PPC740L G3 CPU Daughterboard For Blackbird Powerbooks

[Daniël]

You were going to try it out on a PowerBook 1400 series CPU card that you recently acquired, I think?

I was, until the realization set in that the 133MHz card is in fact not a 603ev, but a 3.3V 603e...
Only the 167MHz is, and thus runs at a lower 2.5V, and that one apparently also needs a matching logic board, which I guess is to do with the power regulation.

Well, at least it'll allow me to sell the PB1400CS on with an included upgrade
I may have found another source for a 745 that doesn't break the bank, so I'm currently pursuing that.

 

[croissantking]

I was, until the realization set in that the 133MHz card is in fact not a 603ev, but a 3.3V 603e...
Only the 167MHz is, and thus runs at a lower 2.5V, and that one apparently also needs a matching logic board, which I guess is to do with the power regulation.

It’s true, the 133 is a 603e not ev. I didn’t know there was a voltage difference. The 166 needing a matching logic board though is to do with the ROM version.

So the next move is to try another 745? Hmm and would it be so bad to try the 740L at 2.42V?

 

[Daniël]

So the next move is to try another 745? Hmm and would it be so bad to try the 740L at 2.42V?

Because 2.2V is the rated max for that chip.
I have actually tried it, but it didn't boot.

No magic smoke, but it's inconclusive to say whether the 740L wouldn't work at all, or if the above max rated voltage simply is too much for it.
Knowing that the buck converter swap also doesn't work, I'm just dropping the idea of trying to get a lower voltage chip working, and sticking to a 745 that is rated for 2.5V max.

It's a matter of eliminating as many variables as possible.

 

[LightBulbFun]

It’s true, the 133 is a 603e not ev. I didn’t know there was a voltage difference. The 166 needing a matching logic board though is to do with the ROM version.

hang on a tick, the 166 needing a matching logicboard due to the *BootROM* is very interesting

do you have any more details on that, ie do you know what happens if you try a 166 card in a 133 logicboard?

this is quite interesting because the 603e and 603ev do have different PVR versions (processor version registers)

and one of the things I was wondering with all of this is the BootROM side, generally older Mac BootROM's are not fussy about the CPU type, so you can stuff any old PowerPC CPU into a PowerMac 7500 be it a 601 604 604e 604ev 750 or various flavours of G4

but I do wonder if in this particular case the BootROM of these PowerBooks is unhappy with the G3 CPU for one reason or another?


(tho there were 1400 G3 upgrades of course, and I think those used a stock BootROM?)

 

[Snial]

hang on a tick, the 166 needing a matching logicboard due to the *BootROM* is very interesting. Do you have any more details on that, ie do you know what happens if you try a 166 card in a 133 logicboard?

I do, because I tried it. I had two PB1400 motherboards. The first one, from a PB1400cs had a 117MHz CPU and the earlier ROM (which I didn't know about at first). That PB1400cs was barely functional, with 16MB of RAM; 750MB HD; a fairly dodgy CS screen; crackly speaker; no CD or FD in the other drive bay. So, I bought a spares/repair PB1400 from a guy in France on eBay. It didn't work. However, it had 16MB of RAM which could be made to work; an active-matrix screen (which I swapped in) and a better speaker (which I eventually put in the first one). I did, however, find that my 117MHz CPU worked in that board. Later I acquired another 16MB of RAM, from eBay, but the seller also sent me (for free: Thank you!!!) a 166MHz CPU.

I found the 166MHz CPU didn't work in my, now fairly functional PB1400c, but it did work in the other PB1400 (now cs). The difference was the ROM. Here's the thread:

Free PowerBook 1400c 166MHz upgrade...

I bought a PowerBook 1400c, 16MB RAM upgrade just before Christmas and it's arrived! The RAM expansion works - I now have 48MB of RAM! Wow, I could run Mac OS 9.1 super slowly! Unexpectedly it also came with a free CPU upgrade, a 166MHz unit. Looks very neat (LHS in images). Of course it...

68kmla.org

From the thread it's worth looking at this post:

Free PowerBook 1400c 166MHz upgrade...

Oh well done @Snial 👏

68kmla.org

To quote: "I have an underclocked 166MHz CPU card @133MHz in my 1400c because I have the older logic board." This gives a definitive answer, if you under clock a 166MHz CPU and put it in an older logic board (with the older ROM) it can work.

 

[LightBulbFun]

To quote: "I have an underclocked 166MHz CPU card @133MHz in my 1400c because I have the older logic board." This gives a definitive answer, if you under clock a 166MHz CPU and put it in an older logic board (with the older ROM) it can work.

interesting! so a 166Mhz 603ev CPU card *will* actually work in a 1400/117 machine but one might have to twiddle with the card to get it to work

so its not that its fundamentally incompatible then, which is what i was wondering about, because again there are much faster G3 upgrade cards for the 1400 that work in all revisions of 1400 AFAIK?

I would love some more details from that from @croissantking ie what does it do if its clocked back up to 166Mhz etc

 

[croissantking]

I would love some more details from that from @croissantking ie what does it do if its clocked back up to 166Mhz etc

Hey, so if you install a 166MHz CPU in a 1400 series logic board with an early ROM, it won’t boot - i.e. it powers on, but doesnt chime or anything else. The same behaviour as if the CPU is not installed.

You can fiddle with the resistors on the 166 card to set it to 133MHz which is what I did and then it does boot and work just fine, and is properly detected as a 603ev.

Later on, I got a donor board with the late ROM and I swapped the chips over. I then set the CPU card back to 166MHz and it worked fine.

The issue is that the earlier ROM does not support the 5x multiplier that the 166 CPU needs, it’s not fundamentally incompatible. I wonder how the G3 upgrade cards would have gotten around the multiplier limitation.

 

[Daniël]

The issue is that the earlier ROM does not support the 5x multiplier that the 166 CPU needs, it’s not fundamentally incompatible. I wonder how the G3 upgrade cards would have gotten around the multiplier limitation.

Is it 5x, or 5x and above?
If it's only the 5x multiplier, then that wouldn't be too much of an issue for 200MHz+ G3s.

Some were also clockdoubled on the system bus, so those could either go under 5x for 200MHz to 300MHz, or over 5x for 366MHz-500MHz, skipping 333MHz (which could be achieved on a 33MHz bus with a 750L and its 10x multiplier, which is remapped onto the 2.5x multiplier of the 750).
Similarly, if one wants to pursue putting a 745 on a PB1400 166MHz card, you could go for a 350MHz model and set the multiplier to 10x to get the max clock speed possible of 333MHz.

System Profiler will likely report a false reading, assuming 2.5x33MHz, but any application like Gauge Pro or Metronome should show the correct 333MHz clockspeed.
However, with the availability of 750 based accelerators with backside cache, I'm not really interested in getting the 166MHz card to do this mod, as it would be in all ways inferior (using L2 cache on the slower, more congested system bus instead), and not doing anything new as those accelerators already provide a G3 for the 1400.

In more positive news, I was given a sample photo of the MPC745BPX350LE part from the supplier, and it does in fact not looked cooked like my others.
Kinda bummed the other ones are all likely fried, though I couldn't have known because of the lack of any reference pictures for these chips at the time.

 

[croissantking]

Is it 5x, or 5x and above?
If it's only the 5x multiplier, then that wouldn't be too much of an issue for 200MHz+ G3s.

Good question, I’m not sure.

In more positive news, I was given a sample photo of the MPC745BPX350LE part from the supplier, and it does in fact not looked cooked like my others.
Kinda bummed the other ones are all likely fried, though I couldn't have known because of the lack of any reference pictures for these chips at the time.

View attachment 80439

Looks good, are you going for it (and can you be sure it’s not remarked)?

 

[Daniël]

Looks good, are you going for it (and can you be sure it’s not remarked)?

I have bought it, yeah.
As for remarking, ironically it's the mask and manufacture date/facility markings that look off, not the part number markings.

In any case, the gold markings on the die carrier package itself match all the 745s I've seen, so even in the case of a remark, it's unlikely to be anything other than a 745, so I'm not too worried about it.

 

[croissantking]

Any project updates?

 

[nottomhanks]

This thread is fascinating. Didn't realize it got updated in 2024. I have the NewerTech 603e 183 MHz card, and am interested in what's going on here.

 

[Daniël]

Hate to break it, but I've officially given up.
I've tried the final 745 a few times, including with 0.65mm balls (as I was having great difficulty getting it reballed with 0.76mm balls), but the behavior remained the same: The PowerBook turns on, but nobody's home.
No chime, no activity beyond the HDD powering up and idling, it just doesn't want to work.

Is the 745 any good?
I right now can't say for certain, because I have no hardware I'm willing to rip the CPU out of to put it in place of, but I want to stop prodding this board before something goes wrong from reflowing over and over, and @CC_333 rightfully wants the card back after all this time.

If I get some hardware to test with, that 745 will get tested, and if it works there, I'm just going to definitively call that this upgrade just isn't possible with the cards and CPUs available.
Sorry folks, I really wish we could have done the crazy upgrade on the Blackbird, but it just wasn't meant to be.

 

[Paralel]

I'm fascinated that nothing ever

and the 740ARX was unlikely to ever work with it.

I'm kinda surprised you came to that conclusion in Oct 2024. You yourself did a transplant in August 2024 of an MPC740ARX266LH into a Performa 6500 that has a 603ev at a slightly higher clock than the NuPower 167 603ev and it was successful:

Coincidentally, I just fitted a Performa 6500 board with a Motorola 740ARX266, still need to tinker with the PLL, but yes, it does work

https://68kmla.org/bb/threads/ppc750gx-vs-ppc750gl.39043/post-541540

That would seem to prove my assertion that the MPC740ARX266LH was a drop-in replacement for the 603ev.

Given that the MPC740ARX266LH failed, and your subsequent attempts with a different CPU also failed, I guess a G3 just isn't meant to happen for the Blackbird unless someone wants to try and make an entirely new CPU daughtercard.

 

[Daniël]

I'm kinda surprised you came to that conclusion in Oct 2024. You yourself did a transplant in August 2024 of an MPC740ARX266LH into a Performa 6500 that has a 603ev at a slightly higher clock than the NuPower 167 603ev and it was successful:

Correct, because the Performa 6500 has a 64-bit wide data path, so the fact the 740ARX does not support a 32-bit wide data path is no problem.

 

[croissantking]

I guess a G3 just isn't meant to happen for the Blackbird unless someone wants to try and make an entirely new CPU daughtercard.

I don’t agree it’s ‘not meant to happen’ - it could just be something marginal that needs adjusting to get it going. I’m of the opinion that someone could get this working if they were dedicated and capable enough. Many feats have been accomplished in the community which were previously deemed impossible, and we’re constantly building on each others’ knowledge.

 

[Paralel]

I don’t agree it’s ‘not meant to happen’ - it could just be something marginal that needs adjusting to get it going. I’m of the opinion that someone could get this working if they were dedicated and capable enough. Many feats have been accomplished in the community which were previously deemed impossible, and we’re constantly building on each others’ knowledge.

True, but given what has been done, especially with the attempts by @Daniël, where could one really go from there with respect to the upgrade card transplant? The system accepts power and then does nothing at all regardless of what has been tried, that isn't much to go on. That's why, to me, it seems like working from first principles and building up to make a new daughterboard would be the way to go.

Plus, making a new daughterboard means that more could be made, using old hardware as the basis would be far less useful given the limited supply of upgrade boards that could be used. All the projects that have succeeded here, for the most part, seem to be built from scratch, possibly using old hardware for reverse engineering as the basis, but in the end they are made using new design and boards.

Hopefully someone will do it with a new daughterboard design someday, more power to them in my opinion.

 

[croissantking]

True, but given what has been done, especially with the attempts by @Daniël, where could one really go from there with respect to the upgrade card transplant? The system accepts power and then does nothing at all regardless of what has been tried, that isn't much to go on. That's why, to me, it seems like working from first principles and building up to make a new daughterboard would be the way to go.

The mods that have been done so far have been superficial in the sense that no attempts (that I know of) have been made to use for example a logic analyser to understand where in the boot process things are hanging up and why. This is not to play down the importance of the work that has been done already but there's plenty more to probe and explore.

It seems like a black box now because there's nothing to go on - but someone will have an 'aha!' moment eventually. It could be something relatively simple.

Plus, making a new daughterboard means that more could be made, using old hardware as the basis would be far less useful given the limited supply of upgrade boards that could be used. All the projects that have succeeded here, for the most part, seem to be built from scratch, possibly using old hardware for reverse engineering as the basis, but in the end they are made using new design and boards.

A new daughterboard design is another good angle to approach this from and something I'm already exploring, although I'd like to produce an 040 card first. Of course I'll make schematics and suchlike available to anyone who wants to build on that for PPC stuff.