Fast Forward: Rethinking Graphics
I’ve had fun shopping for graphics cards, especially when a power user is within earshot. I’ll innocently ask the salesperson, “What’s your slowest graphics card?” The reaction is precious.
As I’ve confessed before, I’m not a gamer. Years ago I edited a videogame magazine and didn’t realize how weary I had become of games until the magazine unexpectedly folded. I stopped playing that day and haven’t resumed since. That’s why I don’t need fast graphics. Playing a YouTube clip is the most taxing graphics workload demanded of my computer.
Often, I won’t even buy a graphics card. I’ll scrounge a hand-me-down from a friend or cannibalize a junked PC. My oldest computer in regular use contains a discarded engineering sample of an Nvidia GeForce4 Ti-4200 from 2002.
Are you cringing yet? Mock me no more, power users. I’m reconsidering my wayward ways.
Not that I’m renewing an interest in games. I’m still such a nerd that I’d rather punch code than shoot pixels. No, what’s making me waver are the nongame programs for GPUs. This software is enabled by Nvidia’s CUDA platform for general-purpose computing on GPUs, AMD’s similar ATI Stream, Microsoft’s Direct Compute API, Apple’s Grand Central Dispatch, and Khronos Group’s OpenCL. I’ve been around the industry a while and don’t use the term “revolutionary” lightly, but GPU computing is the most exciting thing I’ve seen in years.
It also changes the equation for PC shoppers. There are legions of us nongamers, you know. For years we’ve been happy with crappy graphics. But now, if we want to clean up our amateur video, we need a graphics card that can run MotionDSP’s vReveal. If we anticipate transcoding much of that video, we’ll want a system that can run Elemental Technologies’ Badaboom. If we’re editing high-res digital images, we’ll crave a GPU that accelerates Adobe Photoshop.
GPU computing will alter the priorities of users and system designers. Bargain PCs that safely economized on graphics suddenly seem less attractive. Eventually, even casual users will grasp that the performance of some apps will depend more heavily on the GPU than on the CPU. And that’s definitely a game-changer.
Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.