![]() ![]() ![]() It's easier to do multiple things in parallel than to do one thing faster, which is also why today's CPUs have multiple cores. (See this classic piece from Ars cofounder Jon Stokes for background.) Then CPUs became superscalar and started executing more than one instruction at a time. Life was good, but not particularly fast. They executed instructions that were read from memory one at a time. Twenty years ago, CPUs were nice and simple. Still, it's cool that the lowly Air can run the same OpenCL code as that Radeon 5870 number crunching monster GPU, allowing developers to create code once and then run it anywhere-just not on Intel's HD 3000 GPU. The GPU gets a Luxmark score of 4299, the CPU gets 6280, and when they team up the total is 10754. Dave Girard, Ars Graphics and Imaging Guy, on the other hand, has a monster 12 core 2.66GHz Mac Pro with a Radeon 5870 GPU. The MacBook Air gets a reasonable CPU Luxmark score of 1252. Even using both the GPU and the CPU at the same time is slower than only using the CPU, if we can believe the Luxmark OpenCL benchmark. I'm not sure how big a deal that is, though: on my older machines, the CPU can run OpenCL code much faster than can the GPU. As far as I know, this makes the MacBook Air the first Apple computer in several years that doesn't support general purpose computing on the GPU through OpenCL. ![]() ![]() There is no OpenCL support in the Intel HD 3000 GPU. Hopefully we'll soon see Quick Sync support in video encoding software. And the worst part is that Intel's current generation of CPUs have special video encoding hardware on board, known as Quick Sync. It's a real shame that video encoding doesn't benefit from the Air's increased CPU prowess to the same degree as synthetic benchmarks. Standard definition video at 640x360 encoded using the universal preset runs at 79 frames per second, 60 percent faster than the MacBook's 49. That's 50 percent faster than Handbrake does on my 2008 MacBook, where it hovers just under 16 frames per second. On the new MacBook Air, encoding an HD 1280x720 video using the iPhone 4 preset runs at about 24 frames per second. Under Lion, iTunes 10.4 can no longer use QuickTime to play videos, so it's often necessary to convert video to H.264. There is still tons of video out in the world in formats that the iPhone and other Apple devices can't decode. The most CPU intensive real-world task that I need my computer to perform is video encoding. Last year's 1.4GHz Core 2 Duo 11" MacBook Air scored a tiny bit better on the GPU test at 11.25, but did much, much worse on the CPU test with a score of just 0.81. The 1.8GHz Core i7 MacBook Air scores 11.07 on the GPU test and 2.32 on the CPU test. Maxon's Cinebench, on the other hand, runs only two tests: it uses OpenGL to play a short cut scene of two cars driving around city streets (which runs on the GPU), and it has a multithreaded ray tracing test (which runs on the CPU). I'm not a huge fan of the way Geekbench works, taking a number of different benchmarks and calculating an average result. The Core i5/i7 family has come a long way in a short time. (All these results were taken from the Primate Labs Mac benchmarks results page.)Ī comparison against last year's line-up is more favorable. Of course a more sobering comparison is against Apple's current lineup of portables, with the fastest and the slowest iMacs thrown in, too. My three-year-old MacBook and four-year-old MacBook Pro both score in the low 3000s, so this Air is almost twice as fast as those. Geekbench 32-bit output for the MacBook Air ![]()
0 Comments
Leave a Reply. |