Sign in with
Sign up | Sign in

Benchmark Results: Media/Transcoding Apps

Experiment: Does Intel’s Turbo Boost Trump Overclocking?
By

We’ve upgraded our test suite to the latest version of Apple’s iTunes (9.0.2.25), but the program’s behavior remains unchanged. It’s still not optimized for threading, and as a result, Hyper-Threading goes to waste here.

On the other hand, utilization of only one CPU core means Turbo Boost has a fairly profound effect on performance in iTunes. So too does locking each chip’s top speed in at 4 GHz. It’s cool to see theory apply in practice, and that’s what we have here.

Unfortunately, you’ll see that iTunes is the exception in our well-threaded benchmark suite. Let’s move to something a little more parallelized.

MainConcept takes advantage of as many threads as you throw at it. Even still, with Turbo Boost disabled, the Core i5-750 runs at its default 2.66 GHz and the i7-860 is clocked at 2.8 GHz. Although this benchmark is taxing all four cores, operating under its thermal and power ceiling means we still get one bin (133 MHz) of acceleration with Turbo turned on, which is why both CPUs show better numbers with the feature in play.

More so than Turbo Boost, Hyper-Threading gives the Core i7-860 a sizable advantage over the i5-750—proof positive that in well-threaded software, there is a reason to pay a little extra for Hyper-Threading.

Overclocking minimizes the different between the two CPUs, though. At 4 GHz, both shave off significant time off of the stock settings. Of course, the Core i5 sees most of the benefit, since it wasn’t already getting a sizable speed-up via Hyper-Threading.

We compound the results here by mixing a well-threaded DivX codec with a less-optimized Xvid codec.

As expected Xvid sees no benefit (and, in fact, runs a bit slower) with Hyper-Threading enabled on the Core i7-860 versus Intel’s i5-750. Turbo Boost does help speed things up on both CPUs, though.

Interestingly, DivX doesn’t seem to get anything out of Hyper-Threading either, suggesting its limit is four threads. In this case, the Core i7-860 is a tad faster. And both processors get major speed-ups from overclocking—enough so that it’s clear manual overclocking is the best way to improve the performance of threaded apps that don’t get much benefit from Turbo Boost.

HandBrake is new to our test suite. It’s an open source title, freely available, that’s able to take advantage of threading. In our test, we’re converting the first .vob file on The Last Samurai into an .mp4.

Because it’s threaded, Turbo Boost has little effect here. But again, it’s interesting to see that Hyper-Threading doesn’t make as much of an impact as it did in, say, SiSoftware’s Sandra or 3DMark Vantage’s CPU suite. The real way to fly is using manual overclocking—there’s a huge performance improvement to be had by upping the clocks on both of our tested CPUs.

React To This Article