Since the launch of the Core 2 in the summer of 2006, Intel’s line of processors hasn’t evolved in terms of architecture. In fact, there has just been the release of four core versions composed of two Conroe dies and then slight evolutions by adding FSB1333 models. While AMD's Phenom has not yet arrived, with the Core 2 Extreme QX9650 the Santa Clara giant offers its first processor based on Penryn architecture.Penryn : Core v1.1
The Penryn core and its 6 MB of L2 cacheThe Penryn core is an evolution of the Conroe. Still based on Core architecture, there have been several improvements starting with the engraving process which is now in 45nm instead of 65nm. Intel took this opportunity to increase L2 cache (from 4 to 6 MB) which means the total number of transistors explodes from 290 to 410 million. The size of the die however is reduced and is now 107mm² versus the previous 143mm². This allows Intel to increase the number of processors per silicon wafer and therefore lower production costs.
In terms of architectural improvements, the Penryn especially stands apart in two ways. The first involves the enhanced unit responsible for division. Like us, Intel processors use Euclid’s method: a quotient and a remainder is attributed to a divisor and dividend. Processors carry out division in steps, or in other words, for each cycle only a finite number of bits is processed. This operation is relatively slow (the number of cycles necessary depends on the size of the dividend), however it is exact.
While the Conroe processed two bits per cycle (a Radix-4), the Penryn processes four (Radix-16). This innovation is also of benefit to other more complex operations involving division, such as very particular optimizations for the calculation of square roots.
