Mounting PPC750GXs & PPC750FXs on ZIFs

[ArmorAlley]

Hi all,

 I was wondering if it is a difficult/time-consuming thing to take a G3 processor from a broken iBook, fit it onto a ZIF-mount and use it in a Beige or B&W G3.

The processors are soldered in in iBooks, if I remember correctly, (and I've no idea how difficult it is to desolder these) . Some way would have to be found to set the speed in the G3 if the processor is faster than 500MHz.

Sonnet and friends have shown that it is possible, the question is whether it is worth anyone's while to offer such a service.

Is this really unfeasible?

aa

 

[Trash80toHP_Mini]

The clock multiplier bits seem like the most simple part of such an upgrade.

What's the clock differential between said iBook G3s and those on existing ZIF cards?

Wouldn't it be easier to just upgrade an existing Zif card?

Seems more of a Hacks Forum feasibility study topic than a TP topic at this point.

HINT! HINT! :lol:

 

[ArmorAlley]

Agreed. This topic should be in the Hacks forum. I'll ask one of the admins to be so good as to move it.

 

[trag]

Are these the PPC750FX and PPC750GX based iBooks?  I never followed the iBook line much.

I am all for an effort to salvage as many 750FX and GX chips as possible.

However, there are issues with putting them on ZIF cards...

The easiest and probably cheapest thing would be to replace the CPU  on existing slower ZIFs (266 ZIFs would be a good choice as just about hte slowest thing around).   However, that won't work.  The 750FX and GX have a different pinout than eariler PPC750s.

Soooo, new ZIF cards would need to be created.   Challenging.  Perhaps even daunting.   Certainly possible.   The hard part would be finding the Pin grid array that goes on the bottom of the ZIF card.   I looked for those once, and as far as I can tell, they were never in anyone's public catalog and they're probably long out of production at this point.    Getting a new batch made would mean either high unit prices or a large up front order, or both.   Or  a really great stroke of luck and the locating of the 10,000 units that never got used and have thus far been overlooked by the gold reclaimers.

Desoldering is not particularly difficult.   Reballing and resoldering is more challenging, but some forum members have accomplished it with G4s with larger pin counts.    

Again, if we're talking about the PPC750fx/gx, the bus multiplier is set with software.   There may be pins to set the boot multiplier, but those could be hard set to an "always works" value and then let the user adjust with software settings.

 

[Trash80toHP_Mini]

Are ZIF Sockets with a higher pin count that might be used on a pinout conversion PCB available? I'm thinking DOA PC MoBo reclamation for some reason or other.

edit: heck, are pin grids of some form factor available to gang up on pinout adapter PCB for the CPUs? At this point, who even needs ZIF on an old 266MHz card? Clueless here, I'm living in a .1 pitch world. I'm heading back into my cave now. :

 

[trag]

Just to add a little background information:

Original PPC750 (includes PPC750L):   No internal L2 cache.   Special pins on chip package to interface directly with L2 cache chips.  Speeds to 500 MHz (rare)

PPC740:  Introduced with PPC750.  Essentially a PPC750 with the external cache interface removed.  No front side L2 cache possible.  Meant to be pin compatible with PPC603, so a drop in upgrade, but loses the huge benefit of the fast L2 cache.

PPC750CX:   First G3 with onboard L2 cache.  Onboard 256KB L2 cache.   Pinout change because no need for external L2 interface.  Speeds to 500MHz.

PPC750FX:   512KB onboard L2 cache.    Speeds to 900 MHz.

PPC750GX:  1MB onboard L2 cache.   Speeds to 1100 MHz.

If would be really cool, if when the introduced the CX, FX and GX they had made them pin compatible with the PPC740, as the 740 was described as pin compatible with the PPC603...

The PPC750FX and GX are (IIRC) pin compatible.  I think there's a small change between the CX and FX, but the memory is hazy and I'm not going to go chasing the answer right now.

 

[Gorgonops]

Third-party G3 750GX ZIFs were made at speeds up to 1.1ghz so I'm sure it would be "technically possible" to manufacture new ones using chips removed from dead iBooks but, yeah, as Trag points out:
 

The 750FX and GX have a different pinout than eariler PPC750s.

You can't just take existing ZIFs and swap the slower CPU, you'll have to build a new ZIF carrier.

Given that the PowerPC 7400/7410 (aka, G4) *is* pin compatible with the G3 ZIFs upgrading to that CPU is a more practical proposition, even if it's ultimately less satisfying. (Given the lower clock speed ceiling of around 500mhz.)

 

[Gorgonops]

Agreed. This topic should be in the Hacks forum. I'll ask one of the admins to be so good as to move it.

Moved.

 

[trag]

I took a look at Everymac.  It looks like Apple used the 750FX in the iBooks, but never the 750GX (unless there was a quiet upgrade at some point).   Searching for old stock of chips, there are 9 "affordable" 750FX rated for 600 MHz at about $39 each.   I also found 10 750GX at 1.1GHz, but they're north of $130 each.   But new old stock would certainly be easier to manage than reballed reclaimed chips.   But the numbers of affordable units are very limited.

And we still don't have a source for the PGA.

jt, ultimately, and ZIF card or stack of adapters one might build, needs to end in the proper, IIRC, 19 X 19 pin grid array, because that's the socket that's on the Beige G3 board and a host of related machines and adapters.  

As far as I could tell, when I researched this several years ago (but still too late for the 750GX to still be a live part) Mill Max made the PGAs.   And the pin array is a surface mount, probably something like BGA to pin part.   The pins don't go through the circuit board.  They terminated, probably in tiny heads that are balled and surface soldered to the underside pads of the ZIF card.  

So the ZIF card has BGA parts soldered to both the top and the bottom.

 

[Gorgonops]

I took a look at Everymac.  It looks like Apple used the 750FX in the iBooks, but never the 750GX (unless there was a quiet upgrade at some point).

I'm sure that's right. I just mentioned the GX because it happened to be the chip that ended up in the very fastest G3 upgrades and I couldn't remember offhand if the G3 iBook ever made it over 1Ghz; it did indeed top out at 900mhz.

But new old stock would certainly be easier to manage than reballed reclaimed chips.

How's the wild stock of 7400/7410's holding out? It seems like there could still be a market for converting G3 ZIFs into G4s, albeit a small one? (Of course, I haven't checked what G4 ZIFs go for on eBay.

 

[trag]

I just lucked into a lot of 7 PPC750GX in the 933MHz grade, which I think PowerLogix used on their 1.1GHz cards.    So, if someone finds that pin grid array laying around in a bin at an old electronics store, then I can add it to the list of things that I never seem get any closer to doing.

 

[Trash80toHP_Mini]

Sweet find and I hear you about the project thing! :

It's good this topic was moved to hacks, but a more descriptive title edit might be an even better thing.

 

[ArmorAlley]

I'm not against a change of title if it helps generate interest. How about 'Mounting PPC750GXs & PPC750FXs on ZIFs'?

 

[Gorgonops]

I'm not against a change of title if it helps generate interest. How about 'Mounting PPC750GXs & PPC750FXs on ZIFs'?

Done.

 

[trag]

Supplementing the above summary of the differences in the PPC750 family, including a section on the differences in the pinout of the original 750, the 750CXe and the 750FX:

View attachment 750FX_APP_diff_doc_DD2.X_V1.0pr.pdf

The 750GX is pin compatible with the FX.

 

[Daniël]

You know, PowerLogix already did this with their Pismo G3 upgrades. They took the stock card, soldered on an adapter board, then soldered either a 750FX or GX on top of that (the 750FX, 750GX, 750FL, and 750GL are all pin compatible with eachother), and gave you a smaller heatsink to make it all fit. Here is a blog post of someone showing you how to mod the stock Apple cooler for better thermals, which has some good photos of said adapter board. So yes, it should technically be possible to do this, if one was to make an adapter board. I've heard through the grapevine that someone on this forum is looking at doing that, but I'll let them announce that if that is the case.

 

[Trash80toHP_Mini]

Was the PowerPC 750 with off-die L2 cache on the CPU module in the BeigeG3 an off the shelf CPU config or a custom Apple module?

603 pin compatibility has been driving me nuts for so many years now.  :blink:

 

[Trash80toHP_Mini]

About that pin compatibility:

Found an interesting chart: https://www.okqubit.net/ppchikaku.html Looks like these could be signal, if not pin compatible with the 603e:

Motorola XPC740P

IBM PPC740L PID-8p

Both are in the same 255-pin CBGA package as the 603e on the boards in several PowerMacs and PowerBooks.

Thinking about a multilayer 240-pin CQFP to 255-pin CBGA in order to try reworking some redundant PowerBook 1400 processor cards to G3 400 as a first go round.

The QFP to BGA adapter looks OK in my head. ISTR seeing something stupidly expensive to do this kind of prototyping conversion way back when? I wonder if it's a workable notion for a multilayer SEEED  square or even possible?

Clock multiplier would be handled on the adapter so I don't have to locate the friggin' things on my 2300c mobo. h34r:

 

[Trash80toHP_Mini]

Here we go:

The PowerPC 740 is completely pin compatible with the older 603, allowing upgrades to the PowerBook 1400, 2400, and even a prototype PowerBook 500/G3. The 750 with its L2 cache bus required more pins and thus a different package, a 360-pin ball grid array (BGA).

Wikipedia, but I've seen this info many times in legit sources. []