Mac OS X 10.6.3 vs. Windows 7 vs. Ubuntu 10.04 Benchmarks
Phoronix: Mac OS X 10.6.3 vs. Windows 7 vs. Ubuntu 10.04 Benchmarks
Last week we delivered the first of our Windows 7 vs. Ubuntu 10.04 benchmarks to much anticipation, but now we have the results for Apple's Mac OS X 10.6.3 operating system to tack in too. In the first part of that Windows 7 vs. Ubuntu Linux performance examination, we looked closely at the OpenGL gaming performance across six different systems and a whole slew of tests. More articles are on the way looking at the performance and later in the week we already delivered some initial disk benchmarks. However, now it is time to see how Microsoft Windows 7 Professional x64, Ubuntu 10.04 LTS, and Apple Mac OS X 10.6.3 compete with one another.
The results from openarena look a bit strange. Maybe the compiler flags for Linux release needs to be tuned a bit compared to the win build or nvidia drivers really have got enabled optimization for the ioquake engine depending on the exe name, which would be easy to test, just rename it to something else
Overall results are hardly surprising. Microsoft carefully listens to (usual, not open source) gamedevs and puts enough pressure on hardware vendors (directly or indirectly, via its APIs) to streamline their drivers. It also does rigorous testing to ensure that the requirements are met. No surprise that Windows drivers are the ones with the highest quality.
I am not sure whether the similar organization can be created for an open source OS. It would take a company like Red Hat, but dedicated to games. Community lacks organization for the massive QA needed to produce quality drivers itself and clout to make vendors provide binary drivers of better quality.
Looking at the speed at which open source drivers are produced, I don't expect to see a stable open source driver of DX11 functionality (complete with DirectCompute, that is, GPGPU (OpenCL) support) for the next two years (and proprietary devs will move on to something new by that date).
To run a cross OS benchmark on a single hardware means very little.
It's a lot like testing GPUs with a single game
To give an idea on why I believe that: I've run Sunspider with Opera 10.53 on a MacBook (C2D T8300 2.4 GHz) and I got these scores:
- OS X Snow Lepoard 10.6.3: 426.2 ms
- Windows 7 Pro x86: 499.4 ms
I've asked to a couple of friends to run the same test on their Windows 7 laptops:
- C2D E2180 2.0 GHz: 552 ms
- Sony Vaio FZ21M 2.0 GHz: 442.2 ms
Too me benchmark means very little without multiple hardware comparison. Sometimes on slower/different hardware you get better results, clearly it's some type of driver/bios/whatever issue, so I wouldn't take the Phoronix results for granted.
Your milage may vary a lot.