Jump to content

PPC740L G3 CPU Daughterboard For Blackbird Powerbooks


Recommended Posts

The MPC740ARX266LH arrived today, now I only need to find a victim NuPowr 167 (or 183, although I really wouldn't want to risk a 183 if I can at all avoid it) board.

 

I'm a patient man, it will happen eventually. The processor was the harder part, to be honest, and I got exceptionally lucky with that one.

Link to post
Share on other sites
  • Replies 174
  • Created
  • Last Reply

Top Posters In This Topic

According to this datasheet, the 745 is "drop-in compatible" with the 740 and 603ev. I'm not an EE, but I take "drop-in compatible" to mean I can take the 745 and install it in place of the 740 or 603ev without changing anything on the original board. It also suggests that any changes to pin assignments are for non-vital functions and thus ignored by the 745 and host system unless the chip is installed in a specific application that exploits these changes.

This datasheet also specifies the minimum operating frequency of the 745 as 200MHz with a maximum of 350MHz, bus frequencies of 33, 50, 66 and 100MHz, and a maximum 10x multiplier. If a 745 was sourced, our theoretical maximum would be 333MHz on a 33MHz bus. A 40MHz bus may be possible but is technically unsupported, and anything over 350MHz is unattainable anyway.

 

From what I can judge from the pics I would guess that the PLL resistors are those located near the XILINX chip in that picture; they're the only ones with the appearance typical of a bank of selectable components that generally makes up a PLL config.

 

I'd be willing to take a shot at it if there are no other takers, but no guarantees. I have a dead 5500 board I'll use to try to judge the times and temps to use for removal of the original processor and cleaning of the board, which will hopefully translate to the upgrade card and increase my odds of success.

 

I'd figure someone had a damaged upgrade card around somewhere. Those flip-chip packages are extremely fragile and people crack the dies all the time through improper handling or installation; this is the major factor that drove Sonnet to factory-install heatsinks to their upgrades and void the warranty if you removed them.

 

Link to post
Share on other sites

I just got out my PPC-upgraded 540c, and miracle of miracles, it's the 167 MHz model! I think....

 

See for yourself and tell me if it actually is:

IMG_1450.thumb.JPG.3a76b90115dae545f892baa6d9acfbe1.JPG

If I'm reading the writing on the chip correctly, the 3 digits before LC, 166, would represent 166 MHz, yes?

 

I hesitate to donate it to the cause because it works 100% and I don't want it to be damaged, but I will if there's a greater than 50% chance the swap will work, because the prospect of making it into a 300+ MHz G3 is quite compelling.

 

c

Link to post
Share on other sites

Really?

 

Well, I don't think it'll matter here, as I have confirmed unequivocally that what I have is indeed the Newer Technology 167 MHz upgrade, which is precisely what Paralel needs! It's not *too* rare, according to a post somewhere on the last page. Right?

 

Nevertheless, Paralel, it looks like I have the last piece of your puzzle!

 

c

Edited by CC_333
Link to post
Share on other sites

Meh, they're uncommon, but they're not quite at the rarity level of the 183MHz or whatever the top-end one is.

 

On the bright side, even if one of us amateurs botch the processor swap, so long as the card itself is undamaged there's nothing to stop someone from properly performing the upgrade or installing a faster 603ev at a later date; the only guaranteed casualty is the original processor and even then it wouldn't be a problem for someone with a proper BGA reballing kit.

 

So in combination with Paralel's in-hand 740 this would yield a processor card up to 266MHz, which is about 70% faster in clock speed with a doubling of L1 cache over the original 166MHz 603ev.

Link to post
Share on other sites

Shipping is going to be a problem I guess but I can do the swap if needed in case nobody in the US is stepping up to do it.

Have done a lot of CPU swaps already and I am pretty confident that I won’t kill the board or the chip in the process.

Link to post
Share on other sites

Sounds good to me. I can send the chip to Bolle no prob.

 

The only thing that needs to be done besides swapping the processor is to make sure the PLL:CFG 0-3 resistor bank is where we think it is and change it from the current setting to 1-1-0-0, so it will run with the 8x multiplier. Otherwise you'll end up with a G3 that runs at 167 MHz, which is probably still much faster than the 603ev present, but why run at 167 if you can run at 266 MHz?

 

Also the thermal has to be handled before it is ever installed into a working machine and powered up. The current processor @ 167 MHz is putting out 2.1 W thermal for typical use and 3.2 W worst case scenario. The upgraded processor @ 233 puts out 5.7 W thermal for typical use and 7.9 W worst case scenario. I'm thinking thermal tape and a low (possibly super-low) profile heatsink. Possibly securing the top of the heatsink to the rudamentary heatspreader that the Blackbird has to further enhance the thermal transfer. Hopefully that will be enough, Thoughts?

 

The processor I have should work great, its not used or reballed, just NOS, been sitting around somewhere with no one to purchase them.

Edited by Paralel
Link to post
Share on other sites
On 12/26/2018 at 2:56 AM, Franklinstein said:

According to this datasheet, the 745 is "drop-in compatible" with the 740 and 603ev. I'm not an EE, but I take "drop-in compatible" to mean I can take the 745 and install it in place of the 740 or 603ev without changing anything on the original board. It also suggests that any changes to pin assignments are for non-vital functions and thus ignored by the 745 and host system unless the chip is installed in a specific application that exploits these changes.

This datasheet also specifies the minimum operating frequency of the 745 as 200MHz with a maximum of 350MHz, bus frequencies of 33, 50, 66 and 100MHz, and a maximum 10x multiplier. If a 745 was sourced, our theoretical maximum would be 333MHz on a 33MHz bus. A 40MHz bus may be possible but is technically unsupported, and anything over 350MHz is unattainable anyway.

 

From what I can judge from the pics I would guess that the PLL resistors are those located near the XILINX chip in that picture; they're the only ones with the appearance typical of a bank of selectable components that generally makes up a PLL config.

 

I'd be willing to take a shot at it if there are no other takers, but no guarantees. I have a dead 5500 board I'll use to try to judge the times and temps to use for removal of the original processor and cleaning of the board, which will hopefully translate to the upgrade card and increase my odds of success.

 

I'd figure someone had a damaged upgrade card around somewhere. Those flip-chip packages are extremely fragile and people crack the dies all the time through improper handling or installation; this is the major factor that drove Sonnet to factory-install heatsinks to their upgrades and void the warranty if you removed them.

 

 

The datasheet you linked says "The PC745 is a drop-in replacement for the PowerPC 740 microprocessor and is also footprint and user software code compatible with the PowerPC603e microprocessor." The way it is phrased means its a drop-in for the 740, but is only footprint and software code compatible with the 603e, not a drop-in replacement for it.

 

I was correct about there being an issue with B01. It is BVSelect (Board I/O Voltage Selection) an Input, active High, which goes to ground if the I/O bus is 1.8/2.0 V, and 3.3 V if the I/O bus is 3.3 V. On the 603ev B01 is CSE 0, an Output, active High. I'm not exactly sure what a CSE is, but acting as an output on the 603ev and an input on the 740 seems like a problem. Other than that, they are pin-compatible, but Board I/O Voltage Selection is a fairly essential pin.

 

Their core voltages are also not compatible. The 740 & 603ev run at 2.5 V +/- 5%, the core voltage for the 745 is 2.0 V +/- 5%. It is core voltage compatible with the 740L, which was the second revision of the 740, and has a core voltage of 2.0 +/- 5%.

Edited by Paralel
Link to post
Share on other sites

Plus, to make the PC745 compatible with the 740, you'd still need to connect B01, with is NC on the 740, so it can act as BVSelect for the 745, so you'd need to use a jumper wire. So, the PC745 is only "drop-in" if you can add a trace to the board design for the 740 its replacing, or place a jumper wire.

Edited by Paralel
Link to post
Share on other sites

To the membership here as a whole, do the bold sections in this post:

Seem like they could be a problem? I'm not too familiar with some of the abbreviated names for those pins/pads. We need to know if those pads need to be disconnected to act as NC before attempting to graft the G3 to the NuPowr board. We won't really have an option to disconnect those pads after the G3 is in place.

 

I just want to make sure all our ducks are in a row before we take the plunge.

Link to post
Share on other sites

Does anyone know what data bus interface mode (64 or 32 bit) the 603ev operates in when installed in the BlackBird?

 

Does anyone have a BlackBird running a 603ev upgrade module with the proper info or diag utility that can tell us what mode the data bus is operating in?

 

If it's 64-bit, we are good to go. If it's 32-bit, we've hit a potential hard stop, because the 740 cannot operate in anything but a 64-bit data bus mode.

 

It's possible the Pratt chip is still capable of handling a 64-bit data bus interface, but it would be a gamble, as in, we would have no idea until we tried it.

Edited by Paralel
Link to post
Share on other sites
11 minutes ago, Trash80toHP_Mini said:

Damn, looks like it's 32 bit. Found the spec in the PowerBook 550c ServiceSource introduction.

Yep, and it's the same for the Service Source for the rest of the BlackBirds, Series 500 Service Source,. Well, that's a Hard Stop IMO. There is no reason to expect the 603ev to operate in 64-bit mode when the data bus was 32-bit and it could operate in 32-bit mode. The Pratt wouldn't be designed to handle 64-bit data bus translation when the only Apple approved upgrade could operate in 32-bit mode.

 

It looks like those rumors about the guy in Japan that transplanted a G3 were indeed just a rumor, since it is literally not possible.

 

Okay, well, I consider that a wrap. Thank you all for helping with this, your input has been invaluable, but, unfortunately, we have hit an insurmountable problem. I'm glad this was found before CC's daughterboard was sacrificed in vain.

 

Unless someone can provide hard evidence that this isn't a problem, I'll consider this project terminated.

Link to post
Share on other sites

The only hope now is to use a 745, since it can run in 32-bit mode, but the B01 pin issue needs to be rectified, and, of course, the core voltage would have to be adapted to handle a processor with a 2.0 V core. Both issues are way beyond the scope of this thread. So, if anyone wants to pursue adapting the 745 to the BlackBird PowerPC upgrade daughterboard, please start a new thread, since this one should be considered finished, unless someone can think of a way to adapt a processor that expects a 64-bit data bus to a system with a 32-bit data bus

Link to post
Share on other sites

Your project isn't finished, it just got far more complicated! [:)] A new processor card designed around pads for the 745 addresses every problem I can think of offhand. Unlike my theoretical 1400 processor card design, you have the added complication of harvesting Apple ASICs, ROM and misc. components to utilize on your board, but that's not insurmountable. You have but a single interboard connector, where I need to align two of them perfectly as required in a 1400 processor card.

 

BTW, what the blazes is that is that very long, very high pincount white connector on the floor at the front of the processor card cage? Can't figure it out from the block or exploded diagrams. The cubic is occupied by the Modem Card in the exploded diagram, but that's an insanely high pincount for a simple modem interface. Who needs a modem for a PowerBook in this cord cut cellular era anyway? Any clues as to what Apple might have had planned for that crazycool connector would be much appreciated.

 

Good news is that you've got the cubic to use a a 10cm long board with integrated MaxRAM. Without the memory card's board interconnect blocking the way, you can use something like a full size 7cmx10cm copper heat spreader. to keep your cooling budget in line. Added benefit of leaving out the Modem would allow cooling fins to be soldered to the bottom of my proposed copper heat spreader in its place.

 

Such would be a fab subject of another thread for you to explore, even theoretically. [;)]

Edited by Trash80toHP_Mini
Link to post
Share on other sites

Yeah, that honking white connector is for the modem, hard to imagine they need one that size. That sucker is 100 pins. They used 64 of them just for address and data bus. Insane design.

 

Hard to believe there weren't more PDS cards for the BlackBird. It's a typical PDS interface, just like any other Mac.

 

A far as the 745, I'll have to leave that to someone else, once it rises to the level of using an interposer or a whole new card, I have to bow out.

Edited by Paralel
Link to post
Share on other sites
23 minutes ago, Paralel said:

Yeah, that honking white connector is for the modem, hard to imagine they need one that size. That sucker is 100 pins. They used 64 of them just for address and data bus. Insane design.

Is there documentation on the BlackBird Board Interconnect pinout? I need documentation for the 1400 system interconnect as well if someone tries to dig that up. Maybe Inside Macintosh has them?

 

23 minutes ago, Paralel said:

Hard to believe there weren't more PDS cards for the BlackBird. It's a typical PDS interface, just like any other Mac.

PDS Cards were only 16bit as would supposedly be the rest of the I/O bus according to the DevNote block diagram, which appears to be in error.

 

1366669113_BlackBird-LogicBoardBlockDiagram.thumb.JPG.75fcbe63eb89cd684c69fb94489f65a3.JPG

 

Pinout for the PDS has all 32 Address/Data bits, so 16bit system bus may be a typo. But I'm beginning to think it might be a red herring? Again, is there documentation on the System Interconnect pinout anywhere? Is it 16bit or 32 bit? Pincount suggests 16bit to me, so that insanely great Modem Connector (documented!) is beginning to look like a full blown 32bit PDS/Alternate System Interconnect to me because the pinout appears to have all the control lines necessary for a PDS. 

 

I know I'm bat shit crazy, but this might actually not be such a crazy notion at all? :huh:

Link to post
Share on other sites

Here is the pinout for the modem connector in the Developer Note for the BlackBird series:

 

http://mirror.informatimago.com/next/developer.apple.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-68K_Portable/PB_520_520c_540_540c.pdf

 

Here is the Developer Note for the 1400:

 

http://mirror.informatimago.com/next/developer.apple.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-PPC_Portable/PowerBook_1400.pdf

 

I would imagine it has the pinouts for anything you would be interested in

Link to post
Share on other sites
1 minute ago, Paralel said:

That's what I've been working from. I've never seen any Apple documentation on a Processor Card interface. That would likely have been a deliberate roadblock for third party accelerator development. After all, Apple wanted us to by newer/better/faster Macs/Books. Extending the useful life of one we'd already bought with a third party upgrade made/makes no sense at all from Apple's perspective..

 

Just dragged a screwdriver across the System Interconnect bumpage and came up with something like 80 pins. The images in the back of my head suggest that the Modem Connector is almost certainly capable of supporting a full length Processor card in that conveniently long RFI corral. I also see a subterranean 32bit bus between PDS and Modem Connector. Buzzing the three connectors ought to tell that tale, the Modem Connector is almost certainly the place to put a new Processor Card design if my hunches are correct. G4 anyone? :lol:

Link to post
Share on other sites

I noticed that as well when looking at the Developer Note, no processor daughterboard pinout. I figured that was deliberate. Damn Apple!

 

I figured the 3rd party guys probably just grab a system and reverse the pinout themselves using the brute force method.

Link to post
Share on other sites
25 minutes ago, Paralel said:

I figured the 3rd party guys probably just grab a system and reverse the pinout themselves using the brute force method.

Ayup, gotta do just that for the 1400 connectors, what a massive PITA!

 

DRAT! I've got shrink wrapped docs/disk for the  Global Village PowerPort/Mercury in the BlackBird parts bin and no Modem in sight. No BlackBirds, don't really want one, but  .  .  .

 

Best/easiest way to test my theory would probably be to harvest a connector for the Modem Slot from a Modem and one for the System Interconnect from a logic board. Brute force the System Interconnect pinout and design and send Gerbers off for a System Interconnect to Modem Connector adapter and see if a 68040 card fires up in that config.

 

So far as the dearth PDS cards go, that may have been a direct result of the 16bit/32bit disparity. Anything for the PDS hobbled to 16bit would have made it a PCMCIA card candidate?

 

 

edit: just realized ti wouldn't be all that bad to brute force the System Interconnect pinout.  Hold one probe on one of its SMT bumps and drag the other across those of the documented Modem Connector. The missing signals will be the high order data bits and any control signals that don't need to make the trek down to the logic board, but will need to for a full PDS implementation.

Edited by Trash80toHP_Mini
Link to post
Share on other sites

Well, it turns out this project may not be as dead as I thought. It turns out that the memory controller and bus translator IC that NewerTech used for the NuPowr 167 was a later version of the one that is typically used in the BlackBird series and is capable of translating the 64-bit data bus that the processor uses "into single or multiple MC68030 dynamically sized bus cycles. Because the IC seamlessly integrates the two buses, the microprocessor and other bus masters operate as though they were on the same bus." This is from the BlackBird developer note.

 

Based on this information it may still be possible. It's all up to whether CC wants to take that risk or not. I'm still in if everyone else is.

Edited by Paralel
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...

×
×
  • Create New...