IN THE RUN UP to this week's Semicon West show in San Francisco, Global Foundries' veep of manufacturing systems and technology has told the INQ why Automated Precision Manufacturing (APM) isn't just a load of old spin and cobblers, saying that the system really does give the firm an all important edge.
Tom Sonderman told the INQ APM offered specific, quantifiable benefits, not just when it came to line yield, sort yield, and bin splits, but also "significant benefits to cycle time, fab inventory, labour productivity and tool productivity."
All well and good, but don't we just have GloFo's word for this? Sonderman responded by telling us his firm continuously participates in benchmarking through third parties to validate its "competitive edge".
Still uncertain, we pressed Sonderman on what specifically was so unique about the system and why TSMC seemed so eager to get its hands on it. After all, every fab uses some degree of feed forward control, tool level sensor sampling and dynamic wafer monitoring, so why would APM be considered so special?
"It is true that every fab deploys some type of process control," conceded Sonderman, adding "the difference is that GloFo isn't controlling a tool or a recipe - we are controlling the entire manufacturing process." He said GloFo can do this because its APM infrastructure "is a coordinated set of applications that can work cooperatively to solve problems for the entire factory".
For instance, when excursions happen, APM allows GloFo to make adjustments in order to bring products back into spec, Sonderman explained. "Similarly, our yield analysis and WIP management systems are designed to optimise across the entire manufacturing cycle - not just at individual unit operations," Sonderman added.
As for the competition, Sonderman reckons "other players in the industry have been trying to catch up for years now". The reason they haven't managed, according to Sonderman, is that GloFo has not been lying back on its laurels. "If we were complacent about this technology, then our competition might be able to catch up in a few years," he told us, but added that GloFo "continues to drive improvement and value in this area and we expect to keep our lead".
Next, we asked Sonderman if APM didn't work against what a fab firm should want from its manufacturing, as the goal of any manufacturing process worth its salt is stability - to develop as wide a process window on your tools, so you don't need to 'tweak' the process from wafer to wafer and lot to lot. Without stability, we reasoned, the ability to fight excursions in the fab must be extremely difficult, if every wafer can have a slightly different process. Troubleshooting must be a nightmare with so many variables to explore.
But Sonderman was quick to tell us, "Stability isn't the goal. Predictability is." The veep said a stable process was one that would likely have difficulty being adjusted to changing customer and technology demands, while a "well-characterised and predictable process" allowed GloFo to keep it "stable" when necessary and to adapt and respond to changes when needed. "Our processes are capable enough that we don't have to 'tweak' the process on every wafer. But when our customers need that level of control, we have the ability to deliver it," he said.
Another major advantage, according to Sonderman, is that APM gives GloFo the ability to rapidly identify the source of any excursions, due to the "quality of data from the factory and the tools used to analyse that data". Sonderman told the INQ his firm's data infrastructure was "exceptional in that we have very comprehensive data collection and thorough analysis techniques--not just for lots, but also wafer-to-wafer and, in many cases, within wafer". That sort of in-depth data analytics purportedly lets GloFo's technicians adjust processes at the wafer level, which speeds up response time.
At the end of the day, Sonderman told us, "Anything sounds like PR spin if it is not backed up by results. The world-class manufacturing that GloFo has demonstrated in Dresden makes it clear that APM delivers results." µ
How many times are you going to write spin about APM? How was APM working out on 65nm?
Well at least this time you really nailed AMD/GloFo down on the benefits! "They've done benchmarking".... I guess that's specificity... we don't just have to take their word on it because "they've done benchmarking". Just curious any info ACTUALLY AVAILABLE from their benchmarking?
I'm guessing that's proprietary and we should just take their word on it :)
A decades ago, AMD said "REAL MAN HAVE FABS." Now they have been spun off their fabs to become one of chicken in the semiconductor industries. The chicken that initially founded with fabs, currently did not own fabs included LSI, Conexant and several others.
I think if they still retain their fabs, Intel will likely negotiates about the royalty and forget the royalty payments from AMD since the original IA32 had been expired.
All this about APM is pretty cool. Just so happens I'm gonna build a fab of my own so I can be called a Real Man.
Last question though, can they customize it for me so it brews coffee too and parks my car for me in the mornings?
Hope it doesn't spill the coffee on the wafers.
Sighs... all these attempts at sarcasm... how simply purile...
It's simple really. AMD (now GloFo) has invested the resources over the years to develop and institutionalize a culture of continuous incremental improvement. As part of that, they have developed the sampling and analysis tools to support that effort. Their focus is not just on a wafer, but on the process itself over the entire facility. Elementary competitive manufacturing, really. Becoming world class isn't just a matter of PR, it's actually walking the walk and doing the hard work it takes to get there.
As far as telling you specifics, not going to happen. It's a competitive edge. That's business, as you well know. No matter how much you flirt and flutter your baby blues, sweetie.
Also, don't forget that it's more than just GloFo.... IBM is involved deeply. There are some really sharp people involved.