I’ve been a fan of AMD’s 64-bit chips for about three years, when I started working with systems equipped with them. My affection increased when Acer released its Ferrari 4005 laptop, a fast system built atop the 2GHz AMD Turion mobile processor.
I still love the Ferrari, but working with Windows Server 2008 and Windows Vista has necessitated more RAM and more speed, so I recently acquired an HP nx6325 laptop running an AMD Turion 64 X2 Mobile Technology TL-60—a mouthful of a name for a chip that's basically two 2GHz Turions on one chip. The system is definitely fast; that was no surprise. No, the biggest surprise was that it’s cool.
According to the indispensable SpeedFan utility, my single-Turion Ferrari tends to run at a temperature of 60 to 80 degrees Celsius. That sounds fairly hot, but according to the AMD Web site, the Turion is rated to run at up to 95 degrees Celsius. That’s why I was so surprised to discover that the nx6325—with what amounts to dual Turions—tends to stay between 45 and 55 degrees Celsius. How did AMD pull that off? Why, the same way Intel does with its 32-bit Core Duo and 64-bit Core 2 Duo CPUs—by taking advantage of the power of something called “nonlinear heating rates.”
About two years ago, Intel announced that it had pretty much hit a brick wall in terms of CPU speed, and that it couldn’t see a way to exceed about 3.8GHz of speed on a Pentium-family chip. One of the root problems was heat. If you’ve ever tried to “overclock” your CPU—run it at a faster speed than it was rated for—you know that you haven’t a prayer of making it work without a big fan to keep that overclocked CPU cool. (Or you could adopt the approach of some true “speed fans” by using water-cooled CPUs.) A faster clock rate yields a hotter CPU—no surprise. What is surprising, however, is the rate at which the CPU heats up.
Raising a CPU’s clock rate from, say, 3.4GHz to 3.6GHz—about a 5 percent increase—can increase its heat output by 60 percent, according to an Intel guy I spoke with recently. That leads us to the trick that both Intel and AMD are using to get higher speed and cooler processing.
On one hand, AMD has gotten better at making 2GHz laptop processors, and so I’m told that if I were to somehow swap my Ferrari’s circa-2004 2GHz Turion with a modern one, I’d see cooler processing. Alternatively, I could just crank the chip up to 2.2GHz and I’d be back up in the sixties and seventies. (This is all theoretical: It would involve soldering that would require some very expensive hardware.) The better answer—and the one that HP adopted—was to install a Turion X2 and hold it to 2GHz rather than 2.2GHz.
I’m simplifying a bit, but the heat savings works both ways: Just as cranking up a clock by 5 percent increases temperature by 60 percent, dropping the clock rate by 5 percent nets about a 60 percent decrease in temperature. Again, that’s a simplification, and you see that kind of savings only when you drop the clock to a point near the chip’s maximum usable clock rate. But apparently dropping the clock rate on a dual-core system by a mere .2GHz can cause each of the two processors in the chip to drop its power usage by 60 percent, with the effect that you now have two processors running at 2GHz but drawing less power than a single-processor chip would while running at 2.2GHz.
In other words, dual-core processor chips, when utilized in a wisely architected system, generate less heat (which indirectly means they consume less battery power) while at the same time delivering better performance.
I’ve been something of a skeptic about multiprocessor laptops because I haven't been convinced that real-world applications and OSs can typically make use of more than one processor. But the fact that I, like many of you, am running virtual machines on my laptop—combined with the multiprocessor architecture of Vista—has made me a real believer. (And not having to endure thigh burns while using my laptop on my lap is also nice plus!)
