Printable View
Those are only graphic drivers tests, so if an nVidia driver is better optimized for Windows it can be even 0 - 10, 0 - 50 ;) Personally, I think the most interesting will be some parallel benchmarks and benchmarks like Apache, PostgreSQL, Sqlite not interesting at all (at least using Ext4 file system).Quote:
Originally Posted by Apopas
Well, I could take it that if I install one of these OSs on my hardware without performing any tweaking, I should expect Windows 7 to run pretty well comparatively. Sure I guess I could tweak Linux (minorly or heavily), or I have to install other stuff on Windows which will slow it down, or my particular hardware/driver setup will affect results...Quote:
Originally Posted by susikala
But mostly I don't care, nor do most users. Sure different setups might produce different results, but this is a great start... and probably reasonably representative of the state of the world. It's a great place to have that dialog about how well Linux performs in the real world (on the desktop), and what the factors are.
If people go away from this to explore and tweak their little part of Linux (and contribute it back), things will get better. If this switches the dialog away from 'Linux is unquestionably awesome don't ever mention/compare/suggest a feature found in Windows' into a rational discussion of what end users experience (those who have no interest in tweaking their machine). If this starts to establish benchmarks or ways to have rational comparisons between OSes.... it's a good thing.
At the very least, it shows how Michael has expanded his fabulous test suite into a cross platform tool --- which is incredibly invaluable. Now we have another tool to test the performance of open source software across multiple platforms - how is that not a good thing.
So if anything is a waste here, it's the attitude that there is nothing of value -- be open to what value it does bring an we can build on that. This is just a start - being so critical is not helpful.
This is not a universal 'fact'. And considering that Linux can be customized to remove all the crap that consumes resources to fit the form factor/use case of a commercial product, even where it is true it might not be true in the end product.Quote:
Originally Posted by susikala
Regardless, if it gives us a place to measure progress from, or perhaps gives relative performance compared to the rest of the broader world of operating systems (which we are not isolated from), this is a great start!
It looks like NV drivers are better optimized for Windows... So fglrx are "friendly" for 3D in Linux than NV drivers (on my HD4850 I get the same FPS both in Windows Vista x64 Home Premium and Ubuntu 10.04 x64)
It would be interesting (perhaps) if we could re-compile one of the Gallium3D drivers using the Windows winsys (I imagine there must be one out there, even if not open source) to see what the performance is. [ie, how much of the performance differences with some games are within the driver itself versus the other supporting Windows library and/or Kernel]
Vista-bashing in 2010 is quite pathetic.Quote:
Originally Posted by Apopas
Games on Vista get about the same FPS as XP or Windows 7, sometimes more sometimes less but overall it's the same.
Personally, I think that those tests are meaningless.Quote:
Originally Posted by kraftman
Who's the moron that runs Apache on Windows 7?
Who cares about performances of those apps on a client OS?
If the tests were run on Windows Server it would make sense, but on Windows 7 is just totally useless.
A benchmark I would love to see: Photoshop CS5 on OS X and Windows 7... you know... real stuff that matters ;)
You are quite mistaken. I have been working on a cross-platform 3d VR application that is currently running roughly 5% faster on Ubuntu compared to Windows XP. With fglrx of all things!Quote:
Originally Posted by susikala
If you don't the articles, don't read them and/or stop donating to Phoronix. Complaining in the discussion thread for the article results does nothing but annoy those of us who *are* interested in those results.
On the Server results could be worse :) If the server must guarantee data will be immediately written to the disc and on the desktop there won't be so strict mode used, desktop OS will get better results thanks to caching. The same if you do some benchmarks in virtual machine - it can give far better results then a physical host, but those results will be meaningless. When comes to server benchmarks I wouldn't test the default OS's settings. Some applications like Firefox uses SQLite and Amarok uses MySQL for default. I'd like to see some encoding, decoding, unpacking, compressing, computation benchmarks. Speed of creating, deleting, copying files, directories. This is what's important on desktops IMHO.Quote:
Originally Posted by Enrox