OS X 10.10 vs. Ubuntu 15.04 vs. Fedora 21 Tests: Linux Sweeps The Board

[Greijoan]

Would it be interested to see some BSD on that benchmark, for some of the tests presented here, BSD is pretty slow as well.

Exactly my thought, OSX benchmarked against FreeBSD would be pretty interesting.

[Luke_Wolf]

I hope to hear from some Apple authority (not necessarily directly working in the company) what he/she has to say about those results. I would understand 5% differencies, but this?

Part of the bad performance seems to be linked with worse code generation of Clang, because Fedora + Clang exhibits similar tendencies, sometimes. It's still better than OSX though. Maybe it's because the compiler of the kernel?

Anyway I was very surprised by this benchmark.

By the way: does anyone even uses OSX Server edition?

Anyone who has paid any attention to benchmarks of OS X versus any other major OS isn't going to be surprised by these results. It's slow for much the same reasons as it is the most insecure major OS (check out every Pwn2Own ever). Security and Speed are not so much a question of design (although design greatly helps) as much as finding the holes and bottlenecks in production and fixing them. OS X due to the size and nature of its userbase are not doing that, whereas Windows and Linux are hammered daily and as a result are having those areas fixed.

That said I agree it would be interesting to see how FreeBSD compares.

[zoomblab]

Does the first page of benchmarks show a Linux 4.0 performance regression?

[chrisb]

Does the first page of benchmarks show a Linux 4.0 performance regression?

Which result? The only result that stands out is PostMark, and that could be accounted by reasons other than a regression e.g. if the benchmark is done on different partitions for each OS, then the disk might be marginally faster or slower responding (it should be faster as the data gets farther out towards the edge of the disk). If you really wanted to investigate, you would run the test reading/writing a freshly formatted partition, and always the same limited size block device, then if you still replicate a difference try on a ram disk, try some other IO benchmarks etc. It can be a lot of work to verify and track down something like that.

btw The way to quickly compare results is by looking at the standard error bars. If the bars overlap, the results are the same. If the bars do not overlap, there is >95% chance of results being different:



Running the benchmark more times will lower the standard error (i.e. shorter bars on graph), so any potential differences are more obvious and more likely to be true.

[bwat47]

Interesting. The supposedly platform-dedicated OS doesn't win a single benchmark against a completely "generic" one.

Exactly my experience with OSX versus Arch on a 2013 Macbook Retina. Linux was very noticeably faster at all tasks.

in my experience on a 2013 mba, linux (fedora 21) does indeed run faster but also runs WAY hotter and consumes WAY more battery, even after spending hours tweaking with powertop, tlp etc... even when consumed wattage was relatively low (8-9 according to powertop) and very low cpu usage (0-5% idle according to top) it would be noticeably hot to the touch just idling and the fan would start going, in osx it always runs very cool and the fan almost never comes on, and on osx the battery life is insanely good.

Performance isn't the only measure of optimization, on a mac OSX is very highly optimized for battery life and running cool, top notch touchpad integration/drivers, and the performance is 'good enough' in most cases.

[sstuparus011]

Is it me?

It seems to me that tests were conducted on two different machines. The specs table is somewhat confusing.

[Michael]

It seems to me that tests were conducted on two different machines. The specs table is somewhat confusing.

Read the article (especially right below the table....). It was ALL THE SAME MACHINE, as always. Just a matter of how each OS exposes the hw info.

[Ansla]

It's hard to say if MacBooks are overpriced since competition for them is completely inexistent. Did Apple patent the laptop with less than 15" display, SSD and backlit keyboard? Oh, and also below 2 kilos.

I already own a desktop, so I don't need a desktop replacemnt but a mobile device. But since noone except Apple makes such devices, and I would never buy an Apple, I just use my phone (what some would call phablet) for mobile computing.

[liam]

I'm going to guess you are an apple fan

If there is one thing most people can agree on it is that apple is expensive. Overpriced comes as a natural feature from that if you don't use the product as a fashion statement. Everyone knows fashion cannot be overpriced because in that world more expensive == better.

And everyone knows fashion is not supposed to be practical, it is just supposed to be. So there is no need for apple to be fast. The important thing is that it must be different, so why not add cheesy compiz animations for the ui that hide the slowness in a design feature?

I don't, and refuse to, own any apple products, but I certainly wish other companies would try to make such uncompromising products. Asus has some tremendous laptops, and the new Google pixel is near perfect (tiny, not upgradable sad), but there just aren't many options out there for thin, light, powerful, and well made laptops.
As for the price, they really are cheaper than pretty much anyone else if you keep build quality and components in mind.
Their screens are excellent, tremendous battery life and Wi-Fi antennas. Possibly the best track pads around (though that gap has closed, and is perhaps now nonexistent). Keeping all that in mind I've not been able to find a laptop that is either as cheap, let alone cheaper, than a macbook.
Their obscene buying power let's them buy massive runs at factories and maintain their margins since the products get sold.
Osx is extremely flexible. They have a very vibrant add-on community that developers make use of that rivals Linux. If you don't know what I'm taking any talk to someone you know who is a developer and uses Mac.
Frankly, this is why I get so angry at gnome. They piddle their time away with puffery instead of creating things like Alfred, automator and the rest. Things that actually, genuinely make a difference.

[liam]

in my experience on a 2013 mba, linux (fedora 21) does indeed run faster but also runs WAY hotter and consumes WAY more battery, even after spending hours tweaking with powertop, tlp etc... even when consumed wattage was relatively low (8-9 according to powertop) and very low cpu usage (0-5% idle according to top) it would be noticeably hot to the touch just idling and the fan would start going, in osx it always runs very cool and the fan almost never comes on, and on osx the battery life is insanely good.

Performance isn't the only measure of optimization, on a mac OSX is very highly optimized for battery life and running cool, top notch touchpad integration/drivers, and the performance is 'good enough' in most cases.

That's been a serious sore spot for Linux. Most of its development has focused on performance above all else. It's only fairly recently that there's been serious work to get the frameworks in place to make Linux efficient.
There's basically two areas that need work: drivers and software (apps but also the supporting framework like KDE and gnome). The driver work has been going on for several years now, mainly with the effort to get QoS in place (I believe all the framework-level stuff is done but drivers need to be adapted to take advantage). The scheduler could certainly be better, but if the desktop frameworks were better at being chatty and batching work, it wouldn't be needed.