a bloody expensive field to get kicked out of by patent trolls. Mind you, it's also a difficult area to get into, what with TARP being used to fund golden parachutes, bonuses and doubtless a few ladies of the night rather than business loans and venture capital. There's probably a few tens of thousands of mad scientists on Slashdot, and I'm probably one of the maddest. Give each of us 15 million and I guarantee the semiconductor market will never be the same.
(P.S. For the NSA regulars on Slashdot, and if you don't know who you are, you can look it up, feel free to post on your journals or as an article all the nifty chip ideas you've intercepted that have never been used. After all, you're either for us or for the terrorists.)
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
No more speed increases coupled with decreases in power consumption and cost. Fair enough, but who says increasing cost is the way to go? (That's rhetorical, we all know it's the business people saying that). Focus on less power consumption and at least keep costs the same. Use the chips we have to make systems with more processors. Take advantage of the cloud and Hadoop. Refocus on more efficient coding practices. We're so focused on chips getting faster, but parallel processing is a viable method of getting more processing done.
--
Luck is just skill you didn't know you had.
A good, modern C compiler is a lot better than you to find serendipitous optimization points in structured code
Provided that a developer can find and afford a "good, modern C compiler" targeting a given platform. What's the state of the art in compilers for 6502-based* microcontrollers again? Last I checked, code produced by ca65 was fairly bloated compared to equivalent hand-written assembly language. And I'm told that for years, GCC severely lagged behind $6000-per-seat Green Hills compilers.
* Why 6502? Maybe I'm making an NES game for the competition. Or maybe I need to code a hash table for the storage controller in a Terminator.
Nonsense. A lot of tech is finished and does not get better at all. For a common item, look at the pencil or the hammer. They are finished. They were available in the same quality decades ago. They do not get cheaper. Or take paper. Or take gate logic, foil capacitors or discrete transistors. There are countless other examples.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
No, the monkeys cannot. Even making 8088 requires a fab costing a few 100 Millions. These chips are cheap today because they take so little wafer-space, but the wafer and doing anything with it has not gotten cheaper. The other thing is that the fabs that old chips are made in have been paid off a long time ago. Bomb the right 20 places on the planet an the human race would need a few decades to be able to make chips again.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Let me guess, you're a huge Ray Kurzweil fan.
If they try to jack up prices they'll see what happens.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
I see many emerging technologies that promise further great progress in computing. Here are some of them. I wish some industry people here could post some updates about their way to the market. They may not literally prolong the Moore's Law in regards to the number of transistors, but they promise great performance gains, which is what really matters.
3D chips. As materials science and manufacturing precision advances, we will soon have multi-layered (starting at a few layers that Samsung already has, but up to 1000s) or even fully 3D chips with efficient heat dissipation. This would put the components closer together and streamline the close-range interconnects. Also, this increases "computation per rack unit volume", simplifying some space-related aspects of scaling.
Memristors. HP is ready to produce the first memristor chips but delays that for business reasons (how sad is that!) Others are also preparing products. Memristor technology enables a new approach to computing, combining memory and computation in one place. They are also quite fast (competitive with the current RAM) and energy-efficient, which means easier cooling and possible 3D layout.
Photonics. Optical buses are finding their ways into computers, and network hardware manufacturers are looking for ways to perform some basic switching directly with light. Some day these two trends may converge to produce an optical computer chip that would be free from the limitations of electric resistance/heat, EM interference, and could thus operate at a higher clock speed. Would be more energy efficient, too.
Spintronics. Probably further in the future, but potentially very high-density and low-power technology actively developed by IBM, Hynix and a bunch of others. This one would push our computation density and power efficiency limits to another level, as it allows performing some computation using magnetic fields, without electrons actually moving in electrical current (excuse me for my layman understanding).
Quantum computing. This could qualitatively speed up whole classes of tasks, potentially bringing AI and simulation applications to new levels of performance. The only commercial offer so far is Dwave, and it's not a classical QC, but so many labs are working on that, the results are bound to come soon.
Nonsense. A lot of tech is finished and does not get better at all. For a common item, look at the pencil or the hammer. They are finished. They were available in the same quality decades ago. They do not get cheaper. Or take paper. Or take gate logic, foil capacitors or discrete transistors. There are countless other examples.
All the technologies you listed have seen gradual improvements all the way up to this day. The pencil is now available in different hardnesses of the graphite and you can have an eraser on the tip. The manufacturing materials for a hammer have seen various improvements and the cost has dropped tremendously. Gate logic just got an improvement when the "3D" transistor was invented. Even discrete capacitors and transistors get various tweaks and efficiency improvements all the time.
We do more than make parallel processors with silicon chips. We have memory and flash storage too, for instance. those benefit.
Some tasks do actually scale well to parallelism if you want to talk processor. Still, one core needs access to memory usually, so designing buses that are fast and short enough to get all the cores proper access to the memory will be a challenge, unless you can shrink the process so you will have enough room on your silicon to put the pathways and logic to do so.
While most consumer use software can't deal with parallel stuff, there are plenty of reasons to want process shrink. Higher clock speeds is even one, if you insist on performance gains.
I was promised a flying car. Where is my flying car?
They said the same thing back in the 90s when the manufacturing tech of the time was approaching its theoretical 100nm limit, surpassing which would require manufacturing technology so revolutionary, nobody would ever, 3V4R, be able to afford it.
Have these people not learned anything from Bill's famous "640k ought to be enough for anybody" gaffe?
If _nothing_ changes, yes this will all come to pass.
BUT
Want to bet that in a lab somewhere, there is something that will let Moore's continue?
Think: how many times has this prediction been made and then proven wrong? Wonder if these statements are just ploys to jack up prices?
Sit back, relax, and be prepared to be amazed.
"Consensus" in science is _always_ a political construct.
I think memristors are a really interesting development, mainly because as I understand one of the potential applications is for storage densities greater than hard disks with DRAM-like access speeds.
It's not hard to postulate applications where you combine data storage and DRAM together, resulting in big performance increases by eliminating much of the latencies involved with disk access.
It probably wouldn't have as much impact on pure CPU bound tasks but so many workloads now are I/O bound and performance limited by disk systems that having a unified DRAM + storage space could mean performance increases beyond what the additional of CPU power alone could mean.
Storage / RAM is not the only application for memristors. As they can serve as single-element excitation counters, they enable extremely power-efficient neuromorphic chips, as described in this paper from Intel (PDF warning): Proposal For Neuromorphic Hardware Using Spin Devices
We present a design-scheme for ultra-low power neuromorphic hardware using emerging spin-devices. We propose device models for 'neuron', based on lateral spin valves and domain wall magnets that can operate at ultra-low terminal voltage of ~20 mV, resulting in small computation energy. Magnetic tunnel junctions are employed for interfacing the spin-neurons with charge-based devices like CMOS, for large-scale networks. Device-circuit co-simulation-framework is used for simulating such hybrid designs, in order to evaluate system-level performance. We present the design of different classes of neuromorphic architectures using the proposed scheme that can be suitable for different applications like, analog-data-sensing, data-conversion, cognitive-computing, associative memory, programmable-logic and analog and digital signal processing. We show that the spin-based neuromorphic designs can achieve 15X-300X lower computation energy for these applications; as compared to state of art CMOS designs.
It’s not about “green idiots.” It’s about the fact that chips will melt (burn? fry?) if you don’t keep them cool, and you can only dissipate so much heat from air cooling. Water cooling is used in HPC systems, but that too only goes so far. What’s next? Everyone needs a supply of liquid nitrogen to run their desktop PCs?
The “power wall” is a real, practical problem, which we reached somewhere around 2001, where power dissipation hit ~150 Watts in high-end systems. And the challenges go beyond cooling. Did you know that half the pins (around 1000) on a modern CPU are used just for power and ground? Do the math on trying to get 150 Watts at 1 Volt through a single pair if wires.
Oh, and what about mobile computers? Current battery technology can only old so much charge. Do you want your cell phone to get only an hour of useful life before recharging?
Bomb the right 20 places on the planet an the human race would need a few decades to be able to make chips again.
No, you are mistaken.
In the 8088 days the term "fabless semiconductor company" was still a new concept (1978 versus 1979 for the 8088) since every man and his dog had a fab. Any university wth a condensed matter physics department has a clean room and fab facilities. They're not set up for mass manufactuing of chips right now because there's no point.
There are thousands of places on the planet that could make chips on the level of an 8088 pretty much today.
SJW n. One who posts facts.
I think I'll stick with the leading predictions made by people such as Kurzweil over the chief tech officer at broadcom
The CTO (who is responsible to shareholders for turning a profit) of a semiconductor producer is laying down the foundations for why they won't be making processors cheaper?
There's a typo in the summary. It's supposed to say Intel isn't making chips cheaper anymore.
They have. Functional programming. By explicitly avoiding side effects huge chunks of code can execute independently and in different orders. Moreover by organizing the code using functional looping constructors the parallel compilers can tell how to break things.
Functional makes parallelism much easier.
I'm sure there's a lot of great applications, but unifying persistent storage and memory seems like one with a lot of disruptive and performance enhancing possibilities relative to the limitations of RAM vs. disk.
I think it has been a while since I have seen any new software that actually performed a task on new hardware that 10 year old hardware couldn't do just as well. It seems that only bloat and eye-candy take advantage of the latest hardware capabilities. (IMHO)
What I look forward to is being able to buy one set of good high-quality (as in built to last) computing devices and have them last me a decade or more without becoming terribly obsolete. I'm tired of buying new crap every couple of years. I'm also tired of getting cheap stuff that breaks just because I know that I won't want to use it long enough to get the value out of something better built anyway.
Which is actually of surprisingly little concern in most companies. As odd as it may seem, vendor lock-in doesn't faze managers in the slightest. Maybe because it saves them from having to make decisions about stuff they don't know jack about. "Oh, we cannot switch to $other_system, because $mission_critical_app won't run" seems to be an asset for them, not a limitation.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
Matters less now, though. It was valuable when we had a bunch of CPUs in the market - MIPS, SPARC, PowerPC, Hitachi SuperH, ARM, x86, et al. But with most of these having fallen by the wayside, portability hardly matters. Usually, one can just code for ARM or x86 and be done w/ it - it's not likely that the code will be needed in, say, a SPARC
Whether it's electronics manufacture, or oil and gas extraction. There's only so far you can push any technology.
Oh, and by the way, there's no such thing as magic. No Santa. No Jesus. No Tooth Fairy. No infinite power supplies. No infinite computer resources. No infinite supply of money that everyone takes seriously.
Time to grow up kids.
Please do not read this sig. Thank you.
I'm sure there's a lot of great applications, but unifying persistent storage and memory seems like one with a lot of disruptive and performance enhancing possibilities relative to the limitations of RAM vs. disk.
Yes, that one is a great thing, too.
This is how it might go:
- First, they release it as memory modules / ultra-fast SSDs
- Then, someone tweaks Linux and Android to use this memory for the main persistent storage, as well
- Then, special APIs appear to treat more in-app memory operations as persistent (similar to PalmOS)
- In the meantime, the neuromorphic chips bring new low-power AI features to robots and mobile devices
- Then, neuromorphic co-processors are added to "normal" computers
- Then, memristor memory and neuromorphic logic units find their way into the CPUs, building powerful hybrid systems
And if you combine this with the potential performance gains from the other technology I mentioned, the "Moore's Law Ending" articles start looking really lame.
When Moore's law ends, there will be mostly multi-processor computers. The code will all be concurrent. You won't really want to go down ASM and code multi-threaded programs. You will use languages like Scala that have good abstractions like Actors.
Democracy Now! - your daily, uncensored, corporate-free
Not really. This same kind of article pops up whenever a change in technologies starts to become appropriate. I.e., what we have installed now can't make the new stuff, and it's too expensive to build a new factory.
It's not that there isn't some validity to it's points, either. It's just that it's a short term perspective. The curve is bumpy, has been bumpy, and will continue to be bumpy. Sometimes it changes faster than Moore's law predicts, sometimes slower. The average is about right (though if I recall correctly, it's been adjusted in the past to speed it up).
It's also true that EVENTUALLY Moore's law (and it's associates) must end. When is an unanswerable question, though one can be fairly sure it will be before atomic level gates. (The noise level would be too high. You need to bring it down by dealing with multiple atoms. N.B.: This noise level has been a problem since at least the days of vacuum tubes. I'm not sure it was a major problem with gear driven calculators.)
I think we've pushed this "anyone can grow up to be president" thing too far.
...all those stories about the death of the PC were not just bullshit, but harmful bullshit.
Step 1: Create the illusion of digital scarcity
Step 2: Market to rich non-geeks who expect a cell phone's obsolescence timeline
Step 3: Pretend "ooOooooh, well now THESE chips are `SPENSIVE."
Software makers will be milking more and more speed out of quad-core CPUs for the next 15+ years anyway. The industry did this to themselves by squabbling over IP and taking a slash-and-burn approach to manufacturing quality. We (well, you, not me) helped by mindlessly repeating "zomg the PC is dead" for the past several years.
Every trollism an AC posts is prefixed, in my mind, with "A. Coward whined, in a weak and cowardly voice:"
As the silicon era closes, the era of graphene, other metals, and optical will ensue. At least that's been the buzz in the materials and computing sciences over the last several years.
Depending on what exactly you define as an OS, you can either look at Puppy or Damn Small Linux, both of which are (more or less) full desktop OSes, and which have far lower system requirements than any recent versions of Windows.
Stretching the definition of terms a little, there are a half-dozen linux distributions meant to fit on floppy disks. It's extremely common to use Linux in low-spec or embedded devices like routers. The lowest system requirements are on the order of one megabyte of RAM, and the same for persistent storage, and whatever CPU fits within your power envelope at that point. Linux is fairly competitive as an embedded OS, and used on billions of low-spec devices. Windows and OSX are not competitive in the embedded space.
Measuring application performance over a broad range of hardware and OSes is impossible, even if you limit yourself to a single application, due to binary incompatibilities. So with the important caveat that a evaluation of application performance between OSes is likely to be meaningless, we may feel safe in saying that as far as the OS functions are concerned, and all other things being held equal, Linux will use less system resources than Windows/OSX.
Please remedy your incredible ignorance, and refrain from invective, and if you can't do either, please stop polluting this forum with your drivel.
On the more general topic at hand, I concur with rubycodez on the subject of web applications, specifically to note that the comparison (while ill-defined) between web applications and desktop applications of similar functionality, should consider that HTTP is inherently stateless. Your C++ application is going to be best suited to a single user environment, and probably the curves of 'system resources required per number of users n' would intersect at a very low n. Lastly we observe that the exact amount of time required to execute a set of instructions rarely matters, and the choice of programming language is unlikely to be a first-order effect.
Those who advocate genocide deserve every protection afforded by law, and none afforded by common human decency.