IEDM Coverage - Day 2, Online

Article: Coverage of IEDM 2003: AMD and Intel on Day 2
By: David Wang (dwang.delete@this.realworldtech.com), December 13, 2003 11:15 am
Room: Moderated Discussions
G Webb (glwebb@acm.org) on 12/13/03 wrote:
---------------------------
>Thanks.
>
>Does it suggest that Intel is going to get excellent yields
>with Prescott but that clock speed is unlikely to ramp up
>quickly due to high leakage, unless some industry wide
>effort on cooling, board design and casing occurs.

Just like how AMD continuously tweaked the 130nm SOI process until it finally got to where it is now, Intel will no doubt continuously tweak the process until it can support 4+ GHz at roughly the same power envelope or if by agreement with system manufacturers, further change spec to faciliate the heat removal of a larger power envelope.

Basically, a process technology is a combination of hundreds of little knobs, some knobs larger than others, but their interrelated effects are only known in theory, and you won't know eactly if the theory bears out until you've turned them all in all of the combinations conceivable to you. "Process learning" goes on all the time, but you have to be careful, because you don't want to do something that kills your wafer starts for an entire lot.

I think the leakage problem suprised a few people, because of the magnitude of the gate leakage as the oxides were thinned down. Not that intelligent people hadn't known about such effects, but the devil is in the details, and until you build it, you really don't know how good or bad it really is in terms of the relative magnitude of things.

>I've seen it be suggested that future heatsinks may need
>to fastened to the case itself rather than the board,
>and that ducting may become mandatory.

One of the papers released at IEDM is a paper on embedding a closed loop microchannel system (carrying coolant) into the semiconductor itself. The paper was written by a company called Cooligy.

http://www.cooligy.com/

I know that companies such as IBM, Intel and DEC had done this sort of experiement before, but the papers on them are not always released.

I also think that heat is itself not the real problem, but the real problem now become the cost of the cooling solutions (noise, $$$). Are we (as consumers) willing to continue to pay more money to get more performance? The server people will always pay whatever is needed to get the performance, but I think the trend is suggesting that more consumers are choosing lower performance and lower power draw mobile solutions.

< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
IEDM Coverage - Day 2, OnlineDean Kent2003/12/13 01:39 AM
  IEDM Coverage - Day 2, OnlineG Webb2003/12/13 10:38 AM
    IEDM Coverage - Day 2, OnlineDavid Wang2003/12/13 11:15 AM
      Cooling solutionsDean Kent2003/12/13 12:36 PM
        Cooling solutionsG Webb2003/12/13 12:51 PM
      IEDM Coverage - Day 2, OnlineG Webb2003/12/13 12:39 PM
        InterestingDavid Kanter2003/12/13 12:51 PM
        IEDM Coverage - Day 2, OnlineDavid Wang2003/12/13 12:53 PM
      IEDM Coverage - Day 2, OnlineJosé Javier Zarate2003/12/14 04:41 PM
  Good stuff (NT)Singh, S.R.2003/12/13 01:06 PM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell purple?