As interesting as several pages of graphs are, it'd be cooler to read an explanation as to why. Superpages support in FreeBSD? Linux 2.6.38 should have supported it too (as of that version), but perhaps less well?
Also, why did they use different motherboards? For that matter, what the hell is the purpose of providing the page two table of system details if you're going to crop half of it away???
A brief summary of Phoronix benchmarks: nice graphs, meaningless results.
[Seriously, Phoronix benchmarks are often very badly designed/executed. I'm not sure whether it's because they really are completely clueless about benchmarking, or because they just don't care.]
This part is true. It's a spammy site, and annoying.
The criticism of the benchmarks, though, is out of line. All benchmarks suck. Why? Because benchmaking is hard. It takes a ton of patience, a ton of diligence, and a ton of hardware. It's easy to sit back typing on the one laptop you use, running one OS, and whine about someone's inability to be an expert across dozens of environment.
Yet we all want benchmarks, as evidenced by this post. If you guys really think Phoronix benchmarks can be improved (and obviously they can), I'd appreciate it if you'd get off your MBP's and post some.
Where should I be going for Linux news other than Phoronix? I constantly here bad things about it, but I also constantly find out about things that I wouldn't have otherwise. I feel "bad" for reading phoronix or something but I don't know what else to read, and LWN ain't cutting it.
Well, that's the thing. They're the best out there. So personally, I give them some latitude with the spammy ad-driven experience. Most of their numbers are just "more of the same", but occasionally they break something truly interesting and everyone rushes to explain it. This particular post wasn't one of those times though.
According to the thread on the FreeBSD forums linked here somewhere below, the difference in motherboards is just a difference in report on Linux vs BSD.
Other than that, on x86, they seem to compare a PAE-enabled linux kernel to a FreeBSD kernel without PAE. And different versions of Xorg, for some reason. And, yeah, Unity vs KDE.
While most of it shouldn't matter (my money is on "Unity vs KDE makes all other differences negligible"), it's like they didn't even try to give the illusion they were doing things seriously.
In practice, with the current subsystems, it's hard enough to composit (?) windows, let alone unredirect fullscreen windows. For example, there's a tunable in KDE to "unredirect fullscreen windows". Guess how well it works?
If you read the comments on the article, it's pointed out that the results are virtually identical to a previous KDE-vs-unity benchmark (PC-BSD was using KDE for this test).
FreeBSD may or may not be faster than Linux for gaming.. but it does not matter -- freebsd is NOT a desktop OS. I've been using it for 15 years, 14 of those years on my desktop. Buggy desktops, no binary updates for ports, etc, etc. The entire thing is an exercise in frustration.
A decade ago, you could use FreeBSD on a desktop and it would be more or less similar to a Linux desktop. But Linux has more resources, and has invested more than FreeBSD into the desktop.. and FreeBSD has focused their limited resources on the server.
Today it's not even a contest. FreeBSD, if you like it, goes on a server. Linux on the desktop.
Need a workstation to develop the software for your FreeBSD server? - use Linux for the workstation and compile it on a FreeBSD build server. (I'm not kidding.. I actually use this setup for work...)
Some of us still use FreeBSD on a desktop (well, maybe you don't consider xmonad a desktop...) But your point is valid -- it can be an exercise in frustration.
The big advantages Linux has now are mindshare, funding, drivers, and people that devote a lot of time to coming up with OS versions with binary packages that all work well together, so users don't have to build software from source. With FreeBSD that's theoretically possible, but if there's a big public repository of prebuilt binary packages for the latest (9.1) release, I am not aware of it, and my system doesn't know about it. So upgrades can be painful if you have a lot of ports and your machine isn't super-fast. And running into errors in the middle of the process seems par for the course -- then you get to figure out what's wrong and how to fix it.
I have heard good things about PC-BSD (based on FreeBSD) though. Apparently it's much more user-friendly for people looking to use it as a desktop environment.
"if there's a big public repository of prebuilt binary packages for the latest (9.1) release, I am not aware of it"
FreeBSD had a security breach shortly before 9.1.. so some of their hardware was offline. So packages were not available at release. They expect to make them available at some point. Then you can install out of date packages (since packages are frozen at release), and rebuild half of them to fix security or serious bugs.
"I have heard good things about PC-BSD"
I have also heard good things about PC-BSD. I've played with it a bit, and it seems promising. I haven't used it enough to recommend it.. but it seems like it's worth a try if you need a FBSD desktop. Definitely more promising than the other BSD desktop efforts I've seen over the years (all of which are dead AFAIK).
Need a workstation to develop the software
for your FreeBSD server? - use Linux for the workstation
and compile it on a FreeBSD build server.
(I'm not kidding.. I actually use this setup for work...)
Can I ask why do you use a FreeBSD server ? Do you use it for bsd features likes jails or dtrace ?
I've always wanted to try freebsd but I don't really know how it would be better on a server than, say, debian.
There are some things like ZFS, that keep me on FBSD.. but that's only 2 of my servers. The biggest reasons are that it's what I know; and that I have a collection of software projects created over the past 15 years that run on FBSD, most of which do not compile cleanly in Linux or have fbsd-specific dependencies. So until I port them, FBSD is the only option.
I'm not the best person to ask to advocate FBSD. I'm planning a switch to Debian over the next year or so.
Later this year, I'm planning to switch my mail servers and database servers to Debian (since those don't rely on my code). Then the porting should be finished, and everything else will switch except the ZFS servers. The ZFS servers should be gone by mid-2014. Why? Lack of virtualization support is the biggest problem.
I am one of those strange people that still prefer FreeBSD to any Linux desktop/server distribution.
I also used Linux for several years before I started to learn and use FreeBSD, and sometimes I still use Linux, but only when I am payed to do so.
After several years with FreeBSD I even tried to move back to Linux, hearing all these 'advertisements' how good it is on the desktop, how painful FreeBSD is on the desktop etc.
So I used Ubuntu Linux for a whole year on the desktop/workstation without using FreeBSD. My experiences with that quickly got me back to FreeBSD, let me tell You why. First, hangs and crashes that was fixable only by reboot, the sound mostly. I did not turned of that workstation, as that was not needed, so every 2-4 days I was forced to do the reboot just to have the sound back. No matter if I used that sound (play music) during that 2-4 days or the machine just stayed idle, it hanged anyway. Reloading the ALSA modules did not help. Maybe it was a bug, but I always was 'up-to-date' and ALSA and the kernel were upgraded many time, even two time to the 'next big release'.
The other 'awful' thing is the updates. They work the same way like in Windows, for 9/10 times, they fix things, but on the 10/10 You end up with totally broken system that even can not boot, and like with Windows: "With minor problems reboot, with major problem reinstall."
That never happened to me on the FreeBSD land every upgrade/update of the FreeBSD's base system succeed.
There are other distributions You say ... yes there are, but which ones? Linux Mint is basically the same as Ubuntu, but with different default GUI and with some more codecs loaded in by default. I also have an allergy to anything that Lennart did, so for example Arch Linux is dead for me because it uses systemd and pulseaudio. Fedora 'the Lennux'? No thanks.
The whole Linux ecosystem seems broken to me because of all the things explained here: http://www.pappp.net/?p=969
Also, I do not longer want to go back to OSS vs ALSA discussion, where OSS from the FreeBSD base system just works for me with everything I do and ALSA does not on so many ways.
I do not want to 'flame' again the 'initrd' mechanism in Linux where some drivers are in the initrd and some other are in the kernel and others are in the modules, but modules are not for initrd but just kernel, the kernel and initrd is under /boot but the modules are under /lib/modules /... this is just plain mess for me.
On FreeBSD You have one directory with kernel and modules /boot/kernel period. No other subkernels like initrd just to boot and then pass the machine to the 'real' kernel.
Using FreeBSD on the desktop requires knowledge and experience, and that makes it hard to use as a desktop/workstation. PC-BSD tries to change that, we will se how far can it go with it.
A lot of people ask, why You use FreeBSD when there is Linux?
I would ask the opposite, why You use Linux when there is FreeBSD?
I have here ZFS with latest 5000 version, yes, this is the next version after ZFS v28, this is ZFS Feature Flags.
I have ZFS Boot Environments with sysutils/beadm, the same way as it works on Solaris (even better), so I can create a bootable snapshot and destroy everything in my system, even rm -rf /* but after the reboot I may choose to boot from BE created just before that disaster and ... nothing happened.
I do not use DTrace, but many probably do. Jails are very nice thought. I already spoken about OSS in the base system, I already written about deterministic and reliable upgrades. I can also repeat the 'known' properties of FreeBSD like great documentation, great community, a very logical attitude in OS mechanisms and filesystem hierarchy ...
FreeBSD is no panacea to all operating systems problem, it has its own issues, the FreeBSD team has a lot of less resource (and hype) then the Linux world, but taking all the 'cons' and 'pros' its a lot less painful to use FreeBSD then to use any Linux.
Whereas actually they are comparing the speed of Unity in 2011 (notoriously slow) with KDE. This is really dumb, they should be using the same window manager at least.
I mean, they should at least install the same packages on each system if they want to compare "BSD" with "Linux", otherwise this benchmark is spectacularly flawed.
23 comments
[ 2.0 ms ] story [ 50.4 ms ] threadAlso, why did they use different motherboards? For that matter, what the hell is the purpose of providing the page two table of system details if you're going to crop half of it away???
[Seriously, Phoronix benchmarks are often very badly designed/executed. I'm not sure whether it's because they really are completely clueless about benchmarking, or because they just don't care.]
Those tendency combined with the sometimes enormous amount of graphs makes a lot of page views for them.
The criticism of the benchmarks, though, is out of line. All benchmarks suck. Why? Because benchmaking is hard. It takes a ton of patience, a ton of diligence, and a ton of hardware. It's easy to sit back typing on the one laptop you use, running one OS, and whine about someone's inability to be an expert across dozens of environment.
Yet we all want benchmarks, as evidenced by this post. If you guys really think Phoronix benchmarks can be improved (and obviously they can), I'd appreciate it if you'd get off your MBP's and post some.
It would be far better if they'd simply do nothing at all, because as it is, their benchmarks produce misinformation.
Other than that, on x86, they seem to compare a PAE-enabled linux kernel to a FreeBSD kernel without PAE. And different versions of Xorg, for some reason. And, yeah, Unity vs KDE.
While most of it shouldn't matter (my money is on "Unity vs KDE makes all other differences negligible"), it's like they didn't even try to give the illusion they were doing things seriously.
With compiz ~8fps , without compiz ~25fps
In practice, with the current subsystems, it's hard enough to composit (?) windows, let alone unredirect fullscreen windows. For example, there's a tunable in KDE to "unredirect fullscreen windows". Guess how well it works?
https://bugs.kde.org/buglist.cgi?longdesc=unredirect&lon...
http://forums.freebsd.org/showthread.php?t=26204
There also is a section in the FreeBSD Handbook about how the Linux compatibility layer works:
http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/li...
A decade ago, you could use FreeBSD on a desktop and it would be more or less similar to a Linux desktop. But Linux has more resources, and has invested more than FreeBSD into the desktop.. and FreeBSD has focused their limited resources on the server.
Today it's not even a contest. FreeBSD, if you like it, goes on a server. Linux on the desktop.
Need a workstation to develop the software for your FreeBSD server? - use Linux for the workstation and compile it on a FreeBSD build server. (I'm not kidding.. I actually use this setup for work...)
The big advantages Linux has now are mindshare, funding, drivers, and people that devote a lot of time to coming up with OS versions with binary packages that all work well together, so users don't have to build software from source. With FreeBSD that's theoretically possible, but if there's a big public repository of prebuilt binary packages for the latest (9.1) release, I am not aware of it, and my system doesn't know about it. So upgrades can be painful if you have a lot of ports and your machine isn't super-fast. And running into errors in the middle of the process seems par for the course -- then you get to figure out what's wrong and how to fix it.
I have heard good things about PC-BSD (based on FreeBSD) though. Apparently it's much more user-friendly for people looking to use it as a desktop environment.
FreeBSD had a security breach shortly before 9.1.. so some of their hardware was offline. So packages were not available at release. They expect to make them available at some point. Then you can install out of date packages (since packages are frozen at release), and rebuild half of them to fix security or serious bugs.
"I have heard good things about PC-BSD"
I have also heard good things about PC-BSD. I've played with it a bit, and it seems promising. I haven't used it enough to recommend it.. but it seems like it's worth a try if you need a FBSD desktop. Definitely more promising than the other BSD desktop efforts I've seen over the years (all of which are dead AFAIK).
I've always wanted to try freebsd but I don't really know how it would be better on a server than, say, debian.
I'm not the best person to ask to advocate FBSD. I'm planning a switch to Debian over the next year or so. Later this year, I'm planning to switch my mail servers and database servers to Debian (since those don't rely on my code). Then the porting should be finished, and everything else will switch except the ZFS servers. The ZFS servers should be gone by mid-2014. Why? Lack of virtualization support is the biggest problem.
I also used Linux for several years before I started to learn and use FreeBSD, and sometimes I still use Linux, but only when I am payed to do so.
After several years with FreeBSD I even tried to move back to Linux, hearing all these 'advertisements' how good it is on the desktop, how painful FreeBSD is on the desktop etc.
So I used Ubuntu Linux for a whole year on the desktop/workstation without using FreeBSD. My experiences with that quickly got me back to FreeBSD, let me tell You why. First, hangs and crashes that was fixable only by reboot, the sound mostly. I did not turned of that workstation, as that was not needed, so every 2-4 days I was forced to do the reboot just to have the sound back. No matter if I used that sound (play music) during that 2-4 days or the machine just stayed idle, it hanged anyway. Reloading the ALSA modules did not help. Maybe it was a bug, but I always was 'up-to-date' and ALSA and the kernel were upgraded many time, even two time to the 'next big release'.
The other 'awful' thing is the updates. They work the same way like in Windows, for 9/10 times, they fix things, but on the 10/10 You end up with totally broken system that even can not boot, and like with Windows: "With minor problems reboot, with major problem reinstall."
That never happened to me on the FreeBSD land every upgrade/update of the FreeBSD's base system succeed.
There are other distributions You say ... yes there are, but which ones? Linux Mint is basically the same as Ubuntu, but with different default GUI and with some more codecs loaded in by default. I also have an allergy to anything that Lennart did, so for example Arch Linux is dead for me because it uses systemd and pulseaudio. Fedora 'the Lennux'? No thanks.
The whole Linux ecosystem seems broken to me because of all the things explained here: http://www.pappp.net/?p=969
Also, I do not longer want to go back to OSS vs ALSA discussion, where OSS from the FreeBSD base system just works for me with everything I do and ALSA does not on so many ways.
I do not want to 'flame' again the 'initrd' mechanism in Linux where some drivers are in the initrd and some other are in the kernel and others are in the modules, but modules are not for initrd but just kernel, the kernel and initrd is under /boot but the modules are under /lib/modules /... this is just plain mess for me.
On FreeBSD You have one directory with kernel and modules /boot/kernel period. No other subkernels like initrd just to boot and then pass the machine to the 'real' kernel. Using FreeBSD on the desktop requires knowledge and experience, and that makes it hard to use as a desktop/workstation. PC-BSD tries to change that, we will se how far can it go with it.
A lot of people ask, why You use FreeBSD when there is Linux?
I would ask the opposite, why You use Linux when there is FreeBSD?
I have here ZFS with latest 5000 version, yes, this is the next version after ZFS v28, this is ZFS Feature Flags.
I have ZFS Boot Environments with sysutils/beadm, the same way as it works on Solaris (even better), so I can create a bootable snapshot and destroy everything in my system, even rm -rf /* but after the reboot I may choose to boot from BE created just before that disaster and ... nothing happened.
I do not use DTrace, but many probably do. Jails are very nice thought. I already spoken about OSS in the base system, I already written about deterministic and reliable upgrades. I can also repeat the 'known' properties of FreeBSD like great documentation, great community, a very logical attitude in OS mechanisms and filesystem hierarchy ...
FreeBSD is no panacea to all operating systems problem, it has its own issues, the FreeBSD team has a lot of less resource (and hype) then the Linux world, but taking all the 'cons' and 'pros' its a lot less painful to use FreeBSD then to use any Linux.
If You are interested in FreeBSD des...
I mean, they should at least install the same packages on each system if they want to compare "BSD" with "Linux", otherwise this benchmark is spectacularly flawed.