Dear Joe:
Easy! Core i7 was released only for 1S systems. With 2S slated for sometime late Q2 next year and 4S at end of Q4/09 or maybe even Q1/10. The original Opteron was working in 2S on release and when retail boards came out 2 months later on 1S, 4S and 8S. Heck 2 months before the release, you could get a 2S 1.4GHz Opteron server (quickly snapped up by developers). Desktops came later and in two flavors, single channel socket 740 and dual channel 939.
It still had to be validated by IT. But sales ramped up while that happened. Lots of criminal antitrust acts by Intel tried to force them out of the stores and OEMs, but they still sold well. The really big sales gains took about one year after release. A similar thing happened during the Socket F transition. But the earlier, plug in and go, dual core release only looked like it needed a 3 month upgrade validation before sales took off.
Now Core i7 might be ok, but the slowness of the multisocket server types implies that they are having problems. Going IMC and P2P links was always for servers with desktops later. Having that switch around is quite troubling to most people in IT. They will carefully want to test Core i7 servers before allowing them into their enterprise. Especially for any mission critical roles. That is why Intel had to have Dunnington and Harpertown Penryn based Xeons capable of going into the older server MBs. They would have a 3 month CPU upgrade validation cycle and keep the Xeon sales flowing versus Shanghai until Core i7 could finish customer IT validation.
Given the delayed release of multisocket Core i7 servers, it looks like that was a good idea to make them. From appearances though, it looks like they will lose share against Shanghai. And during the chipset/MB HT3.x upgrade, since the CPU/memory/etal is known good, likely would be also a 3 month IT validation. So even if 2S Core i7 is released before HT3.x Opteron MBs, they would still be in the IT validation hold, while HT3.x Shanghai servers get sold in great numbers.
The tables will turn only when the Opteron transitions to Socket G34, unless AMD convinces IT people in general by shipping Shanghais in the G34 socket initially and putting off Instanbul and its derivatives until after the 3 month upgrade validation. They could do that by pulling in the Socket G34 MBs one or two quarters earlier. Then the release schedule makes sense too.
First you send out HT2.0 Shanghais in Socket F. Then send out MB/chipsets for Socket F+ with HT3.x six months later. Then send out Socket G34 Shanghais (and even dual Shanghai die MCMs with 8 cores and 12MB of L3) with the new MB/memory a year from now. Then send out Instanbul CPUs with the now tested MBs/chipsets/memory 18 months after Shanghai. That allows all upgrade validation cycles instead of the much longer full validations. That puts HT3.x MBs out early Q2/09, G34 MBs out early Q4/09 and Istanbuls out Q1/10 with OEMs, suppliers and IT customers all happy.
Pete |
