Electronics Reliability Issues at the 45 Nanometer Node and Below

Most tech-aware people have heard of Moore's Law. Moore was an engineer for Intel in 1965 when he famously observed that the number of transistors on a typical Integrated Circuit (IC) was doubling approximately every 18 months.

More recently, engineers have concerns that a continuation of Moore's law will be impossible. There is significant pressure on the IC industry to continue progress, hence justifying future consumer purchases. If the computer or other electronic module available today, is only equal in performance to the one already owned, why would you replace it?

Advanced fabs are now manufacturing at what is known in the industry as the "28 nanometer node" or below. This refers to the dimensions of features in the circuit. Intel has 4 factories in production with this technology node. They have been joined by Matsushita (Panasonic), AMD (GlobalFoundries), IBM, TSMC, UMC and Samsung. At this point, we are still on course for maintaining Moore's Law, but there are some troubling issues.

The most significant problem is reliability. As the recent problems with Intel's "Sandy Bridge" CPU chipsets revealed, even the best operations are vulnerable. The Cougar Point devices have been recalled for degradation over time. Over $700 million of Intel product was recalled (about 8 million units).

The thinnest layers in these devices are now measured in atomic layers (the number of atoms). At 45 nanometers, the thinnest layers are only 3-5 atoms in thickness, with very little margin for error.

One of the fundamental problems for the new node is heat. The general rule in electronics: for each 10 degrees (Centigrade) increase in operating temperature, a 50% reduction in circuit lifetime will result. The new devices, especially the microprocessors, produce more heat, and are much more difficult to cool.

There are other problems, many related to dielectrics. Voltage stress tends to increase as device feature size decreases. Over time, this can lead to dielectric breakdown. As the dielectric is reduced to less than 2 nanometers in thickness, leakage currents tend to increase significantly, which raises power consumption, and generates more heat.

Reliability studies are projecting a device lifetime for this generation of devices of approximately 3-5 years, as opposed to a reliability of 10-15 years with electronics produced at the 65nm technology node.

Flash memory has demonstrated that the smaller the device features, the less memory read/write cycles prior to failure. System designers, through careful statistical modeling and use of algorithms, idealize randomizing of cell use of available flash memory which allows the load to be spread out.

The reliability problems with this technology nodes 45nm and smaller are commonly discussed in the automotive, medical and military (Milspec) communities. These communities have always been more sensitive to reliability issues. One doesn't want a critical system failing.

Discussion of the reliability problems has been far less common in the general consumer community. This is predictable, since the Semiconductor Industry does not wish to discourage adoption of the technology, and the "consumer" is much less organized.

Many consumers apparently don't care. Even well documented reliability problems, for example with the Sony Playstation 3, did not significantly dampen sales of the product. Certain devices: game-players, cell-phones, MP3-player, etc. are considered disposable. In the future, with the newer tech nodes, it is necessary to modify our expectations of what constitutes a "normal" lifetime. It remains to be seen if the consumer will continue to be tolerant of increasingly reduced product lifetimes.