SI
SI
discoversearch

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 : Intel Corporation (INTC)
INTC 36.38-1.3%Dec 22 3:59 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Charles Gryba who wrote (150197)11/28/2001 6:02:38 PM
From: fingolfen  Read Replies (1) of 186894
 
fingolfen, that's not good. I am assuming by a yield hit you mean that all all speed grades are affected? If not, shouldn't we be seeing a ton of downgraded ( 1500-1600 ) XPs at much lower prices?

It really depends. To the first order, the speed of a CPU is determined by the speed of the transistors. The smaller the gates, the faster the transistor, and therefore the faster the CPU. The catch is, you have to target lots for speed. The other catch is that if you go below a certain gate length on a given process, the transistors tend to stop working (hence the reason ever new manufacturing process has changes beyond simple lithography in the front end). Any manufacturing process is going to have a certain amount of normal variation... picture a Gaussian distribution of gate lengths around the target. If your target for desired speed (in this case 1800+ models) lies close to the process breaking point, normal variation will indicate a number of the die simply won't function.

Normal variation indicates you'll get a few high (and just as many low) speed parts from a "normally" targeted lot. Problem is, with the GHz race on, demand at the higher speed grades typically outstrips your bin to those grades without specifically targeting to a smaller gate length. You therefore run a few lots knowing you'll get less yield to make up the difference, but take a hit in the "total salable die" category.

Intel apparently ran into this initially with the 1.0GHz P3 as they were exceptionally scarce (the 1.13GHz P3 recall was another issue entirely), but it's the fundamental physics behind why "top end" CPUs are more rare than the "sweet spot."

Right now, AMD is running 0.13 micron node gates (~70nm) on their 0.18 micron process. That CAN'T be healthy. Intel is running 0.18 micron at 90-100nm depending on who you believe, and they now have a full 0.13 micron process capable of running 65-75nm gates (and no one can go appreciably smaller than that until 193nm lithography is a go).

Once AMD moves to a full 0.13 micron process, they should get some yield relief, but I don't think they'll be able to inch up the speed of their CPUs too much because they've blown the wad on gate length. AMD clearly thinks the same thing, and have therefore been forced to take the more costly SOI road in the quest for speed. SOI has its own batch of problems, and while they may get the speed, SOI is notorious for yield hits by its very nature.
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext