I was just curious , what made those CPUs crack their solder
joints , was it defective CPU that would blow few transistors and
then heat like crazy or just poor manufacturing (soldering) ?

Those cards look kinda thin , no way for heat to escape but it's
amazing you were able to use anything Apple outside Apple CPU
wise.If it's any help I think that older G4 Macs had both Motorola
CPUs (as far I know fastest Motorola ever was 500 Mhz not counting
later Freescale projects) and IBM PPC CPUs and they are not the
same although they tried to be.Sometimes when Apple would run
out of Motorola it would use IBM and extra cache on card (I thought
it was Altivec and became laughing stock at Moobunny for a while).
I think Altivec by IBM was different because Mac OS X 10.4 which
was first (i think) to use it right from desktop worked but showed
some really strange tearing effects and bugs.

Anyway nice hardware project (and I don't like NG that much)

@blizz1220

Haha. AltiVec is the Motorola version. IBM called it VMX.

@Hypex

I guess it did something

Or maybe Apple just introduced some patches or had
specific CD for those motherboards.I saw some CPU
upgrade modules for them that pushed them way high
up to just above 1 Ghz but forgot about it mostly.

Someone who fixed Macs at the time told me that 500 Mhz
Motorola was performing better than anything other PPC
wise at the time (some still want them for old Macs).
After that I just lost track of what they did and didn't
bother messing if it worked.

Anyway , good luck ...

@blizz1220

I have an old iBook with a 500Mhz G3. The most it runs is Mac OSX 10.4.11. AltiVec was introduced for the G4 which you need for 10.5.8. I suppose the later OSX uses it for vector optimisations. Which should speed some things up but it is slower than the previous one on the same hardware.

Compared to OS4 which can use vector optimisations but doesn't force you to need a G4 just to run it.

Last edited by Hypex on 05-Sep-2015 at 04:27 PM.

@Hypex

Actually , someone asked me to see if I can fix his AmigaONE
(XE?) motherboard and the theory (didn't take it yet) is that CPU
is bad and he wants me to resolder it.

I was thinking of just trying to resolder it in place with heat gun
and flux to see if it will help which is why a thread caught my eye.
Later he said it boots sometimes so I told him to look for answers
in forums.

@blizz1220

You would be talented then. Well "sometimes" could be a bad contact on the CPU which can be caused by soldering fractures. Or a loose module. But it could also be battery, ram or HD playing up. Hard to know without doing testing. And hooking up a debug cable.

@Hypex

Thank you for answers.I have some mac CPUs on cards but
nothing that would seem compatible and no rework station either.

I'll give him 2 tested 3 V batteries though if he asks
It was bought as "non-working" anyway.

@olegil

Quote:

olegil wrote: @Geri If someone really wants a 7448, I think this would be the easiest one to put on a new PCB. All we need is the pinout listing of the megarray connector. A 360 ball BGA with 40 NC pins isn't very difficult to use in 2015

Fully agree!

Quote:

Ah. Lower core voltages than what is supported on the XE. Dang. So would need additional regulators. Not difficult, just a hassle. 1.244V and 1.34V are useful for 1.4 and 1.7GHz 7448, the others are simply too high so would blow something.

I see: the CPU voltage regulator on the A1 motherboard is limited to 1.34V.

Would it be possible to simply reduce the voltage using Schottky diodes in series to the CPU core? I know, it is not at all an efficient solution. The question is, how stable is their forward voltage over current consumption and temperature (which will certainly change under load)? Nah, it is probably easier to simply put a second voltage regulator on the module...

@olegil, busytech

Quote:

olegil wrote: @busytech Oh and please map out the traces from BGA to MegArray while you're at it.

For reverse engineering it would also help a lot to note down the resistance of the BGA pads to the MegArry and to measure the resistance to supply voltage and GND pads. This way we could find out, if there are any series/pull-up/pull-down resistors on these lines.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

diodes are not useful as voltage regulators. Luckily, a regulator is not difficult, even for ~20 amps at ~1V. It can even work directly from the VCore of the megarray connector, if this is the easiest solution (other solutions are to regulate from +12V direct from the PSU, just like on modern ATX12Vv2 boards. Most people should have the 4 pin connector for this already

resistances are usually documented in the datasheets, though.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Quote:

olegil wrote: @Geri diodes are not useful as voltage regulators.

Sure, I was just thinking about using them as load independent resistors on the CPU core supply line to generate the ~1V voltage drop (assuming their forward voltage would be more or less constant).

Quote:

Luckily, a regulator is not difficult, even for ~20 amps at ~1V. It can even work directly from the VCore of the megarray connector, if this is the easiest solution (other solutions are to regulate from +12V direct from the PSU, just like on modern ATX12Vv2 boards. Most people should have the 4 pin connector for this already

Indeed!

Quote:

resistances are usually documented in the datasheets, though.

True. But this would help to verify whether the errata I talked about here has already been considered in the design of the current CPU modules or not.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Quote:

Would it be possible to simply reduce the voltage using Schottky diodes in series to the CPU core?

I used to use resistors for my volrage lowering. On my less complcated electronic circuits.

Quote:

For reverse engineering it would also help a lot to note down the resistance of the BGA pads to the MegArry

Without the CPU as we do the science experiment.

see if this website helps, see PDF docs. you can also buy direct.

http://www.linear.com/

_________________
The Machine: Bride Of The Pin•Bot by Williams Electronics

@all

If this project turns out to be viable, would it be possible and worthwhile to upgrade a MicroA1 (750GX) with it? Mine's dead, you know? So, if anyone comes up with a doable product at an affordable price, then (s)he might count me in. Sure.

Here's just wondering and silently hoping...

OldFart

Last edited by OldFart on 10-Sep-2015 at 01:03 PM.

_________________
More then three levels of indigestion and you're scroomed!

Only a 7451, 7455 or 7457 G4 CPU does fit on a AmigaOne XE G4 Cpu Module.
I have a 7457 CPU on my module with 1267GHZ, clocked with 1266MHZ.


Pegasos II G4 CPU Modules have a 7447 G4 CPU.

@Evillord68

My G4 chips have been ordered and are awaiting delivery.

What multiplier dip switch settings did you use to get 1.26 GHz?

@busytech

we covered this on page one. i posted multiplier setting for 1264MHz there (assuming 133MHz FSB). 0111 (plus the extra pin that is always 0).

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@OldFart

Quote:

OldFart wrote: @all If this project turns out to be viable, would it be possible and worthwhile to upgrade a MicroA1 (750GX) with it? Mine's dead, you know? So, if anyone comes up with a doable product at an affordable price, then (s)he might count me in. Sure.

I don't think we can already count this discussion as a project. But I would love to support such a project! billt, olegil?

@olegil

Quote:

olegil wrote: @busytech we covered this on page one. i posted multiplier setting for 1264MHz there (assuming 133MHz FSB). 0111 (plus the extra pin that is always 0).

"0111" is the PLL_CFG register setting, right? If so, the setting has to be inverted for the DIP switch on the CPU module or you consider 0 = ON and 1 = OFF, IIRC.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

Next stage is done. I have acquired some powermac dual G4 modules. They were advertised as 1.25GHz, but one of them has the following CPUs on them.

XC7455A
RX1333PF
RX1250PF

Quote from the datasheet "The X prefix in a Motorola part number designates a “Pilot Production Prototype” as defined by Motorola SOP 3-13. These are from a limited production volume of prototypes manufactured, tested, and Q.A. inspected on a qualified technology to simulate normal production"



I now need to get them transplanted to the A1 module. I have found someone who can do it and will ship them to him soon.

@busytech

Quote:

I now need to get them transplanted to the A1 module. I have found someone who can do it and will ship them to him soon.

Nice! Keep us informed about progression.

OldFart

_________________
More then three levels of indigestion and you're scroomed!

@busytech

so that's a 1.333 and a 1.25. if you can find the missing pll_cfg bit you could try to go beyond 1.25 now. freescale (like motorola) have ridiculously large prototype runs. could be tens of thousand units or even more. if you own more than one 68k or ppc cpu chances are high one will be an xc. i have several.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.