Slashdot Mirror
Dumping Aqua On Mac OS X For X11?
Sagefire asks: "Aqua is a beautiful interface but it can be incredibly resource intensive (especially for older/low-end machines). And, though the open source community has made great strides in reverse engineering proprietary drivers from Mac OS X, I would love to be able to simply keep using the drivers that came with it, for now. Since there is a fully functional BSD variant under the hood, is it possible (using X11.app, darwinports, and/or Fink) to boot to a command line and simply startx? Would it use less RAM to bypass Aqua?"
Change is certain; progress is not obligatory.
That's because Mach spends a shit of of resources passing messages and servicing real-time interrupts. That has nothing to do with the difference between X11 vs. Aqua in terms of resource consumption. Stick Yellow Dog or NetBSD on that old G3 if you want performance.
There are several Linux distros for Mac hardware, just install one of those. I'd give regular Ubuntu a choice, and if that's too heavy-weight, try Xubuntu.
Ubuntu comes with a lot of software pre-installed, it feels a lot more responsive than OS X on the same hardware, and it has very much a Mac-like feel. I'm running it on an old iMac and have been quite happy with it.
What are you smoking? DPS (and the more recent OS X variant DPDF) goes back to 1988, with the introduction of NeXTSTep on the NeXT cube. Mac OS X is really just NeXTStep. Aqua, however, is an Apple addition and - IMO - is a real improvement. But it's also a resource hog.
A few years ago I compared OS X 10.2 and KDE 3 (YDL) on a G3 All in One, with perhaps a Rage 128 or some other ATI GPU with 2 or 4 MB of VRAM.
With 256 and later 320MB RAM, KDE was much, much faster, by a long shot. It was a shock, since I'd long held the misconception that KDE/Gnome were slow (coming from the days of running Windows 95/NT vs. Gnome/KDE on old Pentiums with 64 MB of RAM).
OS X did not support that machine's video card for any sort of acceleration, and there was no way to turn down the needless eye candy to a level that made the OS usable.
OS X on that machine was slower than Windows 2000 with 48 MB of RAM.
Yeah. I keep telling the scientists where I work, "Do your visualization on Macs; there's no better platform. But do your compute on Linux or one of the mono-kernel BSDs." You're absolutely right that heavy real-time threading tends to make a desktop "feel" faster, but that's simply UI responsiveness. If you want to do heavy compute - where context switching incurs a heavy toll on output - then, a monolithic kernel is still the best approach.
A good analogy is the difference between bandwidth and latency. Microkernels, like as is used with MacOS X, offer the best latency because the message passing allows for very fast interrupts. Just as are monolithic kernels better optimized for long tasks, with as few interrupts as necessary.
Tailor your software to meet your needs.