Nick Race29 September 2006, 2:50 AM
IDF San Francisco |Intel gave us a few minutes today to test their hot-damn quad-core desktop processor. And the results were... fricking good (as you'd expect from Intel's choice of benchmarks on an Intel-built machine).
After the Tuesday announcement of Intel’s move from two to four cores for their performance CPU, there’s only been one question running round and round IDF and the web. Just how fast is this sucker?
Though there’s been some “leaked” pre-production models floating around the web, and some rather interesting politics going along with it, these things are always likely to change at the last minute. Well, the last minute is up, and Intel gave the opportunity to a number of media outlets to have a small slice of one on one time with the new chip.
Similar to the preview APC had of the Core 2 Duo CPU in Singapore earlier this year, Intel had set up a number of preconfigured test benches in private quarters off the main press hall. Though as time was tight, press weren’t allowed to get real deep and dirty with the new chip.
But here’s what we did get: two machines, a Core 2 Duo Extreme at 2.93GHz (stock) and a Core 2 Quad (codenamed Kentsfield) running at 2.66GHz. The Quad runs with a whopping 8MB of cache (to the Core 2 Duo’s 4MB) on the same 1066MHz front side bus.
Both chips were built into identical systems based on an Intel BadAxe2 motherboard (D975XBX2 Rev 303) with a single 1GB nVidia GeForce 7950GX2 card. A Seagate 7200.10 hard disk drive and two sticks of Corsair CM2X1024-8500C5 RAM running at 4-4-4-12 make up the other pertinent parts.
Intel had done the legwork, and printed a form containing the estimated benchmarks for both systems - so our business there consisted mainly of verifying their data, and making sure there was no funny business was going on. Please, keep in mind this is not an “official” APC testbench, but systems put together by the least impartial party.
We fired up PCMark05 and went through a basic run with a CPU test. The Kentsfield pulled out a 7562, to the Core 2 Duo’s Intel estimated 7689, a tad slower, but still within a margin of error. The CPU test of the Kentsfield gave us a tasty 8485, to the Core 2 Duo’s 7425, a quantifiable increase.
We ran POV-Ray, a well regarded, and well aged 3D rendered developed here in Australia with the built in benchmark. The Kentsfield system ran through at 2527 pixels per second, very far ahead of the Core 2 Duo’s 1428 pps. When watching the benchmarks side by side, the difference is obvious.
Intel had provided an HD encoding benchmark based on Sony’s Vegas 7.0a. This combined numerous streams of HD footage, with effects and titling. Once again, the Kentsfield took over. The render took 254 seconds, to the Core 2 Duo’s 382.
Specific results aside, what this really shows us is the Core 2 Quad is going to be a killer for heavily threaded applications.
The benchmarks Intel presented are all known to use multiple threads (most often to make use of multiple CPUs in workstation machines), so seeing a direct increase in speed was to be expected.
Though we’re far from comfortable making a strong conclusion based on this extremely small subset of benchmarks (in an uncontrolled environment, no less) I’m happy to say Kentsfield looks good to go.
