Are Phoronix's performance articles a bit out of real world?
I know that some people will complain a bit about the way my concerns are but the PTS and the articles about performance lose a lot of substance in many ways.
I mean, not sure about you, but most of human beings do run applications that require a lot of things and not only compiler optimizations. So even the compiler can generate a code that is 10% faster (let's say GCC against Clang) in many cases, how it translates to a real game or a real LibreOffice? The package manager will use LZMA which will depend to this 10% speedup, but it will also depend on disk IO, some update scripts that are written in Python or Bash (which in turn are one or two order of magnitude slower than C/C++ code that is sped up).
As for games, as drivers are compared, this is great, but also, doesn't appear that the news are a bit redundant? There are articles over and over, that do show "closed source" driver vs "open source" drivers, but in many games, if they need the proper performance they will take the close source driver. If the closed source driver will not install properly, the user will reduce screen resolution to the level that performance is adequate.
What do you think?
Michael does run gaming benchmarks: I like them, they show me (roughly) what I can expect to get from a given card in a given game with settings X, Y and Z.
As for what packages people use; at the point where you are no longer using exact timings to measure operations there is the problem of quantifying how fast something "feels".
I also like the regular open vs closed source articles, they show how much well the open source drivers are coming along and how much would be gained by switching to a proprietary driver. Whilst users will just drop resolution until the speed is acceptable, having an idea of how a game will run can be invaluable when considering purchasing new hardware (or trying the game in question).
Yes, I was hoping for a bit more input from others, but it looks to me that from practicality's standpoint, the idea of GCC/LLVM/whatever is not so relevant for users but the video drivers are. Maybe I misrepresent you but it seems that you look and like mostly the gaming (OpenGl) side of testing.
Originally Posted by archibald
I don't want to say that the rest is not important or it doesn't have user's impact, but as an user things that may matter are:
- how fast a package does install
- how fast the system boots
- how well Firefox (or Chrome) works in Linux
- if video works accelerated in Flash or Video element
- if he is a developer, how fast it compiles (it is sometimes shown this, but an analysis is fairly just a progress bar how to compile PHP or Linux kernel)
- if he/she uses Java, how fast it performs against other OSes (or distros, but I don't expect a huge variation)
- should he use Firefox or Chrome? Or Opera?
I mean in this way, the articles will give a deeper richness about what Phoronix is, otherwise, or maybe is just me, it looks a bit too geeky
Yeah, I say that except for those benckmarks which show linux is faster then BSD. What a hypocrite.
Originally Posted by archibald
I would like to see more gaming benchmarks done with game titles that are a bit more taxing on hardware. Seeing over 60fps is a little bit redundant a lot of the time and since Michael is such a huge Steam fan. I would love to see him do benchmarks with say Team Fortress and more modern games.
I know this game wouldn't run with open source drivers for the most part but to see performance differences with different graphics cards on more modern titles would be a godsend.