The Benefits Of FD-SOI, by ST-Ericsson.
ST-Ericsson’s FD-SOI NovaThor integrated processor/modem attracted a great deal of interest at CES where it was demo-ed earlier this year. Here ST-Ericsson Fellow Marco Cornero and Andreas Anyuru explain the benefits of FD-SOI.
Thanks to the insertion of the ultra-thin buried oxide substrate, FD-SOI brings a number of fundamental improvements in the transistor electrical characteristics, while using planar manufacturing technology.
The advantages are :
Faster: in the same technology node, the FD-SOI transistor channel is shorter than for the bulk transistor. In addition, in contrast to the bulk transistor, the FD-SOI channel is fully depleted and free of dopants. These two factors together result in significantly faster switching at the same voltage, resulting in up to 35% faster operating frequency for the same power consumption at high voltage, and up to 100% faster at low voltage.
Cooler: several factors contribute to lower power consumption: the fully depleted channel removing drain-induced parasitic effects and lower power operation, better confinement of carriers from source to drain, thicker gate dielectric reducing gate leakage current, and better control of body biasing techniques (voltage applied to transistor body to better control speed and power consumption). The result is very significant power reductions of 35% at high performance and up to 50% at low operating points.
Simpler: process steps are 90% the same as for bulk 28nm technology, with a reduction of 15% of the total number of steps, resulting in shorter cycle time. In addition FD-SOI doesn’t require stressors or similarly complex techniques used in other processes. The resulting process is less complex than bulk already, and far less complex than FinFET technology.
From a microprocessor design perspective the advantages of FD-SOI versus bulk are:
Higher frequencies are achievable for the same voltage/power, or equivalently lower power is consumed at the same frequency;
The maximum achievable frequency is higher;
The processor can operate at lower voltages with still very respectable frequencies (1GHz at 0.65V). In low power modes the FD-SOI relative advantages become more spectacular, reaching up to 100% higher frequency versus equivalent bulk low power modes.
The ~35% increased efficiency at high frequencies is already more than enough for a FD-SOI dual-processors to outperform slower bulk quad-processors in the vast majority of the use cases, due to the currently limited software scalability.
On the low-power side, the impact of FD-SOI is even more fundamental, as it reduces or eliminates the need to adopt the way more complex and still immature heterogeneous multiprocessing solutions.
The extended operating modes described above are achieved through a combination of the FD-SOI advantages described earlier, with body biasing playing a major role. Body biasing is the application of specific voltages to the transistor body, in order to optimally adjust its characteristics to each specific operating mode.
Compared to bulk, FD-SOI enables the application of a wider range of biasing thanks to the elimination of the parasitic diodes between the body and the source and drain present in bulk technology.
In practice we obtain the same effect as if we had two distinct processor designs – one optimised for high performance and the other one for low power – except that we do so using a single circuit and by switching electrically between high-performance and low-power modes, by changing the body bias voltage.
Tags: control speed, design perspective, electrical characteristics, insertion, technology node