The 4Ti shows up from time to time. I've had probably 4 or 5 over the years I think. Sometimes they'll pop up in an otherwise non-descript lot-I actually did really well on my first one because I got it with a Radeon 8500 and boxed copies of Call of Duty plus one or two other good Mac games.
It's bar none the best OS 9 card around, and is certainly no slouch in OS X despite not supporting CI. It's my preferred card for dual boot systems, and right now offhand I'm using them both in my dual 1ghz Quicksilver and my crazy DA that's in a rackmount case and has a dual 1.8 Sonnet(I have a few more around in service-that's not a card I leave on the shelf 🙂 ).
The Mac Edition, there again, is easily IDed by ADC+DVI and has the ADC power tab. Most Mac edition cards you see have several inches of "dead" PCB to allow them to engage the support slots in the case. Supposedly, it was originally offered as a separate purchase card without the extension, and documentation I've seen was adamant about not moving the computer with the card installed. The extra PCB allowed Apple to actually ship the cards installed in systems, and IIRC it was a BTO option on the MDD. All of mine are the extra PCB versions-I only recall seeing one "short" card for sale on Ebay years ago.
One other thing-unless you use an MDD OS 9 installer or the OS 9 Lives universal install, the Radeon 9000 can give you headaches in OS 9.2.1 and 9.2.2. With the card installed, it will boot to the finder and give you a seemingly functional desktop that's otherwise "dead"-you can move the cursor but nothing will "click" or work. You have to track down and install the drivers-after removing the 9000 and installing an earlier card, you need to install the drivers and then reinstall the card. It happens with the 9200 and I think 8500 also. Nvidia cards are 100% plug and play in OS 9 in my experience.
Funny enough, this first "bit" me at work a couple of years ago-we have one very expensive, very specialized scientific instrument that still runs on OS 9. The instrument has some quirks and I've done some bizarre things in my time there to keep it running(like the time we found that a 733mhz Quicksilver was actually incompatible with it, and I "fixed" it by fitting a 466mhz card from a DA). In any case, when I first started helping maintain it, the PI/owner/maintainer asked me if I had a video card that would work and had DVI out to run a 1920x1080 screen. I said sure, so came in the next day and popped a 9000 in it. An hour later, he comes to my office LIVID because the computer was seemingly dead. I researched, found the problem, installed the drivers, and all was good. BTW, I keep a hard drive with what I call the "Wittebort Install" in my office(9.2.2+Radeon drivers+NI interface card extensions+the instrument software) and also have an image of it on my office computer so can create a new one quickly if needed.
The system has been rock-stable now(knock on wood) for a couple of years on a Sawtooth. It actually works better on that than it ever did on a DA/QS. Basically, in a nuclear magnetic resonance(NMR) spectrometer, you take a sample that's in a strong magnetic field, and "pulse" it with a broad-spectrum RF signal appropriate for the nuclei you want to observe and the magnetic field strength(i.e. in the range of 500mhz for hydrogren nuclei in an 11.7T magnetic field). This pulse sets the nuclei spinning, and each of them turns into its own little radio transmitter emitting RF depending on the molecular environment(and that decays over time-say a half second to a couple of seconds depending on a few factors). We collect/measure this decay pattern, called a "free induction decay"(FID) and then do a fourier transform on it to get data that we can read/interpret. When it quit working with a fast computer, we could observe pulses being sent on an oscilloscope, and observe the FID on a second oscilloscope, but the computer wasn't picking up the FID. The best we could figure, the software was designed around about a 233mhz CPU(it as running for a while on a beige G3) and was probably written to expect the FID to return within so many clock cycles rather than a specific amount of time. When the clock speed goes to high, the software "times out" on looking for the FID before it physically has time to get back to the computer. In all of this, though, we'd see weird, sporadic errors sometimes with 133mhz FSB computers, but the 100mhz FSB Sawtooth never gives any of those errors. That's all in the fun of old software that ties operation to clock speed 🙂
forums.macrumors.com
