• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

iMac G3 (Rev A -> C) G4 CPU Upgrade

Daniël

Well-known member
So I've finally ordered a hot air station, and I'm slowly getting closer to finally being able to take the CPU board out of my trashed, dead iMac G3 Tangerine (whose logic board one day will be put in a desktop case, and used as a standalone desktop PPC machine, since it's just the CRT that's gone bad), and upgrade it to a 7410 G4 processor I've had lying around for way too long now.

But, that's not all I'm going to try. It's my intention to attempt to take the processor out of a dead Nintendo GameCube, and put it on an iBook G3 logic board. Now, before you call me absolutely nuts, let me explain. The IBM Gekko processor in the GameCube is based on the 750CXe, and looks identical to it. If my findings are correct, the chip should be the same size as the 750CX line, and have the same amount of pins, presumably all mapped the same except for maybe some special pins for the GameCube that the regular 750CX processors do not use. Either way, I'll be attempting to get one or two dead GameCubes to get the CPU out of, and then compare them size-wise to the CX in one of my iMac G3s. If the size is the same, I'll find a dead iBook G3 logic board with the CX(e) processor, and try my luck with that. There's really no point other than to be the first guy to try this, and potentially have the first ever Mac with a GameCube brain transplant  :p

Also, I've found out that the IBM Broadway processor used in the Wii is a 750CL, a processor that wasn't ever used in any PowerPC Mac, but seemingly has the same pinout and size as the 750FX and 750GX chips, both found in either PowerPC Macs, or upgrade cards. Other than a few discrepancies, specifically the lack of dual PLL on the 750CL, it might just work on an iBook G3 logic board with a 750FX processor. In QEMU-KVM, the 750CL processor option actually did allow OS X to boot, but while About This Mac knew it was a 750 processor, CHUD Tools just reported the processor to be a "PowerPC    ". So it's definitely an interesting thought to see what an actual 750CL would do in a Mac.

This would be the closest thing to Mac OS X running on a GameCube and/or Wii!

And yes, I am a mad PowerPC scientist.  :wink:

 

EvieSigma

Young ThinkPad Apprentice
I don't know how difficult this would be but I have a 700MHz iMac G3 motherboard and I wonder if the CPU could somehow be adapted onto a ZIF card to double the clock speed of my B&W...

 

Daniël

Well-known member
I don't know how difficult this would be but I have a 700MHz iMac G3 motherboard and I wonder if the CPU could somehow be adapted onto a ZIF card to double the clock speed of my B&W...
You can't. The 750CX used in the 700MHz iMac G3 isn't in the same BGA package. It's like attempting to put an LGA1156 CPU in an LGA775 socket. Unless you design a ZIF PCB that can take the CX, which would be fairly difficult as no one has reverse engineered the ZIF cards (yet), as far as I know.

 
Last edited by a moderator:

LightBulbFun

Well-known member
I don't know how difficult this would be but I have a 700MHz iMac G3 motherboard and I wonder if the CPU could somehow be adapted onto a ZIF card to double the clock speed of my B&W...


as Dan says the ZIF that comes with G3 BWs is not compatible with a 750CX

but in theory you could design a new PCB that takes 750CXs. The PowerLogix 750FX/GX ZIF cards are surprisingly simply theres just the CPU it self and just a few passives on the card, i dont think it would be THAT hard in the scheme of things to make a ZIF/PCB to take a 750CXe and allow you to plug it into the 300Pin ZIF socket found in G3 Beiges G3 BWs and some IBM Computers... (a pinout of the 300 Pin ZIF socket would help a lot)

I wonder what happened to all the engineers that worked on all these CPU upgrades over the years for the likes of Sonnet Daystar newertech PowerLogix etc... would be really cool to hear from them :)

 

Compgeke

Well-known member
Worthy of note on the IBM computers with a ZIF: They're just a CPU card without any cache or anything. Good chance it's just the 60x bus directly on the pins from the CPU, possibly even in the same pinout as the BGA chip. If someone wants to grab one and desolder the chip, it'd be possible to cross reference the chip datasheet and the BGA pads to pins. A lot of work, but doable. I've attached the datasheet that has the pinout to this post.

Another IBM note: they won't work with cards that have an onboard L2 cache like the Macs have. Trying a 450 MHz G3 CPU (PPC750L-EB0A450) in an RS/6000 43P-150 will throw an E24C "L2 cache array test fails." error. Knowing the pinout would be interesting as it may be possible to make a G3/G4 upgrade for the 43P-150 (and B50). Not that it'd be a worthwhile time investment, I doubt all 3 of us with such a system would be super interested.

View attachment 604eBGA_Datasheet.pdf

 

Daniël

Well-known member
Another IBM note: they won't work with cards that have an onboard L2 cache like the Macs have. Trying a 450 MHz G3 CPU (PPC750L-EB0A450) in an RS/6000 43P-150 will throw an E24C "L2 cache array test fails." error.
That does make me wonder if a PowerLogix G3 ZIF card, like @LightBulbFun has, would work on those IBMs. Like he said, all there's to those cards is either the 750FX or 750GX, and some passives.

 

LightBulbFun

Well-known member
Worthy of note on the IBM computers with a ZIF: They're just a CPU card without any cache or anything. Good chance it's just the 60x bus directly on the pins from the CPU, possibly even in the same pinout as the BGA chip. If someone wants to grab one and desolder the chip, it'd be possible to cross reference the chip datasheet and the BGA pads to pins. A lot of work, but doable. I've attached the datasheet that has the pinout to this post.

Another IBM note: they won't work with cards that have an onboard L2 cache like the Macs have. Trying a 450 MHz G3 CPU (PPC750L-EB0A450) in an RS/6000 43P-150 will throw an E24C "L2 cache array test fails." error. Knowing the pinout would be interesting as it may be possible to make a G3/G4 upgrade for the 43P-150 (and B50). Not that it'd be a worthwhile time investment, I doubt all 3 of us with such a system would be super interested.

View attachment 21452


indeed I have noticed those 300 pin ZIF 604 CPUs I have been meaning to pick one up to play with but I dont fee like paying £40 for a 20 year old CPU :)  but indeed I too would not be surprised if its just the 60x bus in a socket :)  I want to try one of these 604 ZIF CPUs in one of my BW G3s (I know they work in G3 beiges)

I have wondered what a G3 ZIF might do in those IBM Machines, its interesting that the machine can even detect the L2 Cache in the first place that is on the Backside bus. unless does the IBM machine have any motherboard L2 Cache? if so can it be removed? I know in Beige PCI Macs it helps to remove the onboard L2 if your installing a CPU that has its own L2. (im not very familiar with these IBM machines so i dont know how the cache is done on those)

have you tried a slower G3 ZIF (or one with less L2)? how is the multiplier set on these? maybe the Lobo cant handle 450Mhz CPUs/speeds. it would also be interesting if a G4 ZIF works (and as Dan says if a 750FX/GX works...)

 

Compgeke

Well-known member
The real issue with the IBMs is they have their own non-removable motherboard cache and as such don't like having it on the CPU. If I had a G3 ZIF without any cache it just might work.

No idea on how they handle multiplier. They were offered in two different versions, a 233 MHz 604 and the 375 MHz one so they do support adjustable multiplier and bus. Probably resistors on the CPU card itself. Of course, there's almost no documentation on the IBMs. 

 

LightBulbFun

Well-known member
The real issue with the IBMs is they have their own non-removable motherboard cache and as such don't like having it on the CPU. If I had a G3 ZIF without any cache it just might work.

No idea on how they handle multiplier. They were offered in two different versions, a 233 MHz 604 and the 375 MHz one so they do support adjustable multiplier and bus. Probably resistors on the CPU card itself. Of course, there's almost no documentation on the IBMs. 


http://ps-2.kev009.com/rs6000/redbook-cd/sg245144.pdf

found this has some interesting info but nothing on the clock multiplier for the CPU sadly

I do know when those 604e ZIFs are installed in G3 beiges the CPU speed is still set by the motherboard Jumpers, So I dont think the IBM ZIF CPUs have onboard PLL resistors but I cant say for sure. http://www.applefritter.com/?q=node/24331 https://web.archive.org/web/20140108030322/http://www.ppcmla.com/press/?p=61

is it possible to try and disable the L2 cache by removing some resistor or something? like on the 9600 Tsunami boards...

I also wonder if its worth trying one of these G3 ZIFs that had all IBM components. (these used PPC750Ls and weird IBM BGA L2 cache) kinda rare but iv seen them show up on ebay from time to time. I got this photo from the G3 BW service manual as there are none on ebay atm...

image.png

 

Dandu

Well-known member
I have tried a PowerPC 604 from an IBM machine on my G3 BW : it will not boot, with a long beep. 

 

LightBulbFun

Well-known member
I have tried a PowerPC 604 from an IBM machine on my G3 BW : it will not boot, with a long beep. 
I remember reading about that on your blog (yay for google translate :) ) but I cant seem to find the article these days

which CPU did you try I have noticed theres 2 types of 604 ZIFs one older? black type where you cant see the CPU it self, and a newer type which is a PCB with a 255CBGA 604e with a little metal cap slapped on top both seem to be 300 pin ZIFs tho. also did you try dropping the bus speed down to 66Mhz? (I dont know if the 604e supports a 100Mhz bus speed)

 

LightBulbFun

Well-known member
(forgot to mention, the L2 cache in the iBook, according to the L2 Cache chips data-sheet it says it works at both 3.3V and 2.5V, seeing as the L2 cache is working fine with the 7410 im guessing its running at 2.5V in the iBook I guess I could check with a DMM at some point to find out for sure  :) )

 
Last edited by a moderator:

LightBulbFun

Well-known member
so im happy report some good new developments in the world of swapping CPUs around 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png
:)
 
the first one relates to Core voltage of various PPC CPUs, since the start of this whole thing we had been wondering how do we check the Core voltage of various PPC CPUs/Macs when yesterday it hit me, we can just measure across one of the many Bypass/filter capacitors that go between GND and the Core voltage rail, people do this all the time on MacBook logic-board repair to make sure the CPU is working so I thought why not apply it to PowerPC macs, I gave my Pismo a quick probe this way and I was able to record a Core voltage of 2.045V which is exactly what id expect (the PPC750L is said to have a core voltage of 2V)
 
the other one is to do with, Motorola G3s the MPC750 and XPC750 CPUs, it was thought that all the Motorola parts all had a core voltage that was too high to work with the 7400/7410 CPUs but I did a bunch of internet sleuthing and as it turns out that is only true for the earliest of G3s, but the later ones, actually have a Core voltage that is compatible with 7400-7410 CPUs 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png
:)  the way to tell which 750 is which. is
 
the 2.6V XPC/MPC750s have say on the second line on the chip "ARX" followed by clock speed. the 1.9-2.05V G4 compatible MPC/XPC750 chips say "PRX" followed by the clock speed
 
here is a picture of a "PRX" G3 from a Rev A 400Mhz G3 BW (I measured the core voltage of this one and got 2.045V as well 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png
:)  )
 
2549e5f4d9831dd48060c7d31c4e9e9c.png.2a83f98d5bf6e0c850af3866e54b305f.png

 
this opens up a few more Macs/CPU cards to being G4 upgraded now 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png
:)
 
but the issue still remained of what about the Bus voltage and L2 cache voltages, on the PPC750L 7400 and 7410 the L2 voltage is configured by a pin that is pulled to various states, where as the XPC/MPC750 only supports 3.3V, so these voltage configurations pins are just unused pins on the XPC/MPC750, luckily for us on the 7400-7410 these pads have an internal pull-up which configures the CPU to use the highest Voltage its compatible with, which on the 7400 is 3.3V for both the bus voltage and L2 cache voltage which is the same as what the 750 supports their-for solving those issues 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png
(this explains how the L2 cache continued to function on the Sonnet nuBus G3 card @Bolle stuck a 7400 onto :) )
 
(the Rev 11.3 7410s found in Apple machines are only compatible with 1.8/2.5V bus and L2 cache voltages, but the Rev 11.4 7410s like the ones on ebay are compatible with 3.3V bus voltage but only do 2.5V L2 cache still, so if your planning to G4 swap an MPC/XPC750 machine/CPU card I recommend using a 7400 and not a 7410)
 
if you can adjust core voltage then you XPC/MPC750A based boards should work too with a G4 7400 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png

 
for those looking for the Juicy technical stuff and my references ill link bellow some data sheets and PDFs that helped me figure all this out 
clear.png.8a4e5da0110d13511a537fbd13c057cc.png

 
https://www.nxp.com/docs/en/application-note/AN1812.pdf
 
https://www.nxp.com/docs/en/data-sheet/MPC750EC.pdf
 
https://www.nxp.com/docs/en/data-sheet/MPC750PEC.pdf
 
https://www.nxp.com/docs/en/data-sheet/MPC7400EC.pdf
 
https://www.nxp.com/docs/en/data-sheet/MPC7410EC.pdf
(this is a copy paste more or less from a Post I just made on MR as well so sorry if theres any formatting bugs or the like :p )

 
Last edited by a moderator:

Daniël

Well-known member
@dosdude1 is at it again, this time upgrading a Trayloader iMac G3. He's figured out how to change the core voltage on the card, which is necessary when the G3 chip on the card is a ARX Motorola 750, which run at higher voltages that will damage the Motorola 7400 G4 CPU (see the previous post for more info). @LightBulbFun and I have been trying to figure out what sets the Vcore on these cards, so I am very glad @dosdude1 and him figured out that a set of specific resistors does this. 

 
Top