Silicon Investor (SI) -- The First Internet Community

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) -- Ignore unavailable to you. Want to Upgrade?

To: Ali Chen who wrote (21416)	12/3/2000 10:56:08 PM
From: Bill Jackson	Read Replies (3) \| Respond to of 275872

Ali, I think the chip might have more than 25% of the thermal resistance across it. Remember the heat sinks have to be very flat as is the die surface so the heat sinc coumpound is very thinly spread and they use better stuff that titanium dioxide(white stuff) such as silver powders. In addition the heat sink is very well cooled by higher air flows with enough velocity to trim the boundary layer down as this as possible. It seems reasonable to expect that 30% they tout....in an optimised system of course. recall the heat flow equation is a series flow analagous to series resistors, if they make all the other resistances as small as possible then the lions share of gradient lies across the substrate. Then there is the heat taken off the front by the bumps into the pins. Do they make any efforts to make the socket a heat sink? this is doable. Voltages are low so a copper or aluminum plate in the socket would take away some heat. You would need holes in it and insulation with a fine pitch etc, but doable....every little bit helps. Bill

To: Ali Chen who wrote (21416)	12/4/2000 12:28:12 PM
From: pgerassi	Respond to of 275872

Dear Ali: I misremembered the thermal conductivity of bulk silicon. I thought it was about 8 times lower than it turned out to be. I used a later post's number (I forget whose). The original calculations showed a 40C rise in the silicon. However, these calculations assume that the power is generated evenly in the die. I seem to have missed a few required changes to the post. If the heat was generated in say 7.5% of the die in a region of 3mm by 3mm, the cross section would only average 4mm by 4mm in a die 1mm thick (think about truncated pyramid). This would increase the thermal drop by the ratio of the original cross section to the new one or 42.8C. This could be say in the floating point unit where long sequences of sequential trancedental functions (like atan, log, and y^x) are occuring (this is worst case (peak)). Here the thermal sensor underneath the CPU (as the case with most Athlon motherboards with temperature sensors) would only see the average temperature of the die and not the peak Tj. So my original conclusions would not be bad in this case. As power densities increase, the bulk resistance problem will start to dominate the total temperature drop of cooling solutions. It may have been a few percent (1 to 2%) long ago, but it is no longer insignificant. Pete