Advanced Micro Devices - Moderated (AMD) Message Board

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD)

AMD 237.57

-2.6%

Nov 11 3:59 PM EST

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next

To: Sarmad Y. Hermiz who wrote (217847)	11/29/2006 10:36:51 AM
From: pgerassi	Read Replies (2) of 275872

Dear Sarmad: A 65nm transistor does not have to be a different size than a 90nm transistor. What do you get with a smaller process? You don't necessarily get a smaller transistor, but better definition of where and how accurate the dimensions are. It's just like a laser printer. A 300dpi and a 600dpi laser printer makes the letters all the same size, but the 600dpi letters are sharper and more defined than those from the 300dpi. Yes yu can make smaller letters with the 600dpi one with the same relative sharpness. This analogy also is good in another way. Even though the image is sharper with a 600dpi laser printer, the toner particles are not that much smaller than those used in 300dpi. So the 600dpi dots are more ragged relatively to the size than those at 300dpi. At 1200dpi they are even more ragged and at 2400dpi the dot is almost the same size as the ones at 1200dpi, but the center can move a little over. It gets better only when the toner particles are made smaller. But those cost more money per unit volume (you may get more pages printed per unit volume, but not enough to make up for the premium). Similarly you get that with silicon processing. The material used has a particle (molecule) size measured in a few angstroms (0.1nm). But physics effects start becoming quantum in the 10-100nm range. These effects count for the smaller dots working quite different from the larger ones. So a 325nm transistor works quite different than a 450nm one, even if all of the dimensions are shrunk by the same multiplier. What a smaller process can do is make that 450nm transistor be more defined and consistent with respect to its neighbors. This reduces variability in transistor speed and leakage. This then allows a faster and/or lower leaking CPU. So going to 65nm optically from 90nm doesn't work as well as it used to. Intel likely used those assumptions in their 130nm to 90nm transition. But Prescott samples worked quite a bit different than planned. It leaked a whole lot more than they planned. And it had more variability as well. So instead of getting a well defined high speed warm CPU, they got a poorly defined medium speed hot CPU. As their process got better, they were able to make better defined better speed hot CPUs. A design shift and a smaller process allowed for a larger relatively somewhat better defined medium speed warm CPU. It is also the reason why a process that's 70% smaller doesn't make for CPU dies that are 50% smaller in area. You have to give up some size shrink to reduce variability and leakage. The lower power allows for somewhat higher speeds, but not the huge jumps that used to occur. Long term, the designers will make use of the quantum effects rather than work around them. Perhaps then we will see the big jumps again in performance. Pete

Report TOU Violation

Share This Post

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next