Slashdot Mirror
Quartz Extreme with Unsupported Video Cards
BandwidthHog writes "This thread over at Ars Technica discusses a simple .plist hack to enable Quartz Extreme on the PCI version of a supported video card, i.e. the original Radeon PCI and Radeon 7000, two of the most popular video cards for those of us running on 'unsupported' OldWorld machines."
-A.
What did the walrus say to the penguin? "No soap, radio."
Just a note: I've not actually used this hack since I'm running a G4/867 and a Dual 1GHz Quicksilver. I just posted the link to more information (and because I think xlr8yourmac.com is a great resource; I like to promote it. ;) )
If you follow the link, it looks like most people are just using plain Radeon PCI cards and having pretty consistent success. I mentioned the artifacts only because one person complained of minor screen weirdness (the "preview" in column view wouldn't display correctly) but I wouldn't say that problem is even necessarily the fault of the hack.
With the reduced bandwidth of plain PCI versus AGP, however, I wonder what the real benefits will be. But it looks like it certainly doesn't hurt! Good luck.
-A.
What did the walrus say to the penguin? "No soap, radio."
I think those with Macintoshes with 66MHz PCI slots will see more benefits than those with the 33MHz variety. This is probably why some people see no difference with QE enabled versus disabled.
I did the hack last night. I have both a supported AGP and "unsupported" Radeon PCI. The PCI is my second display. Initially the hack turned QE on for the PCI but OFF on the AGP. That of course was undersirable, but simply removing the preference fixed that and both displays are QE enabled.
For example the orignal is set up as:
<key>GLCompositorRequiredClasses</key>
<array>
<string>IOAGPDevice</string>
</array>
They suggest you change IOAGPDevice to IOPCIDevice. But to make it work on both just remove it, like:
<key>GLCompositorRequiredClasses</key>
<array>
</array>
What did that do for me? Well for one thing I've seen the same artifact issues with column view quicktime previews as others have reported. It's no big deal however. I also had a kernel panic upon my initial reboot after enabling the hack. I rebooted again and it was fine and has been ever since.
I've not noticed a large speed increase, but it is a bit better. What you do get are some effects that are normally turned off when not using QE (transition fading when using automatic wallpaper switching), etc.
I'm sure there are reasons why this wasn't enabled by Apple...besides them wanting everyone to buy a new Mac. But all in all the hack does work and is worth checking out.
John
I have a PCI Rage on my PowerBook, so I'm not gonna bother trying.
The point is though that enabling QE on PCI will take away bandwidth from other PCI cards, such as storage, sound, and networking.
GPL Deconstructed
Flurry is a good one with OpenGL particles. Make a transparent yellow-on-black Terminal window, run 'top', sit back and groove on the juicy goodness.
Can't you see that everyone is buying station wagons?
Unfortunately, people have gotten their brains stuck on the amount of VRAM a card has. It's not so much the VRAM as it is the functionality of the video card (well, more VRAM does help, of course. But it's not the limiting reagent).
-jon
Remember Amalek.
Often these restrictions are placed simply because of the amount of testing that is required. Also the PCI cards are all non Apple standard (although some where sold in the apple store) but this means that you would have to test: Each Different Video Card In Each different Machine Type In different slots With different PCI setups Now I am assuming that there are 5 dfferent video cards sold capable of QE (I just made this up), there are at least 6 setups of hardware architecture (counting revisions in G4 desktops (3), B&W desktops (2), Beige Machines). 3 slots per machine, and a standard set of PCI cards to test against (say 6 cards eg SCSI, IDE controllers, Video Input Cards). Now that makes 540 different tests that needs to be performed. Now if each test only took 30 minutes from setup to pull down, that makes a month of testing for one person. Now that occurs assuming no problems are found! My guess it was to difficult, didn't provide any real benefit and therefore you can leave it out and people can modify there machines with an ***unsupported*** hack. Thanks Luke
Go out and get sailing!
Running on a Beige G3/333 that I've overclocked to:
375Mhz cpu, 75Mhz system bus, 30Mhz PCI. I'm using an ATI 7000 PCI, along with an OrangeLink USB/FW and Apple SCSI card(that drives the boot disk) on PCI. The ATI PCI is driving two displays, so each only has 16MB of video RAM availalbe and I drive them at 1600x1200 each.
The system runs at, but is slightly unstable at the 385/70/35 jumper settings so I'll have to try that to see if the extra 5Mhz of bandwidth on the PCI bus helps at all.
As for performance with the hack: windows sometimes scroll and relocate faster, but I also get random hesitations while dragging windows that I didn't get with QE off. The windows look better when dragged, there's no "tearing" and they seem to float obove the desktop better than with QE off. The genie effect is smoother. The OpenGL screensavers are no faster or smoother than before, not that this should affect them as I understand.
While I'll be the first to say I'd prefer a new G4, I'll also say that this old system is stil quite viable for running OS X on a daily basis. Not for heavy lifting (video clip renderind, 3D modeling/rendering, audio creation, etc). But for surfing, email, coding web apps and the like this is still a nice little box. Beige though it is.
Article X: The powers not delegated... by the Constitution...are reserved...to the people
They suggest you change IOAGPDevice to IOPCIDevice. But to make it work on both just remove it, like:
<key>GLCompositorRequiredClasses</key>
<array>
</array>
Do you see the word "array" there? What part of "array" don't people understand?
<array>
<string>IOAGPDevice</string>
<string>IOPCIDevice</string>
</array>
I mean, how obvious does it have to be? It's not like there aren't 5000 other *.plist files on the system to crib from.
-pmb