Mac OS X 10.6.2 vs. Ubuntu 10.04 Performance

[phoronix]

Phoronix: Mac OS X 10.6.2 vs. Ubuntu 10.04 Performance

While we are just weeks away from delivering the most comprehensive Mac OS X vs. Windows 7 vs. Linux benchmarks, and Apple is on the heels of releasing the major Mac OS X 10.6.3 update, for those impatient ones today we have published an extensive set of tests comparing the performance of Mac OS X 10.6.2 against a development build of Ubuntu 10.04. This is our first time exploring how Canonical's Lucid Lynx can compete with Apple's Snow Leopard.

http://www.phoronix.com/vr.php?view=14723

[RobbieAB]

So on identical hardware, you see identical performance on CPU tests, and slight to dramatic differences on stuff that relates to the kernel.

Given the nature of the pgbench test, if it is a known problem on ext4, why is ext4 being used to run it? Are people running serious PostgreSQL databases stupid enough to stick with the default file-system for a Linux distro when it is demonstrably inferior? In Mac OSX you have little choice about the file-system, in Linux, it's pretty easy to switch file-systems for critical applications.

So yeah, way to go, include a server benchmark in a desktop comparison, and don't worry about the fact that the system is almost deliberately skewed against Ubuntu on that benchmark.

[portets43]

So on identical hardware, you see identical performance on CPU tests, and slight to dramatic differences on stuff that relates to the kernel.

Given the nature of the pgbench test, if it is a known problem on ext4, why is ext4 being used to run it? Are people running serious PostgreSQL databases stupid enough to stick with the default file-system for a Linux distro when it is demonstrably inferior? In Mac OSX you have little choice about the file-system, in Linux, it's pretty easy to switch file-systems for critical applications.

So yeah, way to go, include a server benchmark in a desktop comparison, and don't worry about the fact that the system is almost deliberately skewed against Ubuntu on that benchmark.

a linux newcomer is probably going to stick with the defaults.

as in; people that use mac generally don't customize it because, well, it's not very customizable. and a lot of people that use ubuntu generally don't customize things like filesystems because, well, they're new to linux.

it seems like micheal just likes to test out-of-box experiences. you could run the tests yourself the way you like them, maybe even start a site to compete with phoronix. for more advanced users.

[pvtcupcakes]

a linux newcomer is probably going to stick with the defaults.

as in; people that use mac generally don't customize it because, well, it's not very customizable. and a lot of people that use ubuntu generally don't customize things like filesystems because, well, they're new to linux.

it seems like micheal just likes to test out-of-box experiences. you could run the tests yourself the way you like them, maybe even start a site to compete with phoronix. for more advanced users.

What RobbieAB is saying is that a normal Ubuntu user doesn't care about Postgre performance, even if they're running their own personal website. If I'm the IT for a mega corporation's website that gets hundreds of thousands of views per day, I would have the knowledge to not use Ext4 and Ubuntu for my Postgre DB.

But I also wouldn't be using a Mac Mini like in the test.

[portets43]

ah, okay.

but still, what i said applies to a lot of previous posts in threads. there's just way too much complaining here sometimes.

[RobbieAB]

pvtcupcakes pretty much has my point. Why include a benchmark that is meaningless and only really has the effect of making one platform look bad? With the exception on the pgbench, Ubuntu is actually giving Snow Leopard a decent run, given that Snow Leopard should be optimised far better for the hardware.

[kraftman]

So on identical hardware, you see identical performance on CPU tests, and slight to dramatic differences on stuff that relates to the kernel.

Given the nature of the pgbench test, if it is a known problem on ext4, why is ext4 being used to run it? Are people running serious PostgreSQL databases stupid enough to stick with the default file-system for a Linux distro when it is demonstrably inferior? In Mac OSX you have little choice about the file-system, in Linux, it's pretty easy to switch file-systems for critical applications.

So yeah, way to go, include a server benchmark in a desktop comparison, and don't worry about the fact that the system is almost deliberately skewed against Ubuntu on that benchmark.[/QUOTE]

It should be even better with debugging disabled on Ubuntu and with different ext4 mount options, but the defaults are benchmarked and this seems to be sane.

[RobbieAB]

I have no problems with the rest of the tests. It strikes me as "Oh, Ubuntu is competitive with Snow Leopard for performance".

What I objecting to was the inclusion of a decidely non-desktop benchmark that makes Ubuntu look really bad in stock configuration when anyone who actually cares about the performance on that benchmark won't be running it on stock Ubuntu. What does including that server benchmark in a desktop comparison actually achieve?

[RealNC]

It was a good idea to include that benchmark in the test. When something gets negative publicity, chances of it getting fixed increase.

[kraftman]

It was a good idea to include that benchmark in the test. When something gets negative publicity, chances of it getting fixed increase.

But there's probably nothing to fix, it's probably just matter of configuration.