Geek Patrol

Frequently Asked Questions

Ever since the first release of Geekbench, people have been firing a lot of questions our way. We thought it best to address some of those questions.

Geekbench

Why bother writing a benchmark app?

We were unhappy with other benchmark apps for Mac, and few others that we could find were cross platform. So, we wrote one. Pretty much that simple.

What are you trying to accomplish?

Most benchmarks fall into one of two categories. Synthetic, or real world. Synthetic benchmarks are often written to maximize the results, that is to reach the theoretical maximum performance a system has to offer. Real world benchmarks are usually in the form of encoding times for MP3s or video.

What we’re going for is somewhere in the middle, a real world synthetic benchmark if you will. Geekbench is written in platform neutral code, and is not optimized for any one platform or CPU. In other words, it doesn’t do anything fancy. But then, neither do most applications.

Geekbench is a synthetic benchmark, that makes it useless you know.

Ok, first off? That’s not a question. Secondly, it’s not useless but if you want to think that no one is forcing you to use it or read our web site.

What compilers do you use when building Geekbench?

Geekbench is built with what we consider to be the standard compilers for each platform. Visual C++ 2005 for Windows, GCC 4.0.1 on Mac OS X and GCC 4.0.0 on Linux. Geekbench reports what compiler was used to build it when it runs.

We also use the compiler switches recommended by the compiler vendor for release code.

You know that [insert CPU here] would score better if you optimized for [insert technology here] right?

Yes, but the again the goal of Geekbench is to provide as balanced as possible results. As such, a version optimized for any one technology won’t be released.

Why don’t you release your source code?

Because we don’t want to. If we release the source, then people might download it and built it themselves. They might also use their own set of options/optimizations. Then, there might be a bunch of versions floating around, all producing results that can’t be compared.

Where by “might” I mean “would undoubtedly.”

In other words, by not releasing the source we’re trying to make sure that results generated by whoever can always be compared to results generated by whoever else.

Your tests seem to favour machines with more than one CPU or core!

The majority of Geekbench tests run with one thread. On those tests, a multi-CPU or multi-core machine has little or no advantage. The rest of the tests run with four threads, allowing multi-CPU and multi-core machines to show what kind of performance gain you can expect when running more than one application, or running an application that is multi-threaded.

Geekbench Comparisons

You guys are really pro-Mac!

Also not a question. And no, we’re not. We do happen to have more Macs on hand than PCs though.

You guys are really pro-Windows!

Still not a question. I’m not even sure how this came up (but for the record, it did).

Why don’t you exercise more control over your testing environment?

We don’t have a testing environment, we have users. People run Geekbench on their machines and then submit the results to us. We then draw from those submissions to create comparisons.

Why didn’t you include [insert CPU here] in your comparison?

Again, we don’t have a testing environment, we have users. If our users don’t have access to certain processors, then neither do we.

The comparisons you make are way off base, the machines you use have wildly different amounts of RAM!

Geekbench only uses 100 MB of RAM, which means that as long as a system has 100 MB or more of free RAM the results will never differ.

---