|
Losing the leaks!
|
09/10/2006
|
| |
When Intel founder Gordon Moore proposed his famous ‘law’ in the 1960s, he envisaged semiconductors halving in size and doubling in power every 18months. Since then, this empirical view has been used by the industry as the engine to power a seemingly endless migration towards smaller process technologies.
For many years, following Moore’s Law didn’t create too many problems for process engineers. It was only when process technologies moved into the deep submicron range that physical problems started to make life complicated. In particular, the transition to the 0.13µm node proved difficult.
One reason for this is power; as process technologies get smaller, then dynamic and static power consumption grows in proportion and the phenomenon of leakage becomes dominant. Solving the leakage problem has been one of the industry’s biggest challenges.
It also stands to reason that, the smaller the process technology, the more complex the design process. Assumptions that were made at earlier process nodes no longer hold and a new range of tools has had to be developed. And, unlike earlier processes, it’s the semiconductor manufacturers who have taken the lead, rather than the eda companies.
TSMC, the leading semiconductor foundry, unveiled its 65nm manufacturing process last year. The company took three conflicting factors into account during its development. Firstly, the higher performance that designers are looking for from 65nm makes timing closure more difficult. Power is also pulling in different directions; whilst designers want lower power consumption, leakage is a barrier. Finally, yield becomes critical.
Gareth Jones is director of business operations for TSMC Europe. He said the company has responded to the challenges of 65nm technology with an updated reference flow. “Reference flows are additional tools to help designers get to ‘right first time’ silicon. And, as each node appears, additional areas need to be addressed. Power management is becoming critical and the need to have better timing closure has brought about statistical timing tools, because previous models were not comprehensive enough.”
|
| |
Author
Graham Pitcher
|
| |
| |
|
This material is protected by Findlay Publications copyright 2008.
See Terms and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the sales team.
|
| |
|
| Email this article |
| |
|
|
|
|