Forums | MacLife

You need firewire

Mr T.,

Impressive job at getting Tiger working on the Rev. D iMac. You should contact Ryan Rempel of Xpostfacto... He is working on getting Xpostfacto to boot Tiger on these old iMacs.  Also, thanks for the idea of using the dual IDE cable... (though I've heard that installing Tiger from an external harddrive is possible... USB HD enclosure here we come...). Actually, apparently, you can trade in your Tiger DVD for a set of Tiger CDs... you have to pay Apple for this service though. This should help old iMac users out....no external drive install needed afterall.

For the record, firewire is not a requirement for Tiger, regardless of what Apple marketing says. Its just easier for them to say firewire is required because any mac with firewire also has a Tiger compatible video card. Its easier to look at a mac and see if it has firewire than to boot it up and study the video card model then compare your model to list of video cards.

This is where Xpostfacto could step in... hacking Tiger to work with most "unsupported" video cards.

I sent an email to Ryan.  And I also got another idea for getting Tiger onto these machines.  Install the iMac's hard drive in a Power Mac, Install Tiger, and plop it back into the iMac.  I'm pretty sure this would work too.  But it doesn't quite cut it coolness-wise.

Mr. T,

It sounds like we have very similar iMacs except I'm on Panther and you are on Tiger (also I only have 96 RAM!). Could you please run XBench and post your results? It's simple...XBench.com hosts the results for you. Im curious if Tiger is faster or slower than Panther on an iMac Rev. D.

When running XBench, its important to not have any other apps running. Also close all uneeded windows. Because the iMac has very limited VRAM, closing windows actually makes a noticable difference in performance.

Here are mine:
http://ladd.dyndns.org/xbench/merge.xht … oc2=109397

Unrelated but interesting:
I've been testing the difference between millions of colors and thousands of colors on the iMac running 10.3.9 and Quicktime 7 pro. Apple must have made some optimizations because the performance is nearly the same. This means that Apple's TIL 106154 is no longer relavent... but thats just in my unproffesional testing. Apparently Tiger has much better software optimizations for graphics, but I'm not sure if they apply to the iMac at all. That's why running XBench for comparison would be interesting.

TIL 106154:
http://docs.info.apple.com/article.html?artnum=106154

Can you move your swap file to the 2GB side?  If there is room anyway.

Oh, there's just one more thing.  For some reason Xbench tells me that my backside bus speed is 134MHz, indicative of a 3:1 cache ratio.  According to everymac.com, the iMac 333 has a 166MHz cache speed, indicative of a 2:1 cache ratio.  Since I'm overclocked to 400MHz, the cache should be running at 200MHz.  Why is it saying 134?

I installed CPU Director in OS 9 (to avoid uninstallation problems in X) and it the default setting was 133MHz.  I increased it to 200MHz and it ran just fine.  Is Xbench wrong?  Is everymac.com wrong?  Is something else wrong?

Mr. T,

Thanks for doing this.

Regarding Tiger performance... it is well documented that Tiger is slow for the first several hours while spotlight indexes your hard drive. The rev. D iMac being a very slow machine, might need to be on a day before the indexing is finished. It might be worthwhile to run the benchmarks again in a while to see if it tests faster. Also, as far as graphics goes, the review at Arstechnica.com states that Quartz 2D Extreme is not enabled in Tiger due to last minute bugs. I can't imagine that this would affect the iMac much, since it has an unsupported video card anyway. 10.4.1 will probably have a very different graphics story, but not neccesarily for the iMac.

One thing that your tests show is that upgrading our iMacs from 96 megs to 320 megs speeds up the benchmark %5 (and in the real world, probably a lot more). In my experience, upgrading an iMac's hard drive also makes a very noticeable performance difference...so there is more work to be done...but so far you've shown that you can take a stock rev D iMac and make it %21 faster by just adding $30 of ram and clock chipping the CPU. As I said, to take it a step further, just about *any* hard drive will be faster than the one that ships with the iMac. So it would be worthwhile to drop one in there (as long as its not a Maxtor! Maxtor drives suck!)

As far as the cache goes... everymac.com is reporting 166 Mhz but lowendmac.com is reporting 133Mhz.  I believe everymac.com is mistaken. XBench reported that my L2 cache speed is 134 Mhz before and after clock chipping my iMac. I'll have to try changing the cache speed on mine to 200 Mhz to see if the 2:1 ratio helps out.

Unfortunately the CPU can't simply tell one drive to copy files to another.  It must do the dirty work itself.  In fact, it will probably go slower since it has to perform both reading and writing over the same USB bus.  The best way to do it is the standard way - or to get one of those overpriced Sonnett upgrades with FireWire.  As for the second half of your question, even USB 1.1 offers at least 10 times the bandwith to playback even the highest bitrate mp3s.  So you're cool.

     Edit:  Oops, I misread that.  IMO, USB (and thus the HD it connects) is too slow even without the iPod, so adding one certainly won't help anything.  So it really boils down to your definition of slow.  Percentage-wise, I'd say you'd get about a 10% reduction in HD performance when playing 128bps tunes from the iPod.

Concerning Tiger's benchmarks, I've stumbled on a startling realization.  The scores (at least for the CPU) fluctuate drastically every time you run the test (in both directions).  I have not yet examined Panther for this phenomenon, but after running the CPU test several times, some of the tests were knocking on Panther's door, while others were far worse.  BTW, the Indexing was finnished in just a few minutes after I installed Tiger, and I double-checked before running Xbench just to make sure.  My guess is that Tiger has a lot more running threads than panther did.  For instance X's event handler probably probes the keyboard buffer for "F-12" (Dashboard) at regular intervals.  Of course it would take a lot more threads than this to cause an aggregate irregularity.  But combine this with a slow CPU, and bingo - we've got a problem.  However, it should be noted that this is only a benchmarking problem.  If Xbench were simply modified to give the CPU more work to do or to perform the testing for a longer period of time, the results would be far more consistant.  On faster machines, this would be less of an issue since the "irregularity" consumes only a fractionally small amount of CPU time.

About the Cache, macspeedzone reports 166MHz (2:1) too.  So that leaves the score tied at 2.  Why the hell can't anyone seem to agree?  My best guess is that Xbench is wrong for two reasons.  First, it mentions that there is a bug where sometimes the cache amount and/or speed is misidentified on some upgrade cards.  Secondly, if the cache speed really were 133MHz, running it at 200MHz would be one hell of an OC.

     Edit: Oh, it just occured to me.  Xbench has to be wrong.  If the ratio is indeed 5:2, Xbench would have to report 160MHz, since my CPU is OC'd to 400MHz.  It does not, therefore it must be wrong.

Last edited by Mr. T (2005-05-11 10:25 pm)

According to wikipedia.org:

Firewire allows peer-to-peer device communication, such as communication between a scanner and a printer, to take place without using system memory or the CPU.

So the capability is there, in practice it will be interesting to see what happens.

All I've got are a few straight-through cables.  I lent out my only cross-over, and havn't seen it since.  Oh well, what are friends for?

RE: FireWire.  I'm not actually all that familiar with the FireWire specification, but wikipedia does mention the use of DMA (and "virtual" DMA) as a buffer for shuffling data among devices.  Assuming such a buffer is used, this would obviously still require access to the computer's memory, and consequently, the USB bus.  More importantly, while it's true that FireWire supports direct communication between devices, it doesn't necessarily mean that every FireWire device implements this - In fact most probably don't.  For this to work, the iPod (the "initiator") would have to be specifically designed to read data directly from an HFS+ formatted hard drive, upon an initial request issued by the operating system.  There would be little motivation for Apple to do this.  So, hypothetically, it might be possible, but given the impractical natue of the problem, I think it's safe to assume it wouldn't work.

RE: USB drive.  USB 1.1 should be plenty fast enough.

RE: Cache.  The cache can only be set by Ratio.  Supported values are 1:1, 3:2, 2:1, 5:2, 3:1.  There is also a maximum speed (in MHz) that the cache can safely operate at.  In my particular case, that speed can be narrowed down to 200 <= x < 267. 

Some websites indicate that a stock Rev D uses a 5:2 ratio (133MHz).  Others indicate a 2:1 ratio (166MHz).  The point I was trying to make is this.  If the iMac does in fact use a 5:2 ratio by default, this means that after I overclock the CPU to 400MHz, the cache speed MUST increase to 160MHz to maintain the 5:2 ratio.  It's not going to drop to 3:1.  The cache ratio IS NOT AFFECTED by overclocking the CPU.  As you said before, Xbench reported the same cache speed before and after the overclock.  This is not possible.  It should've reported 133MHz before the overclock and 160MHz afterwords (again, assuming the default ratio is indeed 5:2).  This proves that Xbench is incorrect.  However, this doesn't disprove the 5:2 theory - it just means Xbench doesn't contribute to it.

Apparently, I couldn't have been more wrong about the cache.  Macbench 5.0 (Classic) and Sonnett Metronome (Classic) both say 133.  Add that to Xbench, CPU director, and Apple's own difficult-to-find "detailed" specsheet ([url/]http://docs.info.apple.com/article.html?artnum=58317[/url]) and it is pretty damn well obvious - it's 133MHz!  200MHz IS one hell of an OC.  Black is white; up is down.  Ahhhhhh...

Alright, I'm back.  The reason I thought the way I thought, is really because of what Nightcougar said about lowering the cache speed to possibly get to 433, and some other articles I've read to that effect.  I just assumed the reason for this was to account for the case where the cache reaches it's limit before the CPU.  I further assumed that this was because the cache speed was increasing proportionally to the CPU.  Guess not.  Welp, ya learn something new every day.

You can move your swap partition in OS X. Not sure if it applies to Tiger.

Here is how:
http://www.macosxhints.com/article.php? … =swap+file