G4 7448 and 7457 swap

[B S Magnet]

I don't have this L3 cache chips and don't have the small capacitor-resistor-regulator needed (and don't have the schematics), for me is better a PCB-adapter to 744x

I want a pair of 7448!!!!

If anyone should one day wish to donate a working MPC7448 CPU and/or to help with dropping it into a PowerBook G4 DLSD, to aid in the SL-PPC/Clouded Leopard project, I know I would be grateful, humbled, and indebted to any and all of my supporters.

 

[dosdude1]

If anyone should one day wish to donate a working MPC7448 CPU and/or to help with dropping it into a PowerBook G4 DLSD, to aid in the SL-PPC/Clouded Leopard project, I know I would be grateful, humbled, and indebted to any and all of my supporters.

I mean, I still have plenty of 7448s, but they’ve proven to be pretty hit or miss, in the DLSD especially. I suppose I could try it again, but I would like to hopefully find some better chips first... Which may or may not ever happen. This is the best I was able to get it to run:

 

[B S Magnet]

I mean, I still have plenty of 7448s, but they’ve proven to be pretty hit or miss, in the DLSD especially.

I suppose I could try it again, but I would like to hopefully find some better chips first... Which may or may not ever happen. This is the best I was able to get it to run:

Do you attribute this to quality limitations inherent with the available 7448 chips you have on hand (used previously in other applications, I’m guessing, and rated nominally at 1700MHz), or fundamental issues with the A1138/1139 board having issues in reliably delivering voltages greater than the standard 1.3V for 1.67GHz clock speeds? From what I gather, the fastest 7448s Freescale/NXP produced was for a nominal clock speed of 1700MHZ — meaning, anything higher is a manner of overclocking, even if from factory a particular CPU could be certified reliably at up to 2.0GHz.

As for finding stray, used (or new) MC7448 chips out there, it seems the only viable options these days is to pay a legacy systems support company like Rochester Electronics for a tray of 44 or just hope one can find an old, embedded system (which used MC7448 CPUs) being sold as surplus or scrap.

 

[im_to_hyper]

@JoyBed could you upgrade the CPU on any of these cards?

 

[dosdude1]

@JoyBed could you upgrade the CPU on any of these cards?

They more than likely could be upgraded, but you’d need to remove the heatsinks so I can see what chips they currently have. Any card with a 744x can be upgraded to potentially a 7448, any card with a 745x can potentially be upgraded to a 7457. Also any G3 (minus PPC750CXe) can be swapped out with a 7400 or 7410 G4. The only tricky part is figuring out how to set CPU VCORE voltage appropriately.

 

[dosdude1]

Do you attribute this to quality limitations inherent with the available 7448 chips you have on hand (used previously in other applications, I’m guessing, and rated nominally at 1700MHz), or fundamental issues with the A1138/1139 board having issues in reliably delivering voltages greater than the standard 1.3V for 1.67GHz clock speeds? From what I gather, the fastest 7448s Freescale/NXP produced was for a nominal clock speed of 1700MHZ — meaning, anything higher is a manner of overclocking, even if from factory a particular CPU could be certified reliably at up to 2.0GHz.

As for finding stray, used (or new) MC7448 chips out there, it seems the only viable options these days is to pay a legacy systems support company like Rochester Electronics for a tray of 44 or just hope one can find an old, embedded system (which used MC7448 CPUs) being sold as surplus or scrap.

Yes, the chips I have are rated for a 1700MHz clock speed. However the ones I have are labeled “SC7448” instead of “MC7448”. I haven’t been able to find any sort of information about any differences between them, but from what it looks like, there are none, or are very minimal. It seems the main issue in the case of the DLSD is the 7448s I have don’t like the DFS modes (where the machine decreases VCORE voltage and CPU clock speed under low-load conditions to save power) that the DLSD uses. I don’t know if this is down to them being SC7448s (maybe not designed to support DFS?), or if the DFS modes need to somehow be “tuned” differently for a 7448, but when testing, as soon as the machine turned the VCORE voltage down for DFS, that’s when it would become unstable. Running at full speed, it worked fine at 2 GHz.

 

[im_to_hyper]

They more than likely could be upgraded, but you’d need to remove the heatsinks so I can see what chips they currently have. Any card with a 744x can be upgraded to potentially a 7448, any card with a 745x can potentially be upgraded to a 7457. Also any G3 (minus PPC750CXe) can be swapped out with a 7400 or 7410 G4. The only tricky part is figuring out how to set CPU VCORE voltage appropriately.

Thank you! I will have to do that soon then! Do you know if there are mail-in replacement services to do the swap? Or is it pretty much just Joe here?

(I wanted to learn back when I bought these, but now have a disability so cannot really do anything involving fine motor skills or working with equipment, have to pay someone else do the work so I can enjoy the end result.)

Can the 750CXe CPUs be upgraded to faster G3s?

 

[dosdude1]

Thank you! I will have to do that soon then! Do you know if there are mail-in replacement services to do the swap? Or is it pretty much just Joe here?

(I wanted to learn back when I bought these, but now have a disability so cannot really do anything involving fine motor skills or working with equipment, have to pay someone else do the work so I can enjoy the end result.)

Can the 750CXe CPUs be upgraded to faster G3s?

I perform these PPC CPU upgrades as mail-ins quite often, so I can do any such upgrade for you if you'd like to send them in. Unfortunately the PPC750CXe CPUs are of a unique footprint, so only they can be used (unless I were to design some sort of interposer PCB to convert to another footprint).

 

[B S Magnet]

Thank you! I will have to do that soon then! Do you know if there are mail-in replacement services to do the swap? Or is it pretty much just Joe here?

(I wanted to learn back when I bought these, but now have a disability so cannot really do anything involving fine motor skills or working with equipment, have to pay someone else do the work so I can enjoy the end result.)

Can the 750CXe CPUs be upgraded to faster G3s?

Much as @dosdude1 wrote, the 750CX/CXe were a footprint all their own. So if one has, say, a 366/466MHz iBook G3 running the 750CX, it might be possible to source a a 600 or 700MHz 750CXe from a summer 2001 iMac SE board, but this would be the limit of any speed bump within that line of 750s.

 

[B S Magnet]

Yes, the chips I have are rated for a 1700MHz clock speed. However the ones I have are labeled “SC7448” instead of “MC7448”. I haven’t been able to find any sort of information about any differences between them, but from what it looks like, there are none, or are very minimal. It seems the main issue in the case of the DLSD is the 7448s I have don’t like the DFS modes (where the machine decreases VCORE voltage and CPU clock speed under low-load conditions to save power) that the DLSD uses. I don’t know if this is down to them being SC7448s (maybe not designed to support DFS?), or if the DFS modes need to somehow be “tuned” differently for a 7448, but when testing, as soon as the machine turned the VCORE voltage down for DFS, that’s when it would become unstable. Running at full speed, it worked fine at 2 GHz.

Ahhh DFS mode support — I hadn’t considered that angle to be a factor here, but it does seem plausible, given the way your tests showed instability unless running at a full load.

Without knowing definitively the syntax between MC7448 and SC7448 chip part numbers, the nearest fallback, absent Freescale/NXP data sheets, is to hazard that “S” and ”M” might indicate “single” (as in voltage setting, i.e. no DFS support) and “multiple” (i.e., DFS support), respectively.

 

[im_to_hyper]

Much as @dosdude1 wrote, the 750CX/CXe were a footprint all their own. So if one has, say, a 366/466MHz iBook G3 running the 750CX, it might be possible to source a a 600 or 700MHz 750CXe from a summer 2001 iMac SE board, but this would be the limit of any speed bump within that line of 750s.

A 700Mhz iBook would be a super cool endeavor most certainly.

I perform these PPC CPU upgrades as mail-ins quite often, so I can do any such upgrade for you if you'd like to send them in. Unfortunately the PPC750CXe CPUs are of a unique footprint, so only they can be used (unless I were to design some sort of interposer PCB to convert to another footprint).

Ah very cool, is it OK to DM here or message elsewhere to discuss me sending em in? (Or in the marketplace I guess?) or I can message here xD

Regarding interposer design and delivery, what sort of costs are you talking?

 

[dosdude1]

A 700Mhz iBook would be a super cool endeavor most certainly.



Ah very cool, is it OK to DM here or message elsewhere to discuss me sending em in? (Or in the marketplace I guess?) or I can message here xD

Regarding interposer design, what sort of costs are you talking?

Yeah, DM here is fine. Cost for the interposer shouldn't be too bad, the thing is designing such a thing takes quite a bit of time.

 

[im_to_hyper]

Yeah, DM here is fine. Cost for the interposer shouldn't be too bad, the thing is designing such a thing takes quite a bit of time.

Are there others within the community who could work on the design with you? Have you thought about making the design open source on Github, etc for others to collaborate? As the actual work to reball/do the BGA swap is where the "bread and butter" of the income situation will be located?

 

[indibil]

Perhaps the biggest problem with the interposer board for putting a G4 instead of a 750CX/CXe in an iMac G3 is the heatsink. If I'm not mistaken, the chassis plate has a protrusion where the processor rests to dissipate. By placing the interposer board together with the G4 the final height will not be the same and it will be difficult to modify the chassis.

 

[B S Magnet]

Perhaps the biggest problem with the interposer board for putting a G4 instead of a 750CX/CXe in an iMac G3 is the heatsink. If I'm not mistaken, the chassis plate has a protrusion where the processor rests to dissipate. By placing the interposer board together with the G4 the final height will not be the same and it will be difficult to modify the chassis.

That heat sink protrusion (which protrudes through the chassis plate) is also present in the iBook clamshell design which, in OEM configuration, is bridged to the CPU via a foam-based thermal pad (which disintegrates with age and heat).

If this passive cooling method is the same basic makeup as found in the iMac G3 (iFixit doesn’t have any guides on the later iMac G3s, and I’ve never disassembled an iMac G3), it’s a die-cast piece of metal alloy which does not appear to be high in copper, if any is even present. If anything, the heat sink resembles magnesium with maybe some aluminium and possibly zinc.

From iFixit, it’s a T-shaped metal piece. Its CPU contact protrusion, on the reverse side, is not visible:

This metal protrusion, with the right machining tools (were one to have a garage or workshop with them) could be ground to accommodate the new height of an added interposer.

But perhaps better than grinding away existing metal alloy, however, would be to replicate the metal plate’s design in 3D (with Sketchup or whatever 3D design software y’all use), then output that to a small CNC machining setup (not unlike a 3D printer) to mill from block of metal to accommodate the height tolerances required for an interposer. Even if one is not planning to use an interposer, one could still mill a new piece made from user-specified materials — namely, copper, which will help improve thermal conductivity on the passive cooling architecture for this era of iMacs and iBooks. Specifically, the thermal conductivity of copper over, say, a magnesium/aluminium alloy, is probably double (400W/m K, versus around 195W/m K).

On the matter of designing interposers for dropping in a faster CPU (a question addressed to @dosdude1 and/or other folks who know how to work on that level of circuit design), is there an appreciable difference in difficulty/complexity when designing a 750CX/CXe interposer for, say, dropping in a 750FX CPU (like those used with the final iBook G3/800 and 900s), versus a 7400-based CPU, or is the complexity more or less on par?

 

[dosdude1]

That heat sink protrusion (which protrudes through the chassis plate) is also present in the iBook clamshell design which, in OEM configuration, is bridged to the CPU via a foam-based thermal pad (which disintegrates with age and heat).

If this passive cooling method is the same basic makeup as found in the iMac G3 (iFixit doesn’t have any guides on the later iMac G3s, and I’ve never disassembled an iMac G3), it’s a die-cast piece of metal alloy which does not appear to be high in copper, if any is even present. If anything, the heat sink resembles magnesium with maybe some aluminium and possibly zinc.

From iFixit, it’s a T-shaped metal piece. Its CPU contact protrusion, on the reverse side, is not visible:

View attachment 2114847

This metal protrusion, with the right machining tools (were one to have a garage or workshop with them) could be ground to accommodate the new height of an added interposer.

But perhaps better than grinding away existing metal alloy, however, would be to replicate the metal plate’s design in 3D (with Sketchup or whatever 3D design software y’all use), then output that to a small CNC machining setup (not unlike a 3D printer) to mill from block of metal to accommodate the height tolerances required for an interposer. Even if one is not planning to use an interposer, one could still mill a new piece made from user-specified materials — namely, copper, which will help improve thermal conductivity on the passive cooling architecture for this era of iMacs and iBooks. Specifically, the thermal conductivity of copper over, say, a magnesium/aluminium alloy, is probably double (400W/m K, versus around 195W/m K).

On the matter of designing interposers for dropping in a faster CPU (a question addressed to @dosdude1 and/or other folks who know how to work on that level of circuit design), is there an appreciable difference in difficulty/complexity when designing a 750CX/CXe interposer for, say, dropping in a 750FX CPU (like those used with the final iBook G3/800 and 900s), versus a 7400-based CPU, or is the complexity more or less on par?

There is one major thing that adds complexity to a 750CXe interposer... L2 cache. The 750CXe has internal L2 cache, while other 750s (along with 7400s and 7410s) use external L2 cache. As such, you’d lose L2 cache when doing such a swap with an interposer. There is (probably) not enough room to add L2 cache to the interposer and have it fit in the machine. This could potentially cause another issue: The machine/BootROM assumes L2 cache is there, when it now isn’t. On machines with removable CPU cards, the L2 cache SPD data is loaded from an EEPROM on the CPU card itself, whereas it is located in the system BootROM on machines with non-removable CPUs (that have external L2). However, more likely it works more like the 744x series CPUs, where the machine is able to detect (or “know”) how much L2 cache it has just simply based on the type of CPU it is. In such case, designing an interposer to allow installation of a 744x CPU would be much more beneficial than one converting to a standard 750 footprint. That is, of course, assuming a 744x works properly on a 60x bus, which according to the datasheet, even a 7448 should.

 

[indibil]

There is one major thing that adds complexity to a 750CXe interposer... L2 cache. The 750CXe has internal L2 cache, while other 750s (along with 7400s and 7410s) use external L2 cache. As such, you’d lose L2 cache when doing such a swap with an interposer. There is (probably) not enough room to add L2 cache to the interposer and have it fit in the machine. This could potentially cause another issue: The machine/BootROM assumes L2 cache is there, when it now isn’t. On machines with removable CPU cards, the L2 cache SPD data is loaded from an EEPROM on the CPU card itself, whereas it is located in the system BootROM on machines with non-removable CPUs (that have external L2). However, more likely it works more like the 744x series CPUs, where the machine is able to detect (or “know”) how much L2 cache it has just simply based on the type of CPU it is. In such case, designing an interposer to allow installation of a 744x CPU would be much more beneficial than one converting to a standard 750 footprint. That is, of course, assuming a 744x works properly on a 60x bus, which according to the datasheet, even a 7448 should.

Hello.

The 750CXe into an iMac G3 have 100MHz bus, now I'm writing from one . Swap the CPU with a cheap 7447B can be a great upgrade. But the iMac G3 with 750CX/CXe processors don't have aluminum heatsink block between case and CPU, have a "protrusion" into the case and with some heatsink inside.

If the Firmware can boot with a 744x G4, maybe with GigaDesigns patch, the final machine can be awesome!!! i want to try it!!

Something that perhaps would be fine, would be add at the interposer board, the PADS for the processor speed multiplier, to add or remove here the resistors and don't modify the motherboard.

 

[socom_22]

Yes, the chips I have are rated for a 1700MHz clock speed. However the ones I have are labeled “SC7448” instead of “MC7448”. I haven’t been able to find any sort of information about any differences between them, but from what it looks like, there are none, or are very minimal. It seems the main issue in the case of the DLSD is the 7448s I have don’t like the DFS modes (where the machine decreases VCORE voltage and CPU clock speed under low-load conditions to save power) that the DLSD uses. I don’t know if this is down to them being SC7448s (maybe not designed to support DFS?), or if the DFS modes need to somehow be “tuned” differently for a 7448, but when testing, as soon as the machine turned the VCORE voltage down for DFS, that’s when it would become unstable. Running at full speed, it worked fine at 2 GHz.

Hey dosdude,

In case you haven't seen this, SC means "Custom Device" whereas MC means "Qualified Standard Device". Presumably Freescale manufactured the SC chips to be tailored to a specific customer's requirements. What those requirements were, what product they were produced for, and how they might impact our ability to overclock, I have no idea.

It seems most people are having better luck overclocking the MC chips. Sometime in the next couple of months, I will be doing a few chip swaps with some SC7448s and will report whether or not I can get them to run faster than 2 GHz.

 

[sailorMH]

Hi,
I have Powermac G4 FW800 with 2x 7455 / 1.42 GHz.
I want to swap CPUs with 2x 7457. Please, which enabler should I use?

Is Giga Designs Updater 3.0.1. OK, or I need some other?

 

[socom_22]

Yes, the Giga Designs Updater will do the trick. Alternatively, you can use the Sonnet firmware patch intended for the Sonnet Encore/ST and Encore/MDX. It is available on the Macintosh Garden: http://macintoshgarden.org/apps/sonnet-driver-2002

For the Sonnet patch, use version 3.1 for Panther or Tiger booting. Version 3.2 if booting Leopard.

 

[JoyBed]

@JoyBed could you upgrade the CPU on any of these cards?

Sure, just as @dosdude1 said, remove heatsinks to see whats there.

Hi,
I have Powermac G4 FW800 with 2x 7455 / 1.42 GHz.
I want to swap CPUs with 2x 7457. Please, which enabler should I use?

Is Giga Designs Updater 3.0.1. OK, or I need some other?

Yes, I used that one too.

Also, thanks @dosdude1 for answering in this thread. Idk why no notifications arrive in my mailbox.

Regarding the SC parts, and they worked fine and even overclocked very very nicely, better than the original Apple parts I must say.

 

[sailorMH]

Regarding the SC parts, and they worked fine and even overclocked very very nicely, better than the original Apple parts I must say

Please, do you have some exact results of overclocking SC vs. MC versions ?

Because I had two pairs of 7457 CPUs: SC7457-RX1267LC and MC7457-RX1267LC. Originally I have in mind to use MC versions, because I have no experiences with SC.
I have experience with XC versions ( which are inside 1.42 Powermac ), it means "Pilot Production Prototype" and are generally very bad in overclocking.

 

[im_to_hyper]

Sure, just as @dosdude1 said, remove heatsinks to see whats there.

Will do. Been having a disability flare up, hopefully before “too long” I can get into the garage.

I just want these things to run ArchPOWER… have full access to the AUR for compiling every esoteric piece of software.

(ARMCord, modern WebKit browsers, VLC with modern codecs on a PPC Mac, Transmission, QEMU, Box86, Box64 + Steam and Wine to see if Sim City 3000 will run 😂)

 

[socom_22]

Regarding the SC parts, and they worked fine and even overclocked very very nicely, better than the original Apple parts I must say.

Nice to hear that you've had luck with overclocking the SC chips. I previously said I was going to do a few 7448 swaps myself, but I just haven't had the time to mess around with them yet. Hopefully soon...

I have experience with XC versions ( which are inside 1.42 Powermac ), it means "Pilot Production Prototype" and are generally very bad in overclocking.

Sorry, but I have to respectfully disagree with you on this. The "prototype" XC7455B RX1400PF chips are by far the BEST, highest clocking 7455-series chips Motorola ever produced. Generally, if you encounter stability problems when overclocking them, either you need to run them cooler or pump more voltage (or both).

I ran overclocked RX1400PF chips @ dual 1.67GHz for years in an MDD without any issues. In order to achieve 100% stability 24/7 under full load, I have to run it below 50 C at all times and the voltage is set to 1.800V. With 2MB of L3 cache boosted to 278MHz, this screams on Mac OS 9!

 

[B S Magnet]

Nice to hear that you've had luck with overclocking the SC chips. I previously said I was going to do a few 7448 swaps myself, but I just haven't had the time to mess around with them yet. Hopefully soon...


Sorry, but I have to respectfully disagree with you on this. The "prototype" XC7455B RX1400PF chips are by far the BEST, highest clocking 7455-series chips Motorola ever produced. Generally, if you encounter stability problems when overclocking them, either you need to run them cooler or pump more voltage (or both).

I ran overclocked RX1400PF chips @ dual 1.67GHz for years in an MDD without any issues. In order to achieve 100% stability 24/7 under full load, I have to run it below 50 C at all times and the voltage is set to 1.800V. With 2MB of L3 cache boosted to 278MHz, this screams on Mac OS 9!

View attachment 2191503

View attachment 2191504

View attachment 2191502

You should run that setup with Geekbench 2.2.7 to see how it sizes up.