On November 13, AMD moves to the 45-nm technology node with the launch of its new Opteron server chip, code-named Shanghai. AMD is the third manufacturer to reach this milestone, after Panasonic and then Intel.
Intel's Nehalem is due out soon, but many of the features of what has widely been touted as a "new" architecture are only new to Intel. AMD's previous-generation Barcelona was the first quad-core processor and also included 2 MB of shared L3 cache integrated on the same chip with a Northbridge memory controller. AMD's latest design, Shanghai, triples the L3 cache size to 6 MB, but all the basic building blocks were already incorporated in the 65-nm Barcelona chip.
There is no question that Intel is at the cutting edge of process technology, but that's not true for architecture. Intel probably decided it made more sense to introduce a highly integrated, quad-core design on 45 nm rather than 65 nm just to keep the chip size down. Whatever the reasons, the introduction of Intel's Nehalem architecture will come more than a year later than AMD's Barcelona.
Shanghai Die Micrograph Imaged through backside
Click on image to enlarge.
A quick glance at Shanghai confirms one of its major selling points: It is pin-for-pin compatible with Barcelona. This approach continues a long-standing AMD strategy to allow server manufacturers to maximize the life of their designs. AMD's 65-nm server chip was also designed to be a drop-in replacement, giving server vendors the freedom to simply upgrade the BIOS to take advantage of the latest processor architecture. This easy upgrade path allows server manufacturers a chance to keep their hardware designs going for many generations without new sockets and board redesign.
Ripping open the chip's packaging offered the first glimpse into whether or not Shanghai was simply a shrink of the Barcelona die layout. Moving to the 45-nm node certainly helped AMD to reduce the die footprint by more than 10 percent despite doubling the total SRAM size on the chip from 4 MB to 8 MB. But the designers also made better use of the silicon as well, as the blocks are more efficiently shoe-horned into the 253-mm2 die area. Each of the four processor cores has shrunk by almost 40 percent, to less than 22 mm2.
Of course, process technology is a bit of a different matter when comparing AMD with Intel. Semiconductor Insights began analyzing 65-nm parts from Intel in January 2006. AMD followed up in September 2006 with a dual core Athlon desktop processor. Intel introduced its 45-nm high-K metal gate (HKMG) process in November 2007—a full year before AMD. Intel was not the first to go to 45 nm, so the technology gap for AMD was actually even bigger. The honor for first 45-nm process goes to Japan and Panasonic's UniPhier video chip, which appeared a month before Intel's. Intel's process technology lead appears to be widening.
These dates consider only the date of first production of the new technology as the important milestone. However, AMD process engineers are quick to point out their strategy of rapidly converting all their manufacturing volume to the new node after its launch. They claim production was fully transitioned to 65 nm in mid-2007—prior to the launch of the Barcelona server chip in September 2007.