G3/500 vs. G4/400 in B&W

[jruschme]

Looking for some advice...

I have a B&W G3. It originally a 400mhz system, but a few weeks ago, I upgraded it to a G3/500 with a cheap ZIF from IC-China. It is currently running Tiger.

Today, on a whim, I bid on and won a cheap G4/350 ZIF on eBay.

Now my conundrum... should I stay with the G3/500 or remove the firmware block and install the G4/350 in my B&W (probably overclocking it to 400mhz)? Obviously, I lose some raw speed if I go with the G4, but it seems like I might be better off in terms of sofware compatibility.

Thoughts?

John

 

[Gorgonops]

I have a G4/400-ed B&W, which was a straight swap over from a 400Mhz G3 module. Since the only thing that changed in the swap was the module I suppose that's as good as experience as any to comment on the difference.

Having the G4 module makes a "quite noticeable" difference in performance on certain mostly-unimportant things. For instance, the shrink/genie effect when minimizing windows to the dock is a *lot* smoother, as is the little icon-scale-on-mouseover effect. Having the G4 also vastly improved the performance of the "scan and scale" photo screensaver and significantly speeds up iTunes visualizations. But on more prosaic things like launching applications having it has positively *zero* effect.

(As for increased program compatibility, well, there's not much that runs on Tiger that *needs* a G4. Having it lets my B&W fire up Garageband, I guess, but it's still too slow to really use it. I wouldn't consider it much of a benefit.)

Was the improvement of those scaling/GUI functions more than a 25% clock speed bump would provide? Probably. Almost certainly. Just beware that if you use non-Altivec-aware programs the machine *will* actually slow down, as for "normal" instructions the G4's of the ZIF vintage don't seem to be *at all* faster than the G3s per-clock. So you might be trading, say, faster word processor page rendering for a faster GUI and iTunes.

Since you've already bought the G4 you might as well experiment and try it. Removing the firmware block won't prevent you from going back to the G3 if it better serves your needs.

 

[Cory5412]

Most of the "legit" Apple G3 modules have 1MB of L2 cache, which depending on the configuration of the /350 or /400 G4 chip, may actually cause the G3 to blow the G4 away in those more important (i.e. non-altivec) tasks such as launching apps and rendering pages in firefox.

So if you can, I'd pull up system profiler with each of the chips and find out what they've got in terms of cache. A healthy enough cache in the G3, coupled with the extra few MHz could cause it to be a much better performer.

EDIT: I should say that most of the "legit" Apple /450 and /500MHz G3 modules have 1MB of L2. I know they produced a lot more speeds than those, though.

 

[jruschme]

I went ahead and downloaded the XLR8 firmware block remover. I assume it also works for non-XLR8 G4's.

The G3/500 is a Buffalo (Melco) HG3-G3500 with 1mb cache running at 250mhz. Anybody know how much cache was on the early Apple G4s?

I'm assuming (hoping) that the G4 will overclock to 400mhz, so that I can use the existing jumper block unchanged.

John

P.S. Do I recall reading that Apple's own G4's will not run with a 66mhz bus? I also have a beige G3 which I could upgrade the leftover processor with.

 

[Gorgonops]

I went ahead and downloaded the XLR8 firmware block remover. I assume it also works for non-XLR8 G4's.
The G3/500 is a Buffalo (Melco) HG3-G3500 with 1mb cache running at 250mhz. Anybody know how much cache was on the early Apple G4s?

The online specs say all "Yikes!" G4s had 1Mb, same as your G3 module. Just now for the heck of it I dragged out my B&W and fired it up to look at System Profiler. (Dang, being turned off for three years can make a system cranky. So much for the PRAM battery...) System Profiler agrees, showing "L2 Cache (per CPU)" as 1MB.

P.S. Do I recall reading that Apple's own G4's will not run with a 66mhz bus? I also have a beige G3 which I could upgrade the leftover processor with.

I'm pretty sure that's bunk. When I was given my G4 module I got a pair of them, a 400Mhz and a 350Mhz, and I passed the 350Mhz on to a Beige G3 owner who I'm pretty sure got it running fine at 366Mhz. (Haven't seen them in years either...) As I recall the thing you had to watch out for with Beige G3s is that some of them shipped with a voltage regulator module that's likely to self-destruct with a G4 module. Google for "Royal VRM G3" for more info.

 

[jruschme]

In that case, the only other variable is the fact that the G3/500 is a copper one, though I don't remember if that introduced any performance enhancements, per se.

Based on what I'm seeing... at the same clock speed, the G4 should be mostly the same speed at non-AltiVec instructions (definately not slower, probably a bit faster on some operations). Since I wasn't unhappy at 400mhz with the stock G3, I think I should actually be fine with the G4.

Hopefully, it will arrive before I leave for the inlaws in CA.

John

 

[jruschme]

(I waited to post this until after I got back from vacation and could do some troubleshooting.)

Just before leaving for CA, I received the G4, installed it in the B&W, and appear to have killed something on the motherboard...

Before receiving the G4, I had installed the XLR8 version of the firmware block remover. The system continued to boot with the 500mhz G3, so I assumed everything was working correctly.

After installing the G4 (with the original 400mhz jumper block in place), I pressed the power button and received the normal chime. After waiting more than the usual, though, I noticed that the screen never turned on. The system powered off as soon as I touched the power button, leaving me to believe that it had hung very early in the boot process.

I replaced the original G3, only to have the same thing happen.

Since then, I have tried a number of steps, including:

• Booting with no PCI cards and only a single DIMM
  Numerous CUDA resets
  Apple's G4 PCI "Motherboard reset" procedure

So far, there has been no difference except for the first boot after the motherboard reset. On this attempt, I received 3 beeps (no usable RAM banks) after a very long wait. I'm dubious, though, that the RAM has failed as it worked fine before.

So far, I've found nothing useful on the web except a few instances of similar behaviour following a CPU and/or video upgrade on a B&W. None have offered a solution, though, other than replacing the motherboard.

At this point, I'm at a bit of a loss about what to do. My two choices seem to be either to try and grab a cheap releacement motherboard (preferably a rev2 to replace the current r1) or fall back to my old beige using either the G4 or the G3 500.

All advice appreciated...

John

 

[Gorgonops]

At this point, I'm at a bit of a loss about what to do. My two choices seem to be either to try and grab a cheap releacement motherboard (preferably a rev2 to replace the current r1) or fall back to my old beige using either the G4 or the G3 500.

I suppose it's just speaking the obvious to point out it'd probably be a bad idea to put that G4 module into your beige machine, unless you're willing to take the risk that it'll murder that unit's motherboard as well. I'd consider it damaged goods unless you can possibly think of some other thing you might of done to damage the B&W's motherboard while installing it.

EDIT/PS: It might be worth pulling the firewire module off the B&W's motherboard. I have vague memories of people suffering "weird" problems on those machines solving them by removing said module.

 

[jruschme]

I suppose it's just speaking the obvious to point out it'd probably be a bad idea to put that G4 module into your beige machine, unless you're willing to take the risk that it'll murder that unit's motherboard as well. I'd consider it damaged goods unless you can possibly think of some other thing you might of done to damage the B&W's motherboard while installing it.

Good point, I had already tried it in the beige, though. It didn't seem break anything.

I can't think of anything I did to mess up the B&W when installing the G4. When I was researching the G4 firmware block, I came across a post (on MacInTouch, I believe) from a purported Apple employee who claimed that the block was put in place because a number of Yosemites in their test lab blew their power supplies when fitted with G4 CPUs. IIRC, most of the subsequent comments discredited this claim, though I wonder if there might not be some subtle difference between the rev1 and rev2 motherboards besides the updated IDE controller.

EDIT/PS: It might be worth pulling the firewire module off the B&W's motherboard. I have vague memories of people suffering "weird" problems on those machines solving them by removing said module.

It didn't help.

Thanks for the response, though. Odd thing... I posted this same message almost verbatim to the LEM G-List. So far, I've received two responses, both of which said to do a CUDA reset. Don't get me wrong, I'm not faulting the advice, but I thought I said that I had tried several CUDA resets.

John

 

[Gorgonops]

Good point, I had already tried it in the beige, though. It didn't seem break anything.

I can't think of anything I did to mess up the B&W when installing the G4. When I was researching the G4 firmware block, I came across a post (on MacInTouch, I believe) from a purported Apple employee who claimed that the block was put in place because a number of Yosemites in their test lab blew their power supplies when fitted with G4 CPUs. IIRC, most of the subsequent comments discredited this claim, though I wonder if there might not be some subtle difference between the rev1 and rev2 motherboards besides the updated IDE controller.

That allegation about "blowing power supplies" sounds pretty bogus. I suppose it's *possible* that some B&W motherboards might of had weak onboard power regulators on the motherboard itself (unlike the Beige the VRM wasn't a swappable module) but... still seems iffy.

It's been a long time since I installed the G4 in my unit, and the only problem I recall having was actually getting the firmware patch to take. (Apparently I didn't hold the programmer switch long enough the first time, or something.) I didn't touch the multiplier jumper blocks, though. There's a thread on the Apple support site:

http://discussions.apple.com/thread.jspa?threadID=2023823&tstart=135

Linking to a Knowledge base article:

http://docs.info.apple.com/article.html?artnum=95157

That suggests that bad/incorrect jumpers can cause your symptoms... did you try rejumpering the system as you had it for the 500Mhz module, or just put the G3 back in with the 400Mhz in place? If you haven't touched that since the swap I suppose it's remotely possible the jumper block could of failed by having one of the little metal bits fall out or something. Seems sort of a far-fetched but that's about all I can think of.

(It's situations like this that make me laugh at an old argument actually made by some olde-tyme-y Mac users back in the day claiming that Macs were "more friendly" than PCs because with Macs you don't see "scary" diagnostic BIOS messages on the screen before the OS boots. Oh, heavens no, let's not have the machine always come up in a least-common-denominator console mode and be able to actually tell you in words what might be wrong with it if something goes wrong. That's "hard". A black screen is so much better.) ;^b

 

[Quadraman]

Saying that a G3/500 would be faster than a G4/400 in non-Altivec apps isn't as straightforward a statement as it seems. The G4 introuduced some new efficiencies to the core that make it slightly faster than a G3 at the same speed even without Altivec. A G4/400 won't run even with a G3/500 but it won't be as slow as you would think going purely by the rated speeds. I still think the G4/400 is a better upgrade option.

 

[jruschme]

Well, I finally got tired of the dead B&W sitting on the floor, so I started checking ebay for a replacement motherboard. I ended up scoring a Rev. 2 B&W motherboard (with 300mhz CPU) for $12 and change with Free Shipping.

After confirming it worked with a minimal config, I decided to try the G4 upgrade again. One change, though, this time I went with the NewerTech enabler (from OWC) rather than the XLR8 one.

So far, so good... it booted from a Tiger install and a reports a PowerPC G4 @ 300mhz. Next step is to fix the jumper block, put back in the rest of the RAM and PCI cards and do a clean Tiger install. (Not sure if I didn't strip the G4 code from parts of the current one.)

At this point, I'm not quite sure what happened the first time. My personal theory is that one of two things caused the problem- either the XLR8 enabler or the Rev. 1 motherboard. My gut feeling is that it's the Rev. 1 motherboard, though... probably some minor component which failed with the load of the G4 (back to the power supply comment).

FWIW, if somebody wants to tinker with the partially dead Rev. 1 board, shoot me an address and a couple of bucks for postage and it's yours.

John

 

[johnklos]

If you're really ambitious, you can get Leopard running on a B&W with a G4. Sure, it'd be slow, but there's still a lot of software out there that works fine on 10.5.

 

[jruschme]

I think Leopard on a B&W G4 is a bit adventurous for my blood.

It seems to be running stable at 400mhz (using the jumper block from the original Rev. 1 board) on a clean install of Tiger. ASU even offered the iTunes 9.1 update. It feels snappier than it used to, but I don't know if that is the G4, the on-board IDE (vs. the ACARD I used with the Rev. 1 board), or the Radeon 7000 PCI.

On a whim, I decided to install Handbrake and see if I could encode a DVD for viewing on my iPod... it should be finished about 17hrs from now. :-/