On the heels of Globalfoundries' 28nm hint yesterday, IBM is announcing that its entire bulk processing alliance is ready to manufacture using 28nm.
Recently Globalfoundries hinted to the upcoming 28nm processing in an interview with X-bit Labs, briefly mentioning the process while talking about competing for AMD's graphics business. Although partially owned by AMD, the company is actually one of many manufactures in IBM's Bulk Process Alliance; Chartered Semiconductor, Samsung Electronics, ST Microelectronics, and Infineon Technologies are also part of the Alliance. However, prior to the interview, manufacturers--or rather early access clients--received hands-on access to a 28nm low-power technology evaluation kit back in December 2008. Then, in March, the kit was made available to the general marketplace. With that said, it's really been no secret that 28nm processing would arrive in the near future.
However, early this morning, IBM officially announced that the Alliance companies are jointly developing the 28nm, high-k metal gate (HKMG), low-power bulk complementary metal oxide semiconductor (CMOS) process technology. While that description may seem rather long-winded, the wait for the actual technology will be even longer, as IBM said that "early risk" production for 28nm is slated for Q2 2010. IBM also said that when the time comes, current clients using the 32nm technology will be able to migrate to 28nm without having to completely overhaul the design.
"Clients can begin their designs today in leadership 32nm HKMG technology and then transition to 28nm technology for density and power advantages, without the need for a major redesign," IBM said. "By assuring a path from 32nm to 28nm technology, this migration methodology offers clients lower risk, reduced cost and faster time-to-market."
But why is 28nm such a big deal? As stated by IBM, the low-power 28nm technology platform can provide a 40 percent performance improvement when compared to the larger 45nm chip, and more than 20 percent reduction in power. Additionally, the HKMG technology--allowing for one of the industry's smallest SRAM cells at 0.120 square microns--offers favorable (power) leakage characteristics, stability, and low minimum voltage, providing an optimized battery life for mobile products.
"Through this collaboration, IBM and its alliance partners are helping to accelerate development of next-generation technology to achieve high-performance, energy-efficient chips at the 28nm process level, maintaining our focus on technology leadership for our clients and partners," said Gary Patton, vice president for IBM's Semiconductor Research and Development Center on behalf of the technology alliance.
Other members of the alliance added their comments this morning as well, all boasting the new 28nm process. Jorgen Lantto, chief technology officer of ST-Ericsson, said that the process will make a huge impact in the mobile industry. "Leaders in the mobile industry can utilize 28nm low-power technology to meet the increasingly aggressive demands for performance and improved battery life," he said.
Notice they are comparing 28nm to 45nm... but no mention of 32nm which is also "supposed" to be ~ 40% better. The 28nm is just about shrinking optically - you'll notice no performance comparison BETWEEN 32nm and 28nm, because some folks might say why bother when you consider the cost of masks and qualification time. IBM pulled this same crap on 65nm when they were comparing the process to unstrained devices claiming huge increases and yet when it was compared to 90nm... no so much....
If folks do the products at 32nm, it will be a hill to climb to convince them to move to 28nm.... Perhaps IBM is trying to convince folks to jump to 28nm from 45nm, but then why bother with 32nm? Again when IBM talks about 32nm to 28nm they talk about it being "easy", but no mention of the performance jump (which will be minimal) in that case. IBM doesn't understand that if you are not making high volumes of parts, it really doesn't make economic sense to do a migration of a half node.
Well put my friend...very well put.
With the multi core race this could be used as a half increase in cores from 4 to 6. Generally tho it takes full node just for increase performance on the CPU side so maybe just wishful thinking.
Excuse this noob of a question but what is nm (nanometer?)? and what's a node?