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: NicoV who wrote (219441)	12/7/2006 7:07:51 PM
From: pgerassi	Read Replies (2) \| Respond to of 275872

Dear NicoV: Wire bonded CPUs have long been superceded by flip chip CPUs. The contact bumps are made on the last layer of the die and the die is flipped over and then bonded to the packaging substrate. This allows for bumps to be over the whole area of the die. As for the number of contacts, there are some configuration contacts that are not brought out of the package via pins. So an Opteron socket F is likely to have more than 1207 bumps. However I think that even 80mm2 dies are more than large enough for socket AM2/+/3. The reason for the area around the core section of the dies not shrinking has more to do with the drive currents required to interface to very high speed digital signals like DDR and HT. Those requirements don't drop as the die size shrinks. In fact, there are good reasons to make these more powerful so as to connect to more memory and have faster link speeds. Also these have to operate at higher voltages than the rest of the core like the 2.5V of DDR1 (JEDEC states that a DDR2 controller must be able to attach DDR1 memory) so they have to also have longer channels and thicker gate oxides. The control portions however can shrink, but each time the functionality is increased and that chews up the savings. Also some area is used for decoupling capacitors and that area, if anything, likely grows. The long and short of it is that XBAR/HT/ODMC portion has to remain at about the same size. If the market would have allowed AMD to go to the single DDR2 channel single HT link route, the number of pads would have been greatly reduced and this portion could have been halved in the value end. 20mm2 is a lot of savings for low end both in the number of dies per wafer and the yield would go up as well. So we would go from 120mm2 DCs to 60mm2 SCs. This would increase capacity by 100-150%. They would have been fine in entry level notebooks and desktops. Pete