SemiAccurate doesn’t reprint every rumor out there, and we do confirm our stories to the best of our ability. When the Ivy Bridge delay, said to be about a quarter, surfaced a few weeks ago, we asked our contacts at Intel and other companies what was going on. The short story was that everyone said there was no delay, and 22nm was fine. With that information in hand, we simply didn’t write it up because it was not a real delay, breathless hype aside. The end result was not a delay until June, Q3, or indefinitely, it was a mere three weeks worst case.
A few weeks later, we found out what the rumors were about, Ivy’s launch was delayed because Intel changed production plans at the last minute to put in a different mix of SKUs. A few large customers shifted orders around at the last minute, and asked for harder to make chips. For logistical reasons, Intel decided to delay the launch.
Rumor had it that the mix was skewed more toward lower voltage parts, basically LV and ULV versions, the hardest to make and one of the most sought after. This would indeed explain the not-really-a-delay level delay. That however only turned out to be part of the story, there was one other missing piece that could explain the delay, a shift in GT levels as well.
The non-rumor mill is telling SemiAccurate that in the very recent past, a very large customer shifted their order to get a higher GT level chip. If you recall, large parts of the Ivy die are taken up by the GPU, and there are two flavors, the 6-shader GT1 and the 16-shader GT2. Everything else being equal, a GT2 takes up notably more die area than a GT1.
If a large enough customer decides to change their orders from LV or ULV GT1s to GT2s, Intel is not going to say no to the extra revenue. And that is exactly what happened. At least one big order was moved from ULV GT1s to GT2s, and Intel could not produce enough by the launch day, so that day was moved out for all. Any guesses as to what OEM uses lots of ULV chips, has an interest in graphics that don’t suck as much, and has a lot of sway at Intel?S|A
Latest posts by Charlie Demerjian (see all)
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016
- Everspin and Globalfoundries team up for embedded ST-MRAM - Sep 15, 2016
- Intel’s Xpoint is pretty much broken - Sep 12, 2016
- ARM adds 2048-bit vectors to v8A with SVE - Sep 7, 2016
- AMD releases Bristol Ridge 7th Generation APU - Sep 5, 2016