Ubuntu, Linux Mint, and gOS Benchmarks

[phoronix]

Phoronix: Ubuntu, Linux Mint, and gOS Benchmarks

gOS and Linux Mint are two of the many Linux distributions based upon Ubuntu, but they provide their own spin of things. gOS, for instance, ships with WINE and Google Gears by default and focuses upon providing an easy and rich experienced catered around Web 2.0 services. Linux Mint ships with its own set of customizations and its focus is on providing an easy-to-use Linux desktop by having a distinct user interface, its own set of system, and shipping with various proprietary drivers, plug-ins, media codecs, and other packages. We had a question though from a reader asking whether the performance of these Ubuntu derivatives is vastly different from Ubuntu itself. With that inquiry, we have run a couple benchmarks comparing the performance of Ubuntu 8.10, gOS 3.1, and Linux Mint 6.

http://www.phoronix.com/vr.php?view=13471

[Vadi]

Crap! Now I don't know what OS should I switch to to get a 4% gain in encrypting 4gb-block files. :(

[kraftman]

I would love to know why in last benchmark gOS has such big advantage above others. And I would love to see real world benchmark of Ubuntu 8.10 (or newer with EXT4) vs Macos. Can't wait to Linux crush it.

[russofris]

The difference in performance between distributions can usually be attributed to:

1: Different versions of the application or it's underlying dependencies
2: Different compilation options when compiling the application or dependencies.

One of the neat things about gentoo is that you get to fool around with these options (and often break things in the process). For example, FLAC encoding is almost twice as fast when compiled under ICC7 as it is with GCC4. Granted, doing so can break some applications that depend on libflac.so.

I've often wanted to set up a test system that would do iterative compilations of different applications with various cflags (and compilers) to determine which cflags resulted in the fastest working binaries per app, and which simply resulted in a broken (or half broken)binary.


Frank

[dm0rais]

Good review michael. It was on 32 bit platform right?
Looking forward for some 64bit reviews.:cool:

[kraftman]

The difference in performance between distributions can usually be attributed to:

1: Different versions of the application or it's underlying dependencies
2: Different compilation options when compiling the application or dependencies.

One of the neat things about gentoo is that you get to fool around with these options (and often break things in the process). For example, FLAC encoding is almost twice as fast when compiled under ICC7 as it is with GCC4. Granted, doing so can break some applications that depend on libflac.so.

I've often wanted to set up a test system that would do iterative compilations of different applications with various cflags (and compilers) to determine which cflags resulted in the fastest working binaries per app, and which simply resulted in a broken (or half broken)binary.


Frank

Thank you. It explains a lot. It looks that compiling can give opposite results than expected :)