There are a couple of tried-and-true strategies of the PC world that never seem to get old, and continue to supply us with endless new processor revisions and ever-higher performance. One of these is the venerable die shrink, where process technology moves to the next level, allowing ever-smaller transistors, and potentially higher clock speeds. This does not represent a new architecture, but an enhancement on the existing one, and in the process, extends its lifespan. We've seen this employed in virtually every processor line, and now it's time for the Core 2 to get a facelift, moving from 65nm to a new 45nm world.
The Intel Penryn is the name given to Intel's 45nm processor family, which is made up of the Yorkfield quad core and Wolfdale dual core models, which correspond to the Conroe (dual) and Kentsfield (quad) from the 65nm era. Both Penryn models are based on 45-nanometer (nm) High-k metal gate silicon technology, which uses a combination of high-k gate dielectrics and conductors, rather than silicon, to build the transistor gates in a Yorkfield quad core. A prime advantage of this technology is faster transistor switching speeds, but at reduced power, which in turn allows higher processor speeds at a lower thermal and power envelope.
The physical architecture of the 45nm Penryn family goes deeper than just a die shrink. Instead, Intel has taken advantage of the 45nm SRAM process to upgrade L2 Smart Cache to a full 6MB per dual core (2x6MB for quad), which represents a 50% increase compared to the initial Conroe and Kentsfield processors - the extra L2 increases the transistor count from 582 million to 840 million. The L2 cache also has 24-way associativity, compared to only 16- way associativity on the 65nm Core 2 processors. This increases theoretical performance, as cache hits will be higher with a 6MB L2 compared to only 4MB, not even counting the 24- to 16- way associativity upgrade.
Another bonus to any die shrink is that the basic core architecture has been tweaked and improved since inception, and the Penryn will include the latest microarchitecture enhancements. Other enhancements include SSE4, Super Shuffle Engine, Enhanced Intel Virtualization Technology, and Fast Radix-16 Divider. This last feature effectively doubles the divide speed of the Penryn compared to previous models, which can aid in the performance of any math intensive functions. Power efficiency is also addressed through features like Deep Power Down Technology, which improves on the 65nm Core 2 by allowing, according to Intel, the lowest power state a processor can reach.
The Core 2 Extreme QX9650 is a Yorkfield quad core processor, featuring dual 6MB of independent L2 cache, for a total of 12MB of Advanced Smart Cache. This doesn't change the current Intel quad core design used with Kentsfield, and the 45nm Yorkfield is still a pair of Wolfdale dual cores on a single processor package. The processor supports the LGA775 package, runs at a core speed of 3.0 GHz, and uses a bus speed of 1333 MHz.
It also supports basic features like Intel Virtualization Technology, Execute Disable Bit, Intel 64 architecture, and SpeedStep. Although this is a 45nm processor running at a default core voltage of 1.1-1.2V, Intel still uses a 130W TDP. Motherboard support is also a potential bright spot, as most 1333 MHz platforms should handle a 45nm Penryn with a BIOS update, and there is even *unofficial* Penryn 1333 MHz support on older boards from companies like ASUS. Check your motherboard vendor website just to be sure.
The only real surprise is the 3.0 GHz clock speed. Launching a new processor line at a clock speed that has already been reached is kind of anticlimactic. After all, Intel already has the 3.0 GHz Core 2 Extreme QX6850 quad core, and now we get another 3.0 GHz quad core, albeit at 45nm. This is surely a response to AMD and their nonexistent challenge at the top-end of the processor market. The Yorkfield core is certainly capable of higher than 3.0 GHz speeds, but without AMD pushing Intel, why bother?