The Mac at 25: Processors
Editor’s Note: Celebrating the Mac’s 25th anniversary means more than just looking back—it also requires us to look ahead to what your computer will be like in the years to come. We now turn our attention to the processors that power your Mac.
Microprocessors are the brains that power computers, and graphics processing units (GPUs) help provide fast, high-quality graphics display. The faster they both are, the happier users tend to be.
Microprocessors and GPUs evolve in two primary ways: chip makers shrink the distance between transistors (that distance is known as a chip’s process), which leads to faster and more-efficient chips, and they change processor architecture to improve how processors manage data and instructions, which increases overall performance.
Most of today’s Macs are powered by the Penryn family of chips, which are built using a 45nm (nanometer) process and based on Intel’s Core architecture. They’re fine chips, but Intel’s new 45nm Nehalem architecture features a host of improvements that should significantly boost overall performance. Expect to begin seeing Macs with Nehalem chips in 2009.
Nehalem’s power is especially apparent when you’re running software that takes advantage of multithreading—the ability to split tasks into multiple concurrent parts (threads). Nehalem microprocessors will have two, four, or eight cores, and will be able to simultaneously work on two software threads in each core, using a scheme called hyperthreading. Theoretically, an eight-core Nehalem chip can simultaneously process up to 16 threads—an ability that should make your next Mac quite snappy.
Unfortunately, most software can’t yet take advantage of such extreme multithreading. But fear not—help is on the way. First, Nehalem’s Turbo Mode senses when some cores are idle because software can’t take advantage of multithreading, and it will both shut down the unused cores and increase the clock speed of the ones that are hard at work. Second, Snow Leopard will include Grand Central, a technology that’ll help developers better exploit multithreading.
Multithreading is a trick that GPUs have already mastered—their multicore structures use shader instructions designed for highly parallel tasks such those required for rendering 3-D game frames. Traditionally, there have been three types of shaders—vertex, geometry, and pixel—but many GPUs are now moving to a unified shader system that offers more flexibility and better performance.
GPUs are fast, but their powers have been mainly reserved for drawing graphics. With Snow Leopard’s OpenCL technology, however, the unused clock cycles in their speedy cores will serve your Mac’s CPU. Adobe has already built this trick into Photoshop CS4, enabling the app to perform some amazingly swift image transformations. When OpenCL arrives, more apps will have access to the power of your Mac’s GPU.
One of Intel’s most interesting multicore projects, code-named Larrabee, is a hybrid architecture with the programmability of a CPU and the parallelism of a GPU. Intel isn’t saying how many cores Larrabee processors will have, but it has shown projected-performance graphs describing Larrabee processors with up to 64 cores. The first Larrabee chips will be tuned for high-performance graphics and video and will appear in 2009 or 2010—just in time for Snow Leopard.
After Nehalem and Larrabee, Intel plans a 32nm process named Westmere in 2010, a new Sandy Bridge architecture in 2011, a 22nm process named Ivy Bridge around 2012, and a new Haswell architecture after that; further in the future, the company plans to shrink processes to 16nm and then 11nm, and to develop new architectures that add more—and more-specialized—cores. If Intel sticks to its public plans, the future should be bright.
[Rik Myslewski has been writing about the Mac since 1989. He has been editor in chief of MacAddict (now Mac|Life), executive editor of MacUser and director of MacUser Labs, and executive producer of Macworld Live. He now writes for The Register.]