Matthew Purdy, Advanced Micro Devices
The economics of bringing a new, 300-mm semiconductor factory online are enormous, with capital costs in the billions of dollars. Such large initial capital outlays in the highly competitive semiconductor industry require a rapid return on investment. Such a return can only be done by operating factories at high levels of efficiency and predictability.
When looking to increase factory efficiency, there are two major focus areas for improvement – at the tool level and at the wafer level. Tool-level efficiency focuses on achieving optimal utilization of all of the equipment installed in the factory. Wafer-level efficiency focuses on achieving optimal results for every wafer started in the factory. At AMD, The automated precision manufacturing (APM) technologies being developed today are focused on achieving the goals of optimal wafer and tool results. These capabilities are being developed and proven in our 200-mm facilities in order to make the transition to 300 mm as efficient and effective as possible.
The 2004 International Technology Roadmap for Semiconductors (ITRS) for Factory Integration shows an expected increase in equipment utilization rates from 90% to 92% in the short term, and to 94% by the end of the decade. While this 2–4% change looks to be a small improvement, it represents a 20–40% reduction in overall nonproduction time. In order to meet this much-needed improvement, new technologies and capabilities in equipment diagnostics and material automation are being developed and deployed to improve efficiency. However, while these systems are deployed to improve equipment efficiency, other systems – notably automated process control systems – are creating new potential sources of equipment inefficiency.
In order to achieve the best tool utilization possible, it is necessary to understand all sources of idle time on the tools. Existing standards for tracking tool time comprehend the historical sources of nonproduction time (qualifications, maintenance, engineering time, etc.). However, these existing standards do not adequately comprehend the new sources of nonproduction time that become significant in leading edge 200-mm and 300-mm facilities.
As a specific example, automated control applications frequently have business logic to minimize risk to production (e.g., initialization, guarantee of closed loop feedback, etc.). While this logic does reduce the risk of lost production, the rules can also represent a significant fraction of the overall nonproduction time for a tool. At AMD, we have developed and deployed new systems to track tool utilization that comprehend these new sources of idle time (Figure 1). The information provided by these systems enables us to optimize the automation systems in the factory to minimize unnecessary idle time on tools.