Is there research online to the performance gained with Gent

[dpowers]

I think it's a fair question. The allure of Gentoo is thought of a ports-based system, where package source code is compiled with optimizations, all with a fancy schmancy interface. But Debian has just a just as good/if not better package management scheme, so it basically comes down to the fact that Gentoo compiles, and that's supposed to be a lot faster.

So is it? Are there benchmarks or documentation out there that show the performance gained over a distribution with binaries? And is this performance gain REAL WORLD?

(Don't get me wrong by the way. I love Gentoo. I've learned a lot from playing with it, and it is my primary OS.)

[AutoBot]

I used Debian Woody for a good while and have been using gentoo for a month or so, and it is faster in every aspect. I don't have any benchmarks to prove it but think of it this way:

Your driving your car 55 mph for say an hour and then you speed up to 75 mph for say 5 minutes, do you realize your going faster for those 5 minutes?

>>That may not make any sense to you, but it is the best way I can compare "55 mph" Debian to "75 mph" Gentoo
_________________
This message self destructed a long time ago.

[dpowers]

That could be all in the mind. I was talking to someone who said the difference is <10%. And if that's the case, is it worth going through a 3-4 hour KDE compile over a 1 minute apt-get?

[Guest]

i think beyond the fact that it can be optimized for a particular cpu and that you compile the kernel with less fat perhaps then a stock kernel but i ahrdly think any of it would give you as much of a performance increase as just getting new hardware...what you probably would notice is on lighter systems with out lots of extra memory and power gentoo will run better because its demands are less.
but i ahve all new stuff....so would i see the difference really...probably not...i'll take the 55mph vs 75mph thing a bit further....i had a camaro that i rebuilt adn it would go from 0 to 100 in a little under 6 seconds....a friend of mine had a vette that would get there 3/10ths of a second faster you think you could feel that...even though its there and mesurable would it be really practicle to dwell on it....lol

[AnimalMachine]

If you're not the type of person that likes to squeeze performance out of your machine, then Gentoo probably isn't the best distro for you (IMHO).

If you have a high-end machine and only want a desktop to browse the web and do email, then any prebuilt distro (Debian, Mandrake, or Redhat) will probably suffice, and you won't have to worry as much about system setup.

I still maintain that Gentoo is quite a bit faster (and much more usable) than Mandrake on my P233-MMX / 128mb RAM laptop. So while I had to put 3 or 4 days into boostrapping and compiling X, KDE, and Gnome along with a few other tools, the payoff in speed is well worth it. And once it's set up, it's not like you have to update it by recompiling stuff all the time.

I run Gentoo on my desktops now instead of Debian because I'm the type of masochist that likes recompiling everything to get the extra performance. I may be wierd, but I think that it's fun.

[tod]

Hi!

This is an interesting topic.

Some relevant links are:

http://www.tux.org/bench/

http://lbs.sourceforge.net/

http://oss.sgi.com/LDP/HOWTO/Benchmarking-HOWTO.html

and

http://gcc.gnu.org/benchmarks/

Unfortunately, I don't think these projects are very active


For what its worth, I have found that the -On flag (go figure, ) has a more significant impact on performance that the -mARCH flag. This is based on my playing (very non-rigorous) around with the MDBNCH program and is only applicable to FORTRAN code.


It might be interesting to create a Benchmark Forum here where people could post their results from various benchmarking software.

Regards,

tod

[dpowers]

hey tod thanks for answering! I got ripped a new one in IRC #gentoo for even bringing it up. It's good to know that some developers are giving the thought some attention though.

I think it would be an excellent idea to create a benchmarking forum. At the very least, we can start to see how much of a performance gain Gentoo brings to the table.

Thanks and keep up the good work!

[AutoBot]

[Al'Capone]

Sorry I couldn't help it I needed to jump in on this one

Let me start off by clarifying that performance is a generalized and objective term that depends of different and lots of variables , while speed and velocity are relative terms.

To use the analogy of the car going at different velocities: the car going faster is not necessarily a high performance vehicle.

and this is where it gets interesting. what is a better performance car:
the 60K one that does 0-60 mph in 4.5 seconds using a V-12 engine, ultra sport suspension large and special brake kit, and a half decent transmission ?"A'la Dodge Viper" and consumes tons of fuel (and resources) in the process
or the 35K every day little car weighing half the pounds that does it in 6 seconds using a 4 cilynder engine and consuming only perhaps 1/3 the fuel resources ? A'la Honda 2000

well I like to believe that a good performance car has a lot to do with the engineering and the works under the hood rather than big engines, big brakes, big car, Heavy parts, etc.

The Lotus Elise is a very high performance car hand made slim, nice looking, small engine and fast as hell (darn bad we can get them in North america ) The honda 2000 is the highest performance 4 cilynder engine ever produced period. and it is nicely matched to other components that help in the performance numbers.

Gentoo Linux is the little Honda 2000 car that could it is the streamlined process it uses to compile, only what you need, the optimizations to your specific hardware requirements, and that includes, (not only your CPU) your video card, use of ram, motherboard chipset, bridge, IDE controllers, etc, and more fundametal "the use of source to compile " instead of pre-compile binaries that makes it a high performance distro

I like to think of my compiled gentoo on a toshiba labtop P-III 850 mhz, 128 mb of RAM, with shared 8 mb of Video RAM (128-8=120mb RAM) 15 gig 5400rpm ATA-4 HDD, Ali Magik sound card and cd-dvd combo as the high performance Honda or Lotus car

and my SUSE 8.0 sitting beside it running on AMD T-bird 1200 Mhz with 512 mb of PC-266 RAM, 60 Gig 7200 RPM HDD running at 266 FSB, with 32 mb of Nvidia Geforce II video card SB live, and tons of other crap in it as the Dodge Viper (gentoo is going there too just have to burn all the Mp3 shit that i have in it first)

so you may ask which one is a better performance car? the Big T-bird AMD running SuSE 8.0 ? or the little toshiba running Gentoo?

Well Gentoo on the little guy of course, because there is not bloatware, and gazillions of services running, everything is optimized, and compiled from sources Hence everything is snapy and without a fuzz considering the hardware.

which one is faster ??
The big Dodge Viper running SuSE of course! i have a mechanic friend who say "there is no replacement for displacement"

however you can only imagine when i put gentoo in the big Viper

Hope that helped

Cheers
_________________
if it aint broken..... then... what the heck are you waiting for? ...go mess with it and see if you can make it better

[AnimalMachine]

Well ... Will any benchmark really be an adequate illustration of performance gain?

From www.tux.org/bench:
"There are three kinds of lies: lies, damned lies, and benchmarks"

[AutoBot]