If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
Yeah, I can see how Canonical really has Apple shaking of fear over that MAFFT benchmark. Protein and DNA sequence alignment performance has always been a key deal-breaker in the desktop computer business :-)
What I meant is, what user scenarios were emulated here to illustrate what the two competing platforms bring to the table and tell potential users/customers which one to opt for depending on their own particular priorities?
As often the case with these benchmarks, the average performance delta between the compared entities tends to be something like 10% or less. The weird metric (IMHO) and small differences aside, is the person for whom such differences are relevant even a likely consumer of either product?
Michael and Phoronix insists on sticking to the defaults for things like Ext4/PostgreSQL to emulate out-of-the-box experiences for so-called average users (we assume). This may apply well to some cases, and worse to others. From a Desktop user's perspective, wouldn't a comparison pitching the performance of a typical Free program pipeline, say Ubuntu/Gimp against a Proprietary one, such as Mac/Photoshop be more relevant to illustrate their competitive properties? Or how about things like battery life, Wifi reception, speed of suspend/resume cycles, screen management, audio quality and so on? Surely, Phoronix has the skills necessary to benchmark these too.
I don't mean to rain on anyone's parade, and maybe it is just me being ignorant here, but what am I supposed to be able to extract from this mix of benchmarks? The presentation itself provides no information as to what kind of aspects of the OS that is actually tested and limiting for each test (CPU, memory, IO etc), we are assumed to know that. Some of tests I rant about above will likely be included in the upcoming three-way benchmark, I just hope that they put some more effort into putting things into context in the presentation. This one should really read:
"This is our first time exploring how Canonical's Lucid Lynx can benchmark against Apple's Snow Leopard."
Is it weird that I came to the opposite conclusion? I've seen previous benchmarks of Mac OSX vs Ubuntu on here and Mac OSX always won by a clear margin.
For the first time, the lines are blurred and Ubuntu actually won in most of the benchmarks and certainly the more important ones. The only ones it lost was a test that is known to fail on Ext4 and a ray tracing test (which I do consider important). Almost everything else was a tie or a win in favour of Ubuntu.
No, it's not weird. In fact, that is what your result should be based on the benchmarks run. The only reason for including the pgbench I can think of is to try and make Ubuntu look bad, as it is totally different to the rest being a server benchmark.
Also, Ubuntu getting close to Snow Leopard is not a huge surprise to me. Snow Leopard involved a major change in the compiler targets and as a result is having some serious teething problems in my experience, so of which are pretty bizarre. (Problems saving wireless configs which worked near perfectly in Leopard? I mean, WTF?)
I'll tell you what's dragging Linux down: ATIME. If we had it Ingo Molnar's way, ATIME would have dissappreaed when compiling. Now, even if noatime is specified in fstab and taken into consideration when mounting the FS, the kernel still spins its wheels thinking it has to write a f*cking timestamp after every single disk access.
That's what Linux gets for catering to server loonies who can't evolve from using rsync and mutt. And desktop users have to suffer.
Snow Leopard's biggest selling point was "Oh hey, we've applied the Rixstep tweaks, and we've tuned performance, and now the whole system is multithreaded and some of it runs on the GPU for extra acceleration".
Ubuntu 10.04's biggest selling point is "There's no brown anymore, and you can tweet from the desktop" :-)
The Postgre benchmark is interesting; but Big Iron will often use a database on a raw block device.
And there's one benchmark between HFS+ and Ext4 that wasn't tested: Time before catastrophic filesystem failure. If my experiences are anything to go by, you're lucky you got through the benchmark without your Mac's filesystem going belly-up ;-)