Daniël's PowerPC BGA chip swaps

[croissantking]

The 1400 upgrades all seem to have FPGAs or CPLDs onboard, so I'm not sure if anything's being done there to "make it work", so to speak.
If the 1400 has a 32-bit data bus, then it'd be the only one with G3 upgrades available, I think.

I'm starting to think that the PBX doesn't have a 64-bit data bus option at all.

MacGurus claims that the 1400 has a 32-bit data path, and reading through their descriptions for the various models I get the impression it was done by someone who took care over their research:

PowerBook 1400 Motherboard Layout

Secure online sales of high-quality Macintosh memory products, hard drives, systems, RAID, monitors and other peripherals

web.archive.org

It kind of all fits together. The PBX-based PowerBooks are not fast machines and I guess Apple just wanted to get new models with PowerPC chips out of the door as quickly as possible while keeping costs down. Sadly, it soiled the reputation of the 603e processor which really was capable of more from the outset. Architecturally, I guess these machines have a lot in common with the Performa 5200/5300 and 6200/6300 series.

The only thing in my mind that doesn't add up is how the 1400 series G3 upgrades which used the 750 chip (which allegedly does not have a 32-bit data bus mode) were able to work. As you say, maybe an FPGA or something else. It's really interesting.

 

[croissantking]

EDIT:
If you squint, on that 750 G3 PowerBook 500 card mockup, I think I can just about see a Xilinx logo peek out above the "U13" sticker on the square QFP IC under the cache IC.
So that might also have been using CPLD or FPGA logic

Yep, the NuPowr 500/183 also has that Xilinx chip.

I looked at some photos of CPU upgrades for the PB1400...

We have these two from Interware and Newer, both have 750s and CPLDs.



But then we also have a Sonnet 750L with a FPGA/CPLD, which I wouldn't have expected if the theory of it acting as an interface between 32 and 64-bit data buses is correct (since the 750L has its own 32-bit mode):

 

[Daniël]

But then we also have a Sonnet 750L with a FPGA/CPLD, which I wouldn't have expected if the theory of it acting as an interface between 32 and 64-bit data buses is correct (since the 750L has its own 32-bit mode):

It was most likely designed with the original 750 in mind.
The low-res picture of a green PCB Sonnet accelerator on EveryMac, likely an early preproduction model, shows a Motorola G3, likely the 750:



Some of the later Vimage/Interware cards were fitted with 750Ls as well, despite the chip's 32-bit data bus mode.
Whatever the FPGA/CPLD might be doing to kludge the 64-bit bus chip on a 32-bit bus, would probably have less performance impact than it'd be worth the hassle of adjusting the PCB design over to make it FPGA/CPLD-less for the Lonestar chip.

That, and while by 2000 Motorola was quickly moving away from the G3, it might not be a bad idea to keep the PCB "agnostic" to the data bus issue, should IBM fail to deliver Lonestars and the accelerator manufacturers need to fall back on Motorola or old non-Lonestar IBM 750s.

 

[croissantking]

It was most likely designed with the original 750 in mind.
The low-res picture of a green PCB Sonnet accelerator on EveryMac, likely an early preproduction model, shows a Motorola G3, likely the 750:

View attachment 78207

Some of the later Vimage/Interware cards were fitted with 750Ls as well, despite the chip's 32-bit data bus mode.
Whatever the FPGA/CPLD might be doing to kludge the 64-bit bus chip on a 32-bit bus, would probably have less performance impact than it'd be worth the hassle of adjusting the PCB design over to make it FPGA/CPLD-less for the Lonestar chip.

That, and while by 2000 Motorola was quickly moving away from the G3, it might not be a bad idea to keep the PCB "agnostic" to the data bus issue, should IBM fail to deliver Lonestars and the accelerator manufacturers need to fall back on Motorola or old non-Lonestar IBM 750s.

The theory is sound.

Both NuPowr 500/167 and 183 have Xilinx chips - I wonder what they’re doing.

 

[Nixontheknight]

The theory is sound.

Both NuPowr 500/167 and 183 have Xilinx chips - I wonder what they’re doing.

View attachment 78215View attachment 78217

They're probably FPGAs or CPLDs, so I imagine some LSI voodoo to make the 64-bit bus of the CPU work with the 32-bit bus of the motherboard

 

[croissantking]

They're probably FPGAs or CPLDs, so I imagine some LSI voodoo to make the 64-bit bus of the CPU work with the 32-bit bus of the motherboard

If that is true then indeed the 740 should drop right in. Which throws a curve ball into the theory that a 745 or 740L is needed.

 

[cheesestraws]

I suspect there may be some confusion regarding memory addressing

I think you're right. I think that is referring to the addressing mode the Memory Manager is in, which only has anything to do with the bus size if you're on a 68000.

 

[Daniël]

They're probably FPGAs or CPLDs, so I imagine some LSI voodoo to make the 64-bit bus of the CPU work with the 32-bit bus of the motherboard

That's unlikely to be their purpose on the 603e cards, as the datasheet for the 603e says:

The 603e has a selectable 32- or 64-bit data bus and a 32-bit address bus

I still need to do a deep dive into the differences of the early 603e QFPs, and later 603e BGAs, to figure out why that CPLD/FPGA might be necessary.
As for the 750 mockup, I'm not 100% sure what to make of it yet, as while it looks like it has quite a few chips that make sense, like L2 cache, the seemingly reused Xilinx IC wouldn't be sufficient if it's used for the same purpose and with the same programming.

That, and the Wikipedia article for the PowerBook 500 series, also states the NuPowr G3 would have had a 740 CPU.
It's not properly citated, I admit, but information on this subject is rather scarce to begin with.

I'll keep a watch for any of these cards to become available, so I can put practice to theory.
In the meantime, @finkmac pointed me to something interesting on the Japanese Mercari marketplace:

It's a G3 accelerator card for the Power Macintosh 4400/7220, or Tanzania clone boards.
But crucially, it's missing the actual chip, and the seller thinks it might be a prototype.

In any case, too interesting to pass up, so I've gone ahead and bought it through a proxy.
I might go ahead and see if a 7400 G4 could be fitted in place of a G3, as if it's anything like the Sonnets of the era, it should.

That said, I don't have a Tanzania based Mac (yet), but the L2 cache slot of the Tanzania is identical pinout wise to the Alchemy (5400/6400) and Gazelle (5500/6500/TAM) boards, and uses the same chipset architecture (PSX + O'Hare).
It might not fit in my 6400 chassis, or be a tight squeeze, but if I can set up a test bench type setup with an Alchemy or Gazelle board, I should be able to test it without needing a Tanzania board.

 

[Coloruser]

At least Firmware wise the L2 Slot in Tanzania and Alchemy/Gazelle is different - maybe even some address routing. Something I learned when I and my colleagues developed G3 cards in the late 90s. The Tanzania can directly boot from a CPU in the L2 slot without any additional software (then of course the L2 Cache is disabled). In contrast, Alchemy/Gazelle first start with the onboard 603 and need an init to change the booting process and restart out of the L2 cache slot. So, both need some software to enable the backside L2 cache but Alchemy/Gazelle additionally need an init to switch CPUs.

As a side note, we created a 604 card for the Tanzania based on 603s and included the Tag Ram and L2 Cache so that in principle, a 604 equipped Tanzania 603e showed the same results as a genuine 604 Tanzania with L2 cache Dimm. We even got the 604 Tanzania that was capped at 200Mhz to 300Mhz and faster by this method. But at the end, the upgrade market was dwindling down and we never commercialized these prototypes. Plus, the Tanzania I and II was too limited by its ram capping to be worth such an expensive upgrade.

 

[LightBulbFun]

At least Firmware wise the L2 Slot in Tanzania and Alchemy/Gazelle is different - maybe even some address routing. Something I learned when I and my colleagues developed G3 cards in the late 90s. The Tanzania can directly boot from a CPU in the L2 slot without any additional software (then of course the L2 Cache is disabled). In contrast, Alchemy/Gazelle first start with the onboard 603 and need an init to change the booting process and restart out of the L2 cache slot. So, both need some software to enable the backside L2 cache but Alchemy/Gazelle additionally need an init to switch CPUs.

As a side note, we created a 604 card for the Tanzania based on 603s and included the Tag Ram and L2 Cache so that in principle, a 604 equipped Tanzania 603e showed the same results as a genuine 604 Tanzania with L2 cache Dimm. We even got the 604 Tanzania that was capped at 200Mhz to 300Mhz and faster by this method. But at the end, the upgrade market was dwindling down and we never commercialized these prototypes. Plus, the Tanzania I and II was too limited by its ram capping to be worth such an expensive upgrade.

very interesting to learn that Tanzania machines can boot directly off the CPU in the L2 cache slot without the need for any 3rd party software, thats pretty awesome for anyone who wants to run a more obscure OS on their 4400 or OS X which cant use an L2 cache CPU card otherwise

hypothetically then if the L2 cache CPU cards do indeed work at an OpenFirmware level, then with a G4 card or G3 card thats had a G4 slapped on it, one could run Leopard 10.5.8 on a 4400 if I had a L2 cache card for my own 4400, I would certainly try myself!

 

[Daniël]

At least Firmware wise the L2 Slot in Tanzania and Alchemy/Gazelle is different - maybe even some address routing. Something I learned when I and my colleagues developed G3 cards in the late 90s. The Tanzania can directly boot from a CPU in the L2 slot without any additional software (then of course the L2 Cache is disabled). In contrast, Alchemy/Gazelle first start with the onboard 603 and need an init to change the booting process and restart out of the L2 cache slot. So, both need some software to enable the backside L2 cache but Alchemy/Gazelle additionally need an init to switch CPUs.

As a side note, we created a 604 card for the Tanzania based on 603s and included the Tag Ram and L2 Cache so that in principle, a 604 equipped Tanzania 603e showed the same results as a genuine 604 Tanzania with L2 cache Dimm. We even got the 604 Tanzania that was capped at 200Mhz to 300Mhz and faster by this method. But at the end, the upgrade market was dwindling down and we never commercialized these prototypes. Plus, the Tanzania I and II was too limited by its ram capping to be worth such an expensive upgrade.

Very interesting!
I could imagine Apple might have allowed it on Tanzania, so clone makers could release CPU upgrades for their clones down the line, but locking it down on Alchemy and Gazelle, as those were mainly meant for Apple machines only, where they probably weren't thrilled with third party upgrades (even if there were a couple Alchemy clones, things get a bit foggy with the similarities between Tanzania and Alchemy/Gazelle).

 

[Coloruser]

Very interesting!
I could imagine Apple might have allowed it on Tanzania, so clone makers could release CPU upgrades for their clones down the line, but locking it down on Alchemy and Gazelle, as those were mainly meant for Apple machines only, where they probably weren't thrilled with third party upgrades (even if there were a couple Alchemy clones, things get a bit foggy with the similarities between Tanzania and Alchemy/Gazelle).

Sounds logical. The Supermac C500/600 series went the ZIF way and even incorporated an in-line cache. PowerComputing went the CPU card way. Eliminating the „start 603-switch to G3“ procedure for both Alchemy clones. Interestingly, AFAIK there has never been a Gazelle clone. But the Sonnet/Vimage/NewerTech G3 upgrades still have to go into the PowerBase‘s L2 slot, keeping the original CPU card in place. Why the L2 cards are not compatible with the C500/C600 is another story.

i remember working on a Tatung CHRP Clone board that came with a 604 ZIF Card and we made prototypes of G3 ZIFs for this board as we wanted to make use of that board design. Then his Steveness axed the clones and that mainboard was essentially useless. But I still kept it running until mid 2001 in the lab when the company got bankrupt and then it vanished….

 

[Daniël]

Not PowerPC, but it's still BGA work, so I figure I might as well post it here too:

This weekend, I reballed the PCI Creative Blaster 3DFX Voodoo Banshee a colleague gave me to try and fix.
The card had artifacting issues on cold boots, that would slowly fade as the chip got hot.

Once properly hot, Glide games worked just fine (starting them while the card was still artifacting would result in garbage output and crashing), but to me that indicated poor solder connections under the chip.
With how poorly these chips were cooled with piddly little heatsinks (the PCB had significant discoloration on the area of the PCB under the chip), I guess the PCB has been bent from the large temperature swings, causing cracked solder balls that expand and reconnect due to thermal expansion when the IC gets hot.

I did attempt a simple reflow before this, but that resulted in fully corrupted graphics in the BIOS, so that didn't do the trick.
I also tried a cheap Chinese replacement chip, but once soldered on, it didn't even get hot and had no output, so it's either a fake or already dead.

So, I pulled that chip back off, and practiced reballing it as the chips I've been reballing before were all ceramic BGA chips (PowerPC processors) rather than plastic BGA chips, which are susceptible to popcorning if not handled correctly.
After that went well, I finally went ahead and reballed the original IC, and soldered it back onto the card.

An initial test, after giving the PCI edge connector a good clean, results in a clean picure in the BIOS, free of artifacts.
I've still got to test Glide games, but this is a very good sign already.

 

[Daniël]

Talking of which, I did actually find an AliExpress store claiming to sell those MPC745s, though I failed to take notice of the low I/O voltages.
Means I won't be able to use it if it does show up, but a Plastic BGA G3 chip (instead of the usual Ceramic BGA) will still be a neat thing to have in my little chip collection

Correction, it does support 3.3V.
I was reading a document that said 2.5V I/O wasn't supported anymore, with 1.8V and 2.0V I/O support added, but I failed to see that 3.3V I/O support was still present.

A parcel arrived today, and despite the curious name of the AliExpress seller I got these from, an actual MPC745BPX350LE showed up, though I was surprised this is a dark green, near black PBGA rather than the bright green PBGA in the ad.

Though, then again, that was a different part number (MPC745BVT300LE), but I had no reference online as to what this one should look like.
Very nifty chip, hopefully it can serve a 32-bit data bus only 603e Macintosh at some point in the future

The seller does have more Freescale PowerPC chips, but most of it is the integrated PPC stuff we can't use in Macs (MPC8xxxx series for example).
Other than these, they offer MPC7410s in Freescale Green ceramic, but the Motorola purple equivalents can be found cheaper elsewhere still.

Though the MPC7410VS parts listed in the 745B ad are curious in and of themselves, as they are LGA part numbers, meaning they have gold contacts on the bottoms, not fitted with solder balls.
Another ad lists equivalent MPC7410(T)HX parts, which do have solder balls.

 

[obsolete]

Haha, okay, the name of the AliExpress seller was worth the click-through

Thanks for posting these updates, it's interesting to follow along.

 

[LightBulbFun]

I dont think this is worth creating my own thread over, so Hopefully @Daniël does not mind me slapping it in here, but I am curious if anyone knows just WTF this is? this was something from a few years back, backstory is this was a period G4 Pismo card (as in one of the ones done back in the day when you could send your card in to one of the PowerPC CPU manufactures and they would upgrade it etc) that I was kindly given by someone to use for parts since it sadly became defective, I think crash happy or L2 cache issues? either way the previous owner had tried to bake it, and one of the RAM slots fell off(!) so I ended up with it to use for parts, and yeah when I went to salvage the 7410 from it, this sight greeted me from underneath the 7410!


it held onto the chip somewhat, and had to be actively peeled off the board, it was fairly thick sorta soft? there was no traces on it or the such like IIRC so its not an electrical interposer, but yeah do wonder just WTF was it?

as an aside I do still have this card, I might for giggles try and see if I can get it working by putting a another CPU back on it and see if it still works despite everything

 

[Nixontheknight]

best guess, it's probably a spacer for the chip

 

[croissantking]

Nice addition to your collection and hopefully finds its way into a PowerBook 500 series.

One question, can you be confident it's not remarked?

A parcel arrived today, and despite the curious name of the AliExpress seller I got these from, an actual MPC745BPX350LE showed up, though I was surprised this is a dark green, near black PBGA rather than the bright green PBGA in the ad.

Though, then again, that was a different part number (MPC745BVT300LE), but I had no reference online as to what this one should look like.
Very nifty chip, hopefully it can serve a 32-bit data bus only 603e Macintosh at some point in the future

The seller does have more Freescale PowerPC chips, but most of it is the integrated PPC stuff we can't use in Macs (MPC8xxxx series for example).
Other than these, they offer MPC7410s in Freescale Green ceramic, but the Motorola purple equivalents can be found cheaper elsewhere still.

Though the MPC7410VS parts listed in the 745B ad are curious in and of themselves, as they are LGA part numbers, meaning they have gold contacts on the bottoms, not fitted with solder balls.
Another ad lists equivalent MPC7410(T)HX parts, which do have solder balls.

 

[Daniël]

One question, can you be confident it's not remarked?

I'm pretty much 99% certain it's not.
If you look at the faked MPC/XPC106 chips, they were polished to a clear mirror shine, then re-engraved.

This looks like factory Motorola/Freescale engraving, typeface wise, and it's still got a matte, purple-pink finish to the die.
I would imagine it would be more glossy, if not entirely clear, if they had polished the die for re-engraving, with less legitimate looking engraving.

 

[Daniël]

A parcel arrived today, and despite the curious name of the AliExpress seller I got these from, an actual MPC745BPX350LE showed up, though I was surprised this is a dark green, near black PBGA rather than the bright green PBGA in the ad.

I bought some more, and while I selected the 300MHz part in the thumbnail, as I wanted that variety, I guess the seller too quickly assumed I wanted more of the same as they send me another 350MHz part.
Ah well, can't complain too much given it's the better spec'd part, but if I do later on want the green one, I might order at the "mirror" AliExpress shop to try and see if that at least gets them to pay more attention

500MHz 7410 looks legit as well, so I think I can safely vouch for this Ali shop.



It does seem like it's all reballed with .76mm balls (which the 745s normally come with, but the 7410s do not), and it might be lead free as I find it doesn't melt very well.
I would recommend reballing before using them, which makes the 7410s less appealing as you can get them with leaded balls for $30-ish on eBay still, but I haven't found a source for these 745s elsewhere.