05-06-2010, 04:53 PM
I don't get what you want to benchmark? You can not compare something that does not work on both systems.
05-07-2010, 02:23 AM
Sorry I was not clear.
Originally Posted by Kano
I don't have a complete answer, but I think we should:
- find a way to test performance of non-graphics apps (for example: opening complex documents multiple times, saving them, converting them to different formats, compressing large files, video and audio conversion, video stream).
For most of these tasks you have cross-platform apps (OpenOffice, Audacity, VLC, GIMP, ... and a combination of them in parallel)
My point is/was that if you test "speed" only using 3D apps, you're bound to see the difference between the graphic drivers, not the OS (and by that I mean, scheduler, disk access, memory management ...)
- when talking about games, I think we should have a look at graphics quality as well. I don;t know if any of the games tested have a DirectX version in Windows or not. If so, I think it makes sense to include them in the picture, as well as doing a "quality" comparison between openGL and DirectX.
You ask "why", given the Ubuntu does not support DirectX? Because if you want to compare games between the two platforms you'd need to consider that Windows does have an "advantage" (for lack of a better word) of offering DirectX which MAY give a speed/quality advantage over OpenGL.
Would it be comparing apples to oranges? Partially yes, but, frankly, because of the large difference in development between drivers and games between Linux and Windows, I think we are already comparing apples to oranges, even if we use the same titles.
Hope this makes a bit more sense.
05-07-2010, 05:35 AM
I will give you another example: I was just reading this review:
You can see a comparison using "everyday" software, most of which is independent from the graphic card.
While some of the tools used in this review are Windows-specific, *all* of the tests can be done with cross-platform tools.
I would love to see the same suite applied, on the same HW, to Ubuntu vs Windows.
When comparing two OSes, I would also add multi-tasking tests with a combination of multiple of the tests run in parallel.
05-07-2010, 06:28 AM
lame would certainly work on both, you just need to use a "reference" binary then, on linux you compile it from scratch using the default compiler. 7zip benchmark could be possibly used as multithreaded benchmark. basically the raw speed should not differ that much on a multicore system. what slows win down is not even benchmarked here as the test system runs without virusscan. depending on the actual used anti-virus tool the slowdowns range from a bit to extreme. especially single core systems suffer from av. the only "game" used in the benchmarks that has got a dx render is unigine, and dx10 definitely looks the same as opengl there.
05-07-2010, 07:06 AM
For audio/video coding you could use other applications: xvid comes to mind. Audacity.
Originally Posted by Kano
You can't compare "winrar" but you could use the command line version which is available on both.
PDF document creation: you can do it in many apps (OpenOfice come sto mind): you don't have to use Adobe.
Same for image processing: you can use GIMP instead of Photoshop.
Only two tools are not available cross-platform: Winzip and Fritz.
05-08-2010, 10:47 AM
Excellent 1st instalment, love your work!
I don't suppose you could compare both from a HTPC/PVR usage perspective?
I'm sure there's lots folks that'd love to know how they compare in this regard!
05-08-2010, 11:20 AM
What would you exactly compare? With pretty much every video card out there having DxVA acceleration you would have to give the nod to windows (XBMC for windows for example has already started to switch to directx for playback) and you usually don't have to worry about items like vsync support. Setup wise, windows pretty much wins there as well. The advantages that linux has usually lies in obscure configurations that allow you to do some pretty nifty things (such as spoofing a iptv hd box with a second nic and capturing the pure stream instead of encoding the recording). Power management also is better and more reliable in windows (at least the newer versions of it). Then of course there is the tools that are available in windows but not in linux for video manipulation after capturing. This is still a very neglected issue in linux (just for example software to detect and remove commercials from the recordings, stream repair, proper frame serving, video enhancement, gpu encoding, etc).
Originally Posted by jalyst
05-08-2010, 11:23 AM
The lack of good video manipulation tools is a real sore point with me in linux. It is one of the major reasons why I keep windows around.
05-08-2010, 10:52 PM
I'd like to see a comprehensive comparison for DVB-T/C/S capture (inc encrypted PayTV), transcoding facilities, VA abilities, ripping, depth & utility of PVR functions (choose myth or VDR?) etc.
Anything one might want to do with a HTPC/PVR....
I think each platforms would have distinct pros and cons, seeing this presented in a methodical/formalised review would be awesome!
05-09-2010, 10:38 AM
So you want to see more of a feature comparison then an actual benchmark?
Originally Posted by jalyst