Das Capitolin
07-14-2006, 04:59 PM
ONE OF THE more interesting parts of attending Semicon West is that you can tell what the industry will be doing several years out. You need tools to make semiconductors, without the tools, chemicals and other related items, chips are just pretty patterns on a CAD screen. One of the bellwethers for watching that is Applied Materials (AMat).
This show, the firm had several reassuring themes both for the technology buyer and the AMat stock buyer. They were strain, flash and 32nm, not necessarily in order of importance to either the tech or stock side of things, but AMat can help you will all three.
One of the more dramatic trends it was showing off was the rise of flash, and it's importance in the whole semiconductor ecosystem. Memory in general, mostly dominated by flash, is about to overtake logic in production volume, something that has never happened before in the history of semiconductors. This is largely driven by Buffy's burning need to have the latest corporate pablum on her iPod.
That means that volume goes up for companies like AMat, but more importantly, flash may start having more of an influence on tool and process design than good old CPUs. If you want a good example of why this may be less than optimal for your next iAMD64 gaming rig, look no further than the poster child for not-quite-as-tuned-as-they-wanted processes, Intel's 90nm. This trend may have major ramifications for CPU makers in a few years time.
AMat also pointed out that strain, the current hot topic in semiconductor manufacturing is good to go at 32nm, and can probably be extended well beyond that node. It also said that the smaller you go, the better strain looks, and it can have profound effects on performance.
That brings us to an overarching topic, less of a pronouncement than a general trend. In the past, semiconductor performance was dictated by physical scaling. Every time you made something smaller, it got faster, and life was good for white-suited fab denizens. The last few years have pointed out that this simple trend has some finite limits, or at least trends that were considered rounding error noise suddenly became the dominant performance limiter.
One of the main things that AMat was trying to point out was that physical scaling won't cut it anymore, and if you want performance to scale, you are most likely going to be doing it through materials. Guess what AMat sells?
That doesn't mean it is wrong however, it just means that the business it is in is, more than ever, becoming the most important part of the chain. Again, luckily it has the answer for you, and was showing off technologies to get you down to 32nm in a planar fashion.
Company spokesfolk said that everyone in the industry is looking at dual exposure as a backup plan to immersion lithography at 45nm. Intel has already said this is the way they will be going, and it would not surprise me if the comment made by the AMat people was meant to have a lot read into it. For AMat at least, two exposure passes means more materials to sell, so for them at least, life is good.
Overall, the future looks bright at least to the 32nm node, and after that, well, that is where the research comes in. AMat is hard at work there, dropping hints at sub-32nm processes when it thinks no-one is paying all that much attention. If you are building semiconductors, the roadmap looks good for at least the next few years. µ
Source: http://uk.theinquirer.net/?article=32980
This show, the firm had several reassuring themes both for the technology buyer and the AMat stock buyer. They were strain, flash and 32nm, not necessarily in order of importance to either the tech or stock side of things, but AMat can help you will all three.
One of the more dramatic trends it was showing off was the rise of flash, and it's importance in the whole semiconductor ecosystem. Memory in general, mostly dominated by flash, is about to overtake logic in production volume, something that has never happened before in the history of semiconductors. This is largely driven by Buffy's burning need to have the latest corporate pablum on her iPod.
That means that volume goes up for companies like AMat, but more importantly, flash may start having more of an influence on tool and process design than good old CPUs. If you want a good example of why this may be less than optimal for your next iAMD64 gaming rig, look no further than the poster child for not-quite-as-tuned-as-they-wanted processes, Intel's 90nm. This trend may have major ramifications for CPU makers in a few years time.
AMat also pointed out that strain, the current hot topic in semiconductor manufacturing is good to go at 32nm, and can probably be extended well beyond that node. It also said that the smaller you go, the better strain looks, and it can have profound effects on performance.
That brings us to an overarching topic, less of a pronouncement than a general trend. In the past, semiconductor performance was dictated by physical scaling. Every time you made something smaller, it got faster, and life was good for white-suited fab denizens. The last few years have pointed out that this simple trend has some finite limits, or at least trends that were considered rounding error noise suddenly became the dominant performance limiter.
One of the main things that AMat was trying to point out was that physical scaling won't cut it anymore, and if you want performance to scale, you are most likely going to be doing it through materials. Guess what AMat sells?
That doesn't mean it is wrong however, it just means that the business it is in is, more than ever, becoming the most important part of the chain. Again, luckily it has the answer for you, and was showing off technologies to get you down to 32nm in a planar fashion.
Company spokesfolk said that everyone in the industry is looking at dual exposure as a backup plan to immersion lithography at 45nm. Intel has already said this is the way they will be going, and it would not surprise me if the comment made by the AMat people was meant to have a lot read into it. For AMat at least, two exposure passes means more materials to sell, so for them at least, life is good.
Overall, the future looks bright at least to the 32nm node, and after that, well, that is where the research comes in. AMat is hard at work there, dropping hints at sub-32nm processes when it thinks no-one is paying all that much attention. If you are building semiconductors, the roadmap looks good for at least the next few years. µ
Source: http://uk.theinquirer.net/?article=32980