Thinking about giving gentoo a go

Reasons · 2008-04-25 16:11:03

I've always thought about trying out gentoo but never really had a reason to try; arch works just fine for me. However, I was wondering if it would be worth it now that my specs are a bit lower: 1.06GHz Dual core, Intel 495 graphics, 2GB RAM. I figured that if I compiled everything from source it would run a bit snappier. If so, how long would it take to compile (this would all be in July/August) a fully working KDE 4.1 enviroment and would gentoo even be desirable as I go off to college in September?

brebs · 2008-04-25 16:29:07

See if these stupidities dissuade you

People often notice better performance, simply by changing distro, because reformatting means getting rid of the built-up disk fragmentation that Linux zealots pretend doesn't exist.

Compiling from source for the precise CPU was of benefit many years ago, but not now that GCC has improved.

Dealing with compile-time errors is a PITA. And then people mess up the toolchain and haven't a clue how to recover it, outside of a reinstallation

fiod · 2008-04-25 17:02:42

My suggestion:

Stick with arch, and when you want to compile something for a snappier behavior (like, a commonly used app, etc.),
simply use ABS.

Gentoo simply doesn't worth the headache.
The performance increase is minimal, and the time consumption is terrible.

jbromley · 2008-04-25 17:29:16

Don't do it! It seems that these days, the main benefit of Gentoo is that you can customize the building of each package with USE flags and not the extra performance. In fact any extra performance you might get out of Gentoo will surely be dwarfed by the time spent compiling. It seems that the only real reason you would use Gentoo would be that you have to customize every application because generic binaries lack features you need. I don't know how likely that is. I've recompiled just a few packages for this reason in 10 years of using binary distributions (I do development.) I believe if this level of customization is needed you could get it by building packages through ABS.

Regards,
j

Misfit138 · 2008-04-25 18:00:20

brebs wrote:

See if these stupidities dissuade you
People often notice better performance, simply by changing distro, because reformatting means getting rid of the built-up disk fragmentation that Linux zealots pretend doesn't exist.
Compiling from source for the precise CPU was of benefit many years ago, but not now that GCC has improved.
Dealing with compile-time errors is a PITA. And then people mess up the toolchain and haven't a clue how to recover it, outside of a reinstallation

Brebs, could you expand on what you mean by "GCC has improved" ?
Could a source-based proponent turn this argument around to say that since GCC is improved, then optimizing for architecture would yield even more benefit?

brebs · 2008-04-25 19:05:17

I'm being deliberately vague, to avoid the huge complexity of optimization and over-optimization - so, basically, gcc does a decent job with -O2 compiled for i686/amd64. And particular apps which benefit from different CFLAGS now include them as the default, much more than say in 2002 (and if they don't, then the binary distros will force/patch them in).

Binary distros now use the same go-faster tricks that the Gentoo l33t used to use Without the user having to waste days in speed/stability testing & recompilations, with every program.

It's possible to compile from source in every distro, with custom gcc flags, and there are user-friendly tools. The Gentoo ricer myth should be dead - heck, the typical Gentoo user is far too lazy to even learn how to write an ebuild.

Twist my arm, and I'll admit that "optimizing for architecture" could, potentially, bring speed improvements of say a couple of percentage points (which would have to be measured, otherwise it wouldn't be noticeable). But, for an individual geek, it would be far easier, quicker and saner to just buy a faster CPU.

floke · 2008-04-25 19:53:09

And what about defragmentation?
I've never noticed anything above a few % non-contigious on an fdisk (admittedly I dont' have masses of Gs of films etc.).
I've never heard many people say it doesn't happen at all, just that ext3 is resistant to it, which it is.

TheBodziO · 2008-04-25 20:06:28

brebs wrote:

People often notice better performance, simply by changing distro, because reformatting means getting rid of the built-up disk fragmentation that Linux zealots pretend doesn't exist.

It's not the linux thing anyway—it's the fs thing . Some fs'es are predestined to fragment files some are smarter. Granted: when the space is about to run out the files will have to be fragmented .

@Reasons: If you have time and will—go for it! It's a good thing to experiment!

If it comes to performance gain, frankly I don't think it will be visible over general arch performance. I've done a switch from gentoo to arch about 2 years ago and haven't noticed any slowdown or stability degradation. But it's up to you to verify it! After all you don't have to get rid of arch to try gentoo as long as your disk resources are critically low .

About compilation time I can say that on my ol' Athlon XP 1700+ with 256 megs of ram it took me about 4-5 days to build a working system with… I can't remember gnome or kde. Some of this time was idle waiting for me to get back from somewhere . You can considerably shorten this time by using distcc (check gentoo pages for details). I know a case when a guy installed gentoo with x and some de in about 3 hours with the help of his friends machines and distcc grid (I was a student back then and this guy lived in a dormitory—a nice place to get both: cheap processing power and windows viruses ).

Misfit138 · 2008-04-25 21:17:39

brebs wrote:

I'm being deliberately vague, to avoid the huge complexity of optimization and over-optimization - so, basically, gcc does a decent job with -O2 compiled for i686/amd64. And particular apps which benefit from different CFLAGS now include them as the default, much more than say in 2002 (and if they don't, then the binary distros will force/patch them in).
Binary distros now use the same go-faster tricks that the Gentoo l33t used to use Without the user having to waste days in speed/stability testing & recompilations, with every program.
It's possible to compile from source in every distro, with custom gcc flags, and there are user-friendly tools. The Gentoo ricer myth should be dead - heck, the typical Gentoo user is far too lazy to even learn how to write an ebuild.
Twist my arm, and I'll admit that "optimizing for architecture" could, potentially, bring speed improvements of say a couple of percentage points (which would have to be measured, otherwise it wouldn't be noticeable). But, for an individual geek, it would be far easier, quicker and saner to just buy a faster CPU.

Ah. Thanks for the reply. This seems like the "common sense/uncommon wisdom" that I personally subscribe to. I certainly can't understand or get into the whole source-based thing, but so many people love this kind of stuff.

brebs · 2008-04-25 21:37:22

Misfit138 wrote:

but so many people love this kind of stuff.

Hah - "people" == "stupid kids with an attitude problem, attracted to a performance myth that died years ago, and now one toolchain compilation short of disaster".

Just to balance this out: Gentoo will still be useful for big installations which want multiple copies of the same Linux installation which they can test beforehand for days on their own testbed PCs.

neowolf · 2008-04-26 14:30:23

I can't even wager a guess as to how long it'll take you to compile KDE. (Though KDE does tend to be a hefty compile historically.) It really would depend on how you set up your system. That being said though, seriously count me in the camp of loving Gentoo, but using Arch and loving it more for a reason. It's just easier and the extra benefits of using Gentoo are just too small for me, and I'd wager for even most power users this is going to ring true quite loudly. The most time consuming thing to learn about managing Arch is just the flags you use with pacman. With Gentoo getting a sane install is going to require learning how to specify custom use flags for system utilities. Being able to search the package database in a time frame that even resembles sanity requires using anything but portage itself. Not to mention it's far, far easier to write a PKGBUILD than an ebuild.

Again, don't get me wrong, I really love Gentoo. It's one of the first Linux distros to really win my favor and in a big way. But I'm not using it these days for a reason.

bender02 · 2008-04-26 15:39:37

I used to use gentoo before arch, and I tell you, binary distro is a huge relief after gentoo. I needed to recompile relatively often, in order not to experience a breakage (not the whole system though), compiling OO, xorg (it wasn't split back then), kdelibs, kdebase was a pain... I decided that what I pay for electricity and cooling is enough already, and that I don't want to waste time and money for no visible benefit

But I guess you need to try to scratch that itch... so go ahead and try it out!

bgc1954 · 2008-04-26 15:51:09

If you're really, really bored and have alot of time to kill, Gentoo is good for a go. If I really need to kill alot of time, I'll do a Gentoo install just to keep in practice--but that's it--and it's always been a minimal install with openbox or something like that and it still kills a day. Alot of walking away and coming back later to see how much is left to go. If you're into that sort of thing, as I am occasionally, go for it but as others have said, you'll notice minimal, if any, speed improvement.

oli · 2008-04-27 18:01:05

brebs wrote:

See if these stupidities dissuade you
People often notice better performance, simply by changing distro, because reformatting means getting rid of the built-up disk fragmentation that Linux zealots pretend doesn't exist.
Compiling from source for the precise CPU was of benefit many years ago, but not now that GCC has improved.
Dealing with compile-time errors is a PITA. And then people mess up the toolchain and haven't a clue how to recover it, outside of a reinstallation

Most distros are using patches of Gentoo if they need support for new compilers (like the upgrade from GCC3 to GCC4) and so on. So keep calm and the flame aside. Gentoo is a very valuable distro and you will find their know-how, especially while experiencing problems, in many other distros.

bender02 · 2008-04-27 18:53:39

If you want to learn something new and see some other approaches to various low-level stuff and are into compiling - go for FreeBSD

neowolf · 2008-04-28 02:12:44

oli wrote:

Most distros are using patches of Gentoo if they need support for new compilers (like the upgrade from GCC3 to GCC4) and so on. So keep calm and the flame aside. Gentoo is a very valuable distro and you will find their know-how, especially while experiencing problems, in many other distros.

He wasn't really knocking on Gentoo there. He listed some genuine complaints, and that the speed difference isn't that big a deal. Nothing about the distro not having value. No need to look for stones being tossed when they aren't.

yesterday · 2008-05-01 15:57:14

I think the joke goes that all the time you gain with optimizations in Gentoo you lose waiting for the bloody thing to compile...

ProzacR · 2008-05-01 17:43:35

brebs ~2-3% fragmentation is nothing. In windows you sometimes have ~40% and still it is ok.

kensai · 2008-05-01 23:35:42

I only have fun on Gentoo using ~x86 (Testing) since Stable seems to be too outdated for my taste. All in All, Gentoo was my favorite for years, but well, Arch Linux just satisfies me now, I just love Arch Linux. But, is true Gentoo has its things, like USE flags, they are just awesome, and a great selection of packages. But even me being an avid Gentoo user, can feel quite a pain to maintain.

iphitus · 2008-05-02 04:12:48

The thing I realised after a while about 'optimising' things....

Is that in nearly all cases, the time spent optimising, outweights any time gained by those optimisations.

Suddenly... they're not worth it.

Redroar · 2008-05-02 04:23:06

USE flags are at once awesome and kind of annoying....way too much to keep up with for me. They can let you do great things, but sometimes ebuilds have default USE flag settings that are ridiculous, and when you're looking at a list of 40 possible flags (for more intricate programs), it just gives me a headache...

handy · 2009-06-29 09:24:31

brebs wrote:

Just to balance this out: Gentoo will still be useful for big installations which want multiple copies of the same Linux installation which they can test beforehand for days on their own testbed PCs.

My ISP, which is the most highly regarded by users in Oz(tralia), run their servers on Gentoo, which then host virtual machines.

Allan · 2009-06-29 09:43:26

@handy: The last post in this topic was over a year ago. Don't bump such old topics.

Arch Linux

#1 2008-04-25 16:11:03

Thinking about giving gentoo a go

#2 2008-04-25 16:29:07

Re: Thinking about giving gentoo a go

#3 2008-04-25 17:02:42

Re: Thinking about giving gentoo a go

#4 2008-04-25 17:29:16

Re: Thinking about giving gentoo a go

#5 2008-04-25 18:00:20

Re: Thinking about giving gentoo a go

#6 2008-04-25 19:05:17

Re: Thinking about giving gentoo a go

#7 2008-04-25 19:53:09

Re: Thinking about giving gentoo a go

#8 2008-04-25 20:06:28

Re: Thinking about giving gentoo a go

#9 2008-04-25 21:17:39

Re: Thinking about giving gentoo a go

#10 2008-04-25 21:37:22

Re: Thinking about giving gentoo a go

#11 2008-04-26 14:30:23

Re: Thinking about giving gentoo a go

#12 2008-04-26 15:39:37

Re: Thinking about giving gentoo a go

#13 2008-04-26 15:51:09

Re: Thinking about giving gentoo a go

#14 2008-04-27 18:01:05

Re: Thinking about giving gentoo a go

#15 2008-04-27 18:53:39

Re: Thinking about giving gentoo a go

#16 2008-04-28 02:12:44

Re: Thinking about giving gentoo a go

#17 2008-05-01 15:57:14

Re: Thinking about giving gentoo a go

#18 2008-05-01 17:43:35

Re: Thinking about giving gentoo a go

#19 2008-05-01 23:35:42

Re: Thinking about giving gentoo a go

#20 2008-05-02 04:12:48

Re: Thinking about giving gentoo a go

#21 2008-05-02 04:23:06

Re: Thinking about giving gentoo a go

#22 2009-06-29 09:24:31

Re: Thinking about giving gentoo a go

#23 2009-06-29 09:43:26

Re: Thinking about giving gentoo a go

Board footer