LONDON – Foundry chipmaker Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) – said to have started trial manufacturing of the ARM-based A6 processor for Apple – will put the IC through another tape-out for the "production design" in the first quarter of 2012, according to the Taiwan Economic News. As a result of the respin of the design, production volumes of the A6 will not be available from TSMC until the second quarter of 2012 at the soonest, the report said referencing unnamed industry sources.
The A6, rumored to be a quad-core design, was expected to debut inside a yet more powerful iPad 3 tablet computer from Apple, which some thought could arrive this year.
One potential reason of the respin is that TSMC plans to use 3-D stacking technologies along with its 28-nm manufacturing process in the production of the A6 for Apple. The use of a specialized silicon interposer and bump-on-trace interconnect may produce specific requirements in the main processor die.
Samsung has been the sole supplier of the previous iterations of the Apple processor, the A4 and the A5. However, the use of company-proprietary packaging technology would count against the possibility that Apple is seeking to qualify both Samsung and TSMC and operate them as mutual second-sources.
A major reason why TSMC has not handled Apple's processor manufacturing to date is because the company has effectively been sold out with existing customers such as Nvidia and Qualcomm, the report said.
TSMC now expects manufacturing capacity utilization to dip to 92 percent in the third quarter, from 99 percent, and has forecast that its combined revenue will drop by between 6 and 8 percent sequentially in the third quarter, normally a quarter when sales are growing.
Packaging house Advanced Semiconductor Engineering Inc. partnered with TSMC to develop the 3-D chip packaging technology and so should benefit from the A6 processor business in 2012, the report said.
Is TSMC the only foundry capable for the 28nm and 3-D packaging together? If yes, Apple doesn't have to worry anything about. They just need some more marketing work to extend the lifetime of iPad2. I still believe Apple can dominate the tablet market by next Q1. So, if TSMC can fix the issue in time while it leaves enough time for Apple to refine the product, the new iPad may be more sexy! BTW, it is not too long to wait from iPad2 ato iPad3 (at least I don't think so).
Not sure if package need major change, yes some internal changes are required.
Regarding risk, probably its worth taking that risk since stcked dioe give benifit of speed and power which will give them market edge. Even if they fail for first time atleast they will be ahead of market by one step to implement this.
@dylan; I agree with you. Samsung has good revenue in fab coming from Apple. But key thing is if Apple withdrew then can Samsung keep its fab busy? I don't think Samsung fab want to give up just for Galaxy at the same time Apple is preparing for worst case too.
3-d stack memory module integrated on an interposer w/ a CPU on the other side would definitely help both power consumption and bus throughput( shorter lengths ) in handheld systems.
Apple often takes leadership at the system integration level but is not known for sticking its neck out to the bleeding edge of component ( e,g. die ) technology. Indulging in 3-d stack memory ( wide I/O bus ) would require getting involved in Semiconductor Physics ? Does Apple have the manpower to handle these problems yet ?
What would be a Killer App for Apple to justify it ? TV in a Smart Phone ?
3-d stacking has issues. First heat dissipation ( so initial applications will not include the CPU ), then stress ( efect of flexing in through silicon vias affecting device performance, hence requiring wide keep - outs / die real estate ).
Perhaps right now Apple is just testing the waters and getting the Suppliers ready. Application in real systems will have to wait for 2 - 3 years.
Any stacked memory dies would have to be compatible with the new package, but we did not hear about adjustments by them.
For Xilinx, FPGA dies were side by side on same interposer, but all the dies were made at tsmc, as well as interposer.
I don't know why Apple and tsmc would take such risk so soon. I think the 28 nm yield is hard enough already, and could be the real issue.
This may be another good example of how Apple respect its designers. Design team of A6 may need more time to optimize product and specification and they may need one or more run. It is is not good practice to fix product launch date before design team agrees to design progress. One wrong example from Apple was iPhone4 Antenna. They should have given more time to RF/Microwave engineers for testing and refinement.
@p_g- But can Apple continue to rely on Samsung for producing its processors while at the same time engaging in a legal battle with the South Korean firm over the Galaxy Tab? Seems like something's gotta give.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.