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."
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.
350 may bring costs down, but it isn't a process node advancement and won't help cram more transistors per unit area into a chip.
Instead it will just let them process more chips at once in most time-consuming processing steps such as deposition and oxide growth. The photolithographic systems, which are the most expensive equipment in the entire fab on a cost-per-wafer-processed-per-hour basis, gain somewhat due to less wafer exchanging, but the imaging is still done a few square cm at a time repeated in a step-and-scan manner a hundred times or more per wafer per step. Larger wafers however are posing one hell of a problem for maintaining film and etch uniformity, extremely important when you have transistor gate oxides on the order of a few atoms thick.
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.
The problem is not so much in the hardware than in the software nowadays...
Oh look the 100th executive to predict the end of moore's law in the last month.
This hits the nail on the head. For decades, software developers have been able to play fast and loose, while counting on the ever-faster hardware to make up for bloated, inefficient programs. Those days are ending. Programmers will need to be a lot more disciplined, and really engineer their programs, in order to get as much performance as possible out of the hardware. In a lot of ways, it will be similar to the early days of computing.
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)
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.
"For decades, low skilled software developers have been able to play fast and loose,"
FTFY
Embedded system programmers are the only real programmers anymore.
Do not look at laser with remaining good eye.
It's not cheap to get rid of that much processor power without improving anything.
Office XP system requirements: Single core processor at 0.133 GHz minimum. 0.4 GHz recommended. RAM 0.024 GB (OS) + 0.008 GB (Office). Storage 0.21 to 0.26 GB.
Help stamp out iliturcy.
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?