Intel Core 2 Extreme QX9650 - first look

Posted at: 3:49am 29th October 2007 by James Gorbold

Meet the world's first 45nm Penryn CPU ...

It’s hard to think of something that Intel has done wrong this year – it’s released a massive selection of aggressively priced processors, ranging all the way from the bargaintastic Pentium E21-series to the ludicrously overclockable and cool running G0 stepping Core 2 Quad Q6600. Meanwhile, AMD has struggled to get a single quad-core desktop CPU out the door, only managing to launch a few disappointing Barcelona architecture Opterons last month. However, despite the fact that Intel seems to be left to compete with itself, it’s already prepared to start shipping the industry’s first 45nm CPU this month – the Core 2 Extreme QX9650.


LET’S GET PHYSICAL

We could start off by with all sorts of crass and clichéd statements about how much smaller a 45nm transistor is than a human hair or an amoeba – but what really matters is what benefits 45nm CPUs have over 65nm CPUs. However, to understand why 45nm transistors are such an important step forward, it’s important to understand how they’re made.

Although it’s only been two years since Intel introduced the first 65nm CPUs, the company has been working on the development of the 45nm manufacturing process for close to ten years. This incredibly long development time is because it’s becoming increasingly hard to make smaller transistors.

Without delving too far into theoretical physics, it’s fair to say that the process of scaling down transistor size has been fairly straightforward over the last few decades. That’s a gross simplification of course, but around the early 1990s Intel began to realise just how hard it would be to make transistors any smaller than 65nm. This is because the SiO₂ (silicon oxide) used as the gate dielectric between the silicon substrate and polysilicon gate in a 65nm transistor is just five atomic layers thick, and to make it any thinner would mean current would easily flow through it.

As a result, Intel had to find a replacement for SiO₂; the material that’s been used as a gate dielectric since the 1960s. The replacement for SiO₂ turned out to be a so-called high-k gate dielectric, the precise make up of which is a very closely guarded secret, although Intel confirmed to us that it’s based on Hf (hafnium). At the same time, the polysilicon gate was replaced by a metal gate. These two advances meant that Intel was able to scale down the size of the transistor as a whole, and make much more energy efficient CPUs, as the gate leakage was cut by around a factor of ten.

At the same time, 45nm transistors can switch off and on almost 20 per cent faster than 65nm transistors, so a 45nm CPU can scale to much higher frequencies than a 65nm CPU. In addition, because each individual transistor is smaller, the CPU die is smaller, which makes it cheaper to produce as more dies can be fit onto a single wafer. For example, a 65nm dual-core Core 2 Duo has 293 million transistors and is 143mm² while a 45nm dual-core Core 2 Duo is just 107mm² despite having considerably more transistors (410 million).