Slashdot Mirror
Intel On Track For 32 nm Manufacturing
yaksha writes "Intel said on Wednesday that it has completed the development phase of its next manufacturing process that will shrink chip circuits to 32 nanometers.
The milestone means that Intel will be able to push faster, more efficient chips starting in the fourth quarter.
In a statement, Intel said it will provide more technical details at the International Electron Devices Meeting next week in San Francisco. Bottom line: Shrinking to a 32 nanometer is one more step in its 'tick tock' strategy, which aims to create a new architecture with new manufacturing process every 12 months. Intel is obviously betting that its rapid-fire advancements will produce performance gains so jaw dropping that customers can't resist."
I'm just finishing a rebuild of my system, going from an Athlon64X2 to a Core i7. 3DMark06 is downloading now; can't wait to see how well it does on that and Flight Simulator X.
...Now if they could only make some progress on coordinating RAID implementations across motherboards, so a MB swap doesn't have to mean that the path-of-least-resistance is a complete reinstall...
Paleotechnologist and connoisseur of pretty shiny things.
At WinHEC 2008 the Intel speakers continued to hint at the fact that they had operating, packaged cores at this size. On track for manufacturing? More like they've been making it for 9-12 months already. At any rate, it's cool, though not surprising.
Newton-metres? You mean Joules?
What could possibly make you confuse N which is a symbol for Newton with n which is a prefix for nano.
You're definitely not geeky enough.
At some point, it will stop getting smaller.
It's great that Intel are working on die shrinks for their processors, but I wish they would do the same for their support chipsets. It's annoying that on most laptops the northbridge for Atom processors uses more power than the processor does.
"I bless every day that I continue to live, for every day is pure profit."
Am I the only one feeling we might have reached the point of diminishing returns, at least for desktops, in the last 2-3 years. All the shrinkage past 90 nanometers just feels underwhelming. Stuff beyond Pentium 3 has not been revolutionary, performance wise, for a desktop.
For this reason the SI standard dictates that metric units such as "km" or "nm" are never capitalized, even on a sign that is written ALL-CAPS.
If Intel is able to shrink its die size every 12 months AMD is in trouble. A more efficient design is usually beaten by a less efficient design fabricated in less space. That is if you think AMD's design is still more efficient.
Intel: I'm a chip company. I make chips, that's all I'm programmed to do.
AC: Were you any good?
Intel: Are you kidding? I was a star. I could make a chip to any size. 30 nm, 32 nm, you name it. 31... But I couldn't go on living once I found out what the chips were for.
AC: What for?
Intel: MacBooks.
This is one case where shrinkage is damn good.
Don't take that out of context.
In my day, getting to one micron feature sizes was a big deal. And we were grateful!
You kids get off my lawn!
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Faster computers are going to be generally irrelevant to about 85% of the population. They only really use computers for surfing the internet, checking e-mail, MS Office, iTunes, organizing photos, and playing The Sims occasionally. Most people play video games on consoles (PS3, WII, Xbox 360). There are few things that 90% of the population regularly do that require a faster computer. These advancements are going to affect businesses and scientists who need super computers to perform large amounts of computations, or servers that need to respond to heavy demands. The only thing, I think, that needs to be improved is the hard drive. Right now they're just way too slow.
It's all about splitting hair nowadays
Intel has always enjoyed a much better manufacturing technology than AMD. But, Intel made some stupid architectural decisions with the P4 architecture.
Once Intel came out with the Core series, then the combination of a decent architecture and terrific fab capabilities really started eating away at AMD. This will only continue the rally.
The sad thing is that this will actually be a step back in pricing... it's getting back to where AMD simply cannot touch the higher-end Intel territory, and so Intel is back to enjoying terrific profit margins on those chips.
Oh, you're not stuck, you're just unable to let go of the onion rings.
Could someone explain to me hoe smaller chips are better? Doesn't that decrease the surface area and make it even harder to dissipates heat from the chip?
What's a 'micron', USian?
What's a 'micron'
It's a term used in semiconductor manufacturing to refer to one millionth of a meter.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
I think AMDs strategy is overclocking and lots of it. Look at what it's introducing in it's latest and upcoming hardware. Features that make overclocking easier. Also I wouldn't count AMD out too soon. Amd is just one design correction away from having perfect hardware for HTPCs And their IGP is still better than Intels.
Shai Schticks:"You don't make peace with friends, you make peace with enemies"
"Intel is obviously betting that its rapid-fire advancements will produce performance gains so jaw dropping that customers can't resist.""
Two things. One it doesn't matter how awesome your hardware is. If the majority can't afford it then it doesn't matter? Second as Microsoft is learning prior success can be a barrier to future growth. How many are going to throw out their Core 2 Duos in order to have the most amazing hardware from Intel?
Shai Schticks:"You don't make peace with friends, you make peace with enemies"
I found out from my wife that our home server died and won't reboot. AMD Athlon 3200+ running Fedora.
It is almost certainly a hardware problem, and that server has been running 24/7 for years now... time to upgrade.
My hardware philosophy has been to buy big and milk it for a long time. You pay more up front for that power, but the fact that it has power means it doesn't get obsoleted immediately either.
So then, cut through the marketing crap. Assume a desktop PC purchase in the May-ish time frame, to run Linux. What is likely to be the way to go from a hardware perspective?
DG
Want to learn about race cars? Read my Book
Stupidly uppercasing everything in a headline will regularly backfire if scientific units are used.
What's meant here is nanometer, not Newtonmeter - which, by the way, is equal to Joule.
And now here I am, unable to think of a good pun about a 32 Joule chip...
What about not only putting the more cores on die bit also RAM and the chipset. As said the motherboard would be different from anything we've seen so far. Would this result in any meaninful gains compaired to the cost and limites on options?
I have an issue of Byte up in my attic where the process shrink to 1 micron is the cover story. I read it a couple of years ago, around the time of the 65nm process shrink. It really gives you a sense of the speed at which process technology is improving.
I am TheRaven on Soylent News
premature optimization is sometimes as you say, bad. however there is an idea of mature optimization where you know something needs to be written in such a way as to be fast.
say your task has to run in realtime, and it involves iterating over most of the machine's memory. if it doesn't run fast, you have a real problem.
always choose the correct read/write patterns, the correct architecture, and then make that code as clear as possible...
music - http://www.subatomicglue.com
Can anyone remember the last time an incremental advance in chip speed was anything close to "jaw dropping?" Having been in this industry a while I can't count the number of times people like Steve Jobs and Andy Grove claimed speed increases of more than double with almost no apparent effect on anything but benchmarks. The early days of 3D accelerators was about the only time I really went "wow!"
Syntax error: loose != lose, affect != effect, then!=than
Really, this is just a matter of having limited manufacturing capacity. Every time they create a new manufacturing process, they have to upgrade a factory to use it. This puts the factory out of service for however long it takes to roll out the new tech, and costs billions of dollars in the process. In other words, even Intel doesn't have the resources to upgrade all of their factories at once.
Instead, they take one or two factories running the oldest tech, and upgrade them. Once they are ready, they start manufacturing the high-end processors. The last-generation tech manufactures lower-end processors. The generation before that manufactures chipsets, graphics chips, etc. The generation before that manufactures DRAM / flash / whatever else is needed. This is just an example, I have no idea what the split is in reality.
So from Intel's perspective, they are always using the newest manufacturing tech for their most important products (high performance, high profit margin). This in turn gives them the capital needed to develop their next manufacturing process, and the cycle continues...
Just to clarify: the tick-tock strategy means that one year gets a new architecture, the next year gets a new manufacturing process, and the cycle repeats. This means that there is a new architecture and new manufacturing every 24 months, not 12, and in alternating years.
The whole separate northbridge thing is kind of a legacy idea. AMD ditched it some time ago, now Intel is ditching it. Well, that being the case, little point in pushing forward with advances on it, only to then deprecate it immediately after. It'll probably be till the next "tick" before it is totally gone, but it should happen soon.
The other problem people have to remember is that they have a limited amount of the highest tech fabs. It isn't as though you flip a switch and the fab suddenly is on a smaller process, it takes a massive retooling effort, months of work. So if you have a limited amount you can crank out of this new tech, you do it where it matters most: The processor. That is your real bread and butter in terms of money. You have the other, lesser fabs do your other stuff.
So as for the Northbridge, this should be solved soon. I can't imagine they will make any 32nm processors for desktop or laptop that aren't based on the Core i7 architecture. As for die shrinks in general, I think you can always expect other chips (northbridge/southbridge network, flash, etc) to be on a process that's at least one generation out of date of their current one.
I was looking at the range of low power CPUs and noticed that Intel's Atom seemed to do ok compared to the other low powered chips but then noticed that all the other chips were being built on a 65nm process while Intel had the Atom on the 45nm process. Looking at Intels standard "Core" processors showed that their newest CPUs were also on the 45nm process but not the majority of them.
This was a few months ago but it made me wonder why all the other low power CPU manufacturers were able to get the power and performance from 65nm while Intel had to jump to the newest and smallest process(45nm) for the Atom to get in the ballpark on power and performance. And they had to be eating into production space for the high profit high performance larger CPUs they'd just released on the 45nm process.
They seem to have a strategy of winning by using their ability to shrink the process instead of optimizing the designs. It's a much much more costly exercise and if anyone else with massive amounts of process experience steps in to manufacturer for those with better chip designers, they'll not be gaining much more than a few months. And this push to low power and high performance is also bringing ARM into play with their Cortex A8 and multi-core Cortex A9 processor. It's amazing what a Cortex A8 can do on 3W.
LoB
"Anyone who stands out in the middle of a road looks like roadkill to me." --Linus
Its a common term for what would, in normal SI parlance, be a "micrometer", because "micron" predates the adoption of the SI system, and "micrometer", in English at least, has a well established meaning referring to a particular kind of measuring device.
Emphasis mine:
No, Intel aims to create a new architecture *OR* a new manufacturing process every 12 months. NetBurst to Core 2 in 2006, 65 nm to 45 nm in 2007, Core 2 to Core i7 in 2008, 45 nm to 32 nm in 2009.
Alternating between architecture and process.
I first learned the importance of getting things right. (A payroll program that gets the wrong answers gets the doors torn off the front of cookie factory. [That was my predecessor's mistake. :-])
Then I learned the importance of getting them to run fast. (I had a twelve hour window for calculation that suddenly got chopped in six as the company spread over a wider geographic area. The company bought their competitor. Now I had more impatient people to deal with [See previous 'font door' problem.])
Squeaked by and the machine eventually got upgraded with a mainframe.
I eventually learned to write Smalltalk (interpreted once, cached compiled code) that could outperform any C or C++ code by exploiting the structure of the problem. (I had people generating huge reports multiple times because they couldn't get it through their thick skulls that the program was not malfunctioning, it has actually finished.)
First, do it right.
Then do it fast.
Lastly do it small. (RAM is simultaneously the most expensive part of a system and if you know how to program, since time is money, the cheapest.)
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
O-O code can be optimized by knowing how (and therefore where) to cut up your code.
The code itself doesn't need to be any different, but how and where you cut it up can make an enormous difference in performance.
If you can take advantage of RAM to cache intermediate results of seek (find/get) operation, you can get incredible speed out of otherwise 'dead code'.
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
or make a jaw-dropping investment with such potential for failure it will bring us back to the electronic stone age.
what does Intel have that Ford didn't before?
--- ask me about nihilism, I will have nothing to tell you.
I want that in a Terry Pratchet book!
I know tobacco is bad for you, so I smoke weed with crack.