Yesterday we announced a significant milestone at Fab 1 in Dresden, Germany—the shipment of 250,000 wafers based on 32nm High-k Metal Gate (HKMG) technology. A quarter million wafers is a lot of silicon. If all of these wafers were laid side-by-side the trail would run for 47 miles, or about half the distance from Dresden to Prague, the largest city in the Czech Republic.
But of course our customers don’t sell wafers—they sell computer chips. So what does this milestone mean to our customers?
Our largest 32nm customer is AMD. They make several products using this process technology, including Accelerated Processing Units (APUs) for notebooks and desktops, and high-performance processors for servers and workstations. AMD’s president and CEO Rory Read summed up the news from their perspective:
In just one quarter, we were able to see more than a doubling of yields on 32nm, allowing us to exit 2011 having exceeded our 32nm product shipment requirements. Based on this successful ramp of 32nm HKMG, we are committed to moving ahead on 28nm with GLOBALFOUNDRIES.
We have been very open about the challenges we faced at the beginning of the 32nm ramp in Fab 1. Not only were we moving to a new technology node and manufacturing the foundry industry’s first HKMG product, but the transition also represented our first integration of a CPU and GPU on the same die and the first time a GPU had ever been manufactured on a Silicon-on-Insulator (SOI) substrate. However, we made several organizational and operational changes in the second half of the year that led to a dramatic increase in production velocity and major breakthroughs in yield learning. Now we have not only met all of AMD’s 32nm product shipment requirements, but we have exceeded them.
In fact, our calculations show that the 32nm ramp has actually outpaced the 45nm ramp by a significant margin. Cumulative die shipments for the first five quarters of wafer production are more than double that achieved during the same period of the 45nm technology ramp, despite the integration of a number of new and complex elements at 32nm.
And the good news does not end there. Since our 28nm technology uses the same HKMG implementation as 32nm, AMD and many other 28nm customers will benefit greatly from the learning we have achieved during the high-volume ramp at 32nm. Our 28nm HKMG technology is qualified and ready for design-in today. We began running our first 28nm customer products in Q4 2011 and we have multiple more planned for 2012. 2011 was clearly a year of transformation for GF, but we have gained a tremendous amount of momentum and we are starting off 2012 on the right foot.