Jump to content
Paralel

PPC740L G3 CPU Daughterboard For Blackbird Powerbooks

Recommended Posts

2 hours ago, kev7112001 said:

its a stupid design that could not scale hence why all others faster are flip chip

you can not cool such a chip properly that was just a cheap package all ibm power chips beside that are flip chip as that is the only way to cool any cpu thats is worth using

the reason why they chose that was to be cheap and its lower power so it was ok to do

its doesnt make it better in anyway like you say it actually is worse due to you sending all that heat into the PCB and surrounding components

if you look thats why there is a heatspreader on top of that version to help with its huge disadvantage

flip chip is way more expensive to produce so if they dont have to in this case a 750cx/cxe they wont

under a flip chip die are more tiny solder balls no bonding wires ilke that garbage package

many other things i can list as to why that package is not used simple google search will tell you why besides what i have said

Regardless of all of the supposedly amazing things you're doing with grafting high-end chips into Amiga desktops, the purpose of this thread is to install a new processor in a vintage laptop that has virtually no ventilation and only the most meager of heat sinks. Thus, if you want to use the thing on battery for more than 30 minutes and/or not have it shut down from thermal overload in about the same amount of time, a low-power cool-running chip such as the CX or CXe, with a maximum power draw of 6.7W@366MHz (and less at the maximum 333MHz it would run in a PowerBook 5x0), is something to be desired, not scorned. I mean if you want to build your own 7447-based PowerBook 5x0 upgrade (which will draw up to 23W@1GHz and top out at just 924MHz (28x33MHz)), then please do, but the rest of us really aren't ready to take it to that level just yet, seeing as how they're still trying to arrange to do the original 603e-to-740 processor swap being discussed here.

 

As far as the 750CX/CXe's package goes, yes it's wire-bonded and that's not the fastest technology ever, but it has a max of 700MHz and since its multiplier tops out at 10x (just like its contemporaries until the 750FX and 7450), the fastest it could possibly go in a 5x0-series machine is 333MHz, so it's kind of irrelevant. Also, yes, it has a heat spreader integrated to the top of the otherwise plastic BGA package. How is this a problem? It provides a large surface area directly bonded to the die, able to safely withstand a 2.2kg static load, and utilized to transfer heat to a proper thermal solution. Also the encapsulant on the underside of the package is designed to insulate the board from the die's heat, not transfer heat to it (the encapsulant is also a minimum of .244 mm away from the system board so there's no direct heat conduction). It doesn't transfer any more heat to the system board than would any comparable chip. In fact, I'd wager that it conducts less because normal FCBGAs have balls directly beneath the die. 

 

Anyway it's just a recommendation for one of the Plan B variants: 750CX on an interposer on the original BGA-style upgrade card. The low ball count and low package profile of the 750CX make it easy to work with and keeps the total installed height within original specs so you don't have to fab custom heat sinks or anything, and keeping the original upgrade card greatly reduces cost and effort compared to designing and building a whole new card. You'd still have to reduce voltages though since the CX only runs a 2.5V I/O instead of the 3.3V of the 603e, but this is a requirement in order to use pretty much any chip outside of the originally desired 740.

 

Share this post


Link to post
Share on other sites

Side topic related: since you're deeply into the specs for CPU alternatives here, how might a 750FX on a new card work out? 1400c board (CPU only) Accelerator experimentation leading up to a CQFP to BGA interposer for Duo 2300c et al advice requested. I'm figuring the 2300c board with PBX would need to be re-created for plan B1 PCB to proceed. If not a duplicate that portion of the 2300c very closely resembles the existing Blackbird PPC upgrades. Reverse engineering a 2300c board's splayed out implementation should be fare easier and risk free compared to tearing down a rare Blackbird upgrade card. Baby steps.

Edited by Trash80toHP_Mini

Share this post


Link to post
Share on other sites
6 hours ago, Franklinstein said:

Regardless of all of the supposedly amazing things you're doing with grafting high-end chips into Amiga desktops, the purpose of this thread is to install a new processor in a vintage laptop that has virtually no ventilation and only the most meager of heat sinks. Thus, if you want to use the thing on battery for more than 30 minutes and/or not have it shut down from thermal overload in about the same amount of time, a low-power cool-running chip such as the CX or CXe, with a maximum power draw of 6.7W@366MHz (and less at the maximum 333MHz it would run in a PowerBook 5x0), is something to be desired, not scorned. I mean if you want to build your own 7447-based PowerBook 5x0 upgrade (which will draw up to 23W@1GHz and top out at just 924MHz (28x33MHz)), then please do, but the rest of us really aren't ready to take it to that level just yet, seeing as how they're still trying to arrange to do the original 603e-to-740 processor swap being discussed here.

 

As far as the 750CX/CXe's package goes, yes it's wire-bonded and that's not the fastest technology ever, but it has a max of 700MHz and since its multiplier tops out at 10x (just like its contemporaries until the 750FX and 7450), the fastest it could possibly go in a 5x0-series machine is 333MHz, so it's kind of irrelevant. Also, yes, it has a heat spreader integrated to the top of the otherwise plastic BGA package. How is this a problem? It provides a large surface area directly bonded to the die, able to safely withstand a 2.2kg static load, and utilized to transfer heat to a proper thermal solution. Also the encapsulant on the underside of the package is designed to insulate the board from the die's heat, not transfer heat to it (the encapsulant is also a minimum of .244 mm away from the system board so there's no direct heat conduction). It doesn't transfer any more heat to the system board than would any comparable chip. In fact, I'd wager that it conducts less because normal FCBGAs have balls directly beneath the die. 

 

Anyway it's just a recommendation for one of the Plan B variants: 750CX on an interposer on the original BGA-style upgrade card. The low ball count and low package profile of the 750CX make it easy to work with and keeps the total installed height within original specs so you don't have to fab custom heat sinks or anything, and keeping the original upgrade card greatly reduces cost and effort compared to designing and building a whole new card. You'd still have to reduce voltages though since the CX only runs a 2.5V I/O instead of the 3.3V of the 603e, but this is a requirement in order to use pretty much any chip outside of the originally desired 740.

 

if you plan on making a interposer then you should be looking at 750FX/GX only im adding 1 now to a blizzard ppc which i have way less space than this apple part

this is the last guy who tried http://leblogdecosmos.blogspot.com/2018/08/blizzardppc-740.html

just fyi there are a few IBM IOP cards all on ebay you can steal 750FX/740 from cheap

and the glob top on that chip does not contain heat and it DOES shoot all that heat too PCB which can warp it over time this is fact

and will pop BGA pads

 

Edited by kev7112001

Share this post


Link to post
Share on other sites
14 hours ago, Franklinstein said:

I would have to assume that the previous TREX controller doesn't support it, or at least the '030-era video controllers in these older 'Books didn't have any provisions for it.

I'm sure it's a combination of both but the video controller would definitely be a blocker all by itself. By contrast, the 2400/3400 used the dirt-common-in-PCs C&T 65550 VGA chip instead of a proprietary Apple ASIC, and the datasheet for that chip makes a big deal about the presence of its "Zoom Video Port". (And this chip came in both PCI and VESA local bus flavors, which is why 486s with ZV are indeed a thing.)

Thinking about it now I'm kind of snickering at the idea that maybe instead of continuing to reuse all that '030-bus garbage they had lying around the lab perhaps Apple would have been better off slapping together a 603-to-486/VESA bus transceiver and going all-in on using off-the-shelf PC components even before the PCI era. I mean, sure, given the small window where this would have been relevant it would have ultimately been a total waste of time but given how the Mac software architecture works it would have at least been theoretically feasible. LC PDS video cards based around VLB VGA controllers do actually exist. ;)

 

Anyway, yeah, again, sorry for the derail.

Share this post


Link to post
Share on other sites
7 hours ago, Trash80toHP_Mini said:

Side topic related: since you're deeply into the specs for CPU alternatives here, how might a 750FX on a new card work out? 1400c board (CPU only) Accelerator experimentation leading up to a CQFP to BGA interposer for Duo 2300c et al advice requested. I'm figuring the 2300c board with PBX would need to be re-created for plan B1 PCB to proceed. If not a duplicate that portion of the 2300c very closely resembles the existing Blackbird PPC upgrades. Reverse engineering a 2300c board's splayed out implementation should be fare easier and risk free compared to tearing down a rare Blackbird upgrade card. Baby steps.

That's an interesting idea. I must admit, I am not terribly familiar with the 2300's setup. The 750FX was indeed what I was thinking for B1, it would be the least difficult of the most advanced chips to implement. 

 

13 hours ago, Franklinstein said:

Regardless of all of the supposedly amazing things you're doing with grafting high-end chips into Amiga desktops, the purpose of this thread is to install a new processor in a vintage laptop that has virtually no ventilation and only the most meager of heat sinks. Thus, if you want to use the thing on battery for more than 30 minutes and/or not have it shut down from thermal overload in about the same amount of time, a low-power cool-running chip such as the CX or CXe, with a maximum power draw of 6.7W@366MHz (and less at the maximum 333MHz it would run in a PowerBook 5x0), is something to be desired, not scorned. I mean if you want to build your own 7447-based PowerBook 5x0 upgrade (which will draw up to 23W@1GHz and top out at just 924MHz (28x33MHz)), then please do, but the rest of us really aren't ready to take it to that level just yet, seeing as how they're still trying to arrange to do the original 603e-to-740 processor swap being discussed here.

 

As far as the 750CX/CXe's package goes, yes it's wire-bonded and that's not the fastest technology ever, but it has a max of 700MHz and since its multiplier tops out at 10x (just like its contemporaries until the 750FX and 7450), the fastest it could possibly go in a 5x0-series machine is 333MHz, so it's kind of irrelevant. Also, yes, it has a heat spreader integrated to the top of the otherwise plastic BGA package. How is this a problem? It provides a large surface area directly bonded to the die, able to safely withstand a 2.2kg static load, and utilized to transfer heat to a proper thermal solution. Also the encapsulant on the underside of the package is designed to insulate the board from the die's heat, not transfer heat to it (the encapsulant is also a minimum of .244 mm away from the system board so there's no direct heat conduction). It doesn't transfer any more heat to the system board than would any comparable chip. In fact, I'd wager that it conducts less because normal FCBGAs have balls directly beneath the die. 

 

Anyway it's just a recommendation for one of the Plan B variants: 750CX on an interposer on the original BGA-style upgrade card. The low ball count and low package profile of the 750CX make it easy to work with and keeps the total installed height within original specs so you don't have to fab custom heat sinks or anything, and keeping the original upgrade card greatly reduces cost and effort compared to designing and building a whole new card. You'd still have to reduce voltages though since the CX only runs a 2.5V I/O instead of the 3.3V of the 603e, but this is a requirement in order to use pretty much any chip outside of the originally desired 740.

 

I agree, you have some strong points here in favor of the CX/CXe. The thermal budget is one of the biggest challenges we have and >6.7W is within the realm of reasonable, with some modification we are looking at for the thermal management system.

Edited by Paralel

Share this post


Link to post
Share on other sites
2 hours ago, Paralel said:

That's an interesting idea. I must admit, I am not terribly familiar with the 2300's setup. The 750FX was indeed what I was thinking for B1, it would be the least difficult of the most advanced chips to implement. 

 

I agree, you have some strong points here in favor of the CX/CXe. The thermal budget is one of the biggest challenges we have and >6.7W is within the realm of reasonable, with some modification we are looking at for the thermal management system.

you have tons and tons of cooling options today if heat is that serious for you they sell heatpipes copper sinks etc etc etc and obviously in your case nothing but copper based stuff you should be looking at

another version my team uses on amiga is a PCI carrier to PMC powerpc boards

you have next to no space for a heatsink since it is sandwiched

what a decked out 1GHz 750FX amiga 4000 looks like

second card down from top is the PCI carrier with PMC PPC

also view of what the cards look like https://4.bp.blogspot.com/-Sv9hh1mNNH8/WxxcLVuUTzI/AAAAAAAAPYU/_MGtgDu8nmY6tQ_U9zcwHlItWsT3S0RwgCLcBGAs/s1600/20180610_004431.png

IMG_7782.jpg

Edited by kev7112001

Share this post


Link to post
Share on other sites

Share this post


Link to post
Share on other sites
4 hours ago, Paralel said:

That's an interesting idea. I must admit, I am not terribly familiar with the 2300's setup. The 750FX was indeed what I was thinking for B1, it would be the least difficult of the most advanced chips to implement.  

750FX seems the obvious choice for my 1400 project, 20x gets it to 666MHz with 512k  of on die cache vs. my Crescendo PB at 466MHz with 1MB on the backside.

 

Here's your Duo 2300c Block Diagram, it's exactly what's on your Blackbird PPC Upgrade target as I understand it.

 

2300c-FSB-Block_Diagram.thumb.JPG.14b5fe0d92204a5bc53fec3ae48fea11.JPG

2300c

 

The block diagrams of 2300c and 1400 are all but indistinguishable.

 

1400-FSB-Block_Diagram.JPG.69e3a45232bbabd7349840200325b03b.JPG

1400

 

edit: at this level (FSB?) the 2300c and 1400 are identical but for the CPU board interconnect.*** The Delta DevNote for the Enhanced 1400 (166MHz/Cache) doesn't provide a block diagram update for the CPU card unless it's on one of the pages that appear to be blank.

 

*** Different Max RAM provisions are mere details related to intentional hobbling of the 2300c/PPC upgrade path that Apple had promised, but never really wanted to ship for the Duo System. With IDE AND SCSI buses on board for HDD, Color AND Grayscale connectors for LCDs on board makes the 2300c the Swiss Army Knife of the PPC menagerie. Plans were to hang T-REX/PCMCIA Card Cage of the 1400 off the Duo's Docking Connector. That's why there is so much vacant cubic underneath the PCB in the UltraDock series. Getting that setup running has been a sporadic journey of fifteen years.

 

Edited by Trash80toHP_Mini

Share this post


Link to post
Share on other sites
2 hours ago, kev7112001 said:

Those are exactly the type of heat sinks I was considering. Looks like he FX's heat dissipation could be manageable.

 

1 hour ago, Trash80toHP_Mini said:

750FX seems the obvious choice for my 1400 project, 20x gets it to 666MHz with 512k  of on die cache vs. my Crescendo PB at 466MHz with 1MB on the backside.

 

Here's your Duo 2300c Block Diagram, it's exactly what's on your Blackbird PPC Upgrade target as I understand it.

 

edit: at this level (FSB?) the 2300c and 1400 are identical but for the CPU board interconnect.*** The Delta DevNote for the Enhanced 1400 (166MHz/Cache) doesn't provide a block diagram update for the CPU card unless it's on one of the pages that appear to be blank.

 

*** Different Max RAM provisions are mere details related to intentional hobbling of the 2300c/PPC upgrade path that Apple had promised, but never really wanted to ship for the Duo System. With IDE AND SCSI buses on board for HDD, Color AND Grayscale connectors for LCDs on board makes the 2300c the Swiss Army Knife of the PPC menagerie. Plans were to hang T-REX/PCMCIA Card Cage of the 1400 off the Duo's Docking Connector. That's why there is so much vacant cubic underneath the PCB in the UltraDock series. Getting that setup running has been a sporadic journey of fifteen years.

 

 

I had no idea the 2300 was so similar in architecture. I can completely see what you are talking about. It appears that our two projects do certainly intersect. Anything you learn can be applied to future plans for the BlackBird project, and vice-versa.

Edited by Paralel

Share this post


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

Those are exactly the type of heat sinks I was considering. Looks like he FX's heat dissipation would be reasonable.

750FX/GX are the fastest coolest per clock of all G3 due to its process it uses ~4w at 800MHz

also they have software controlled clocking there are 2 PLL's those cards you see in that pic the white chip 1 holds a 750FX@800mhz and that heatsink can keep it stable at 1GHz same the TSI107 is a 7410 which can hit 600+ with that sink

we wrote a program to control the clocks within AmigaOS 3.X

Edited by kev7112001

Share this post


Link to post
Share on other sites
7 minutes ago, kev7112001 said:

750FX/GX are the fastest coolest per clock of all G3 due to its process it uses ~4w at 800MHz

also they have software controlled clocking there are 2 PLL's

 

4W, that's it? That would be 50% of the maximum that the 740 puts out. Very nice.

 

Software controlled clocking? Hmm. I'd have to read up about that and how Mac's work with those.

Share this post


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

 

4W, that's it? That would be 50% of the maximum that the 740 puts out. Very nice.

 

Software controlled clocking? Hmm. I'd have to read up about that and how Mac's work with those.

it will stay at default hardwired speed till set by software or powersaving states

Edited by kev7112001

Share this post


Link to post
Share on other sites

From my read the FX uses a standard PLL setup with the option of software control. Later versions jettisoned standard PLL clock multipliers. Haven't looked at the GX since I found an unlimited supply of the FX.

Share this post


Link to post
Share on other sites

Just looked at the 750GX pinout and it shows five lines: PLL_Config 0-4 so I guess it's not a "later" version of the G3 than the FX. Anybody got a link handy to the 750FX User Manual?

 

edit: never mind, I just hadn't moved the PDF out of the downloads folder to documents as yet. Same thing PLL_CFG[0:4] on the FX and something called "Dual PLLs" whatever that is? Is that what you meant by "Two" PLLs, kev?

Edited by Trash80toHP_Mini

Share this post


Link to post
Share on other sites
14 minutes ago, Trash80toHP_Mini said:

Just looked at the 750GX pinout and it shows five lines: PLL_Config 0-4 so I guess it's not a "later" version of the G3 than the FX. Anybody got a link handy to the 750FX User Manual?

 

edit: never mind, I just hadn't moved the PDF out of the downloads folder to documents as yet. Same thing PLL_CFG[0:4] on the FX and something called "Dual PLLs" whatever that is? Is that what you meant by "Two" PLLs, kev?

yes

Share this post


Link to post
Share on other sites

and i have all manuals and datasheets and errata for PPC's and many bridges/chipsets besides the impossible marvell series stuff which i really need

well i have some but i need the programmer manual

Edited by kev7112001

Share this post


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

I had no idea the 2300 was so similar in architecture. I can completely see what you are talking about. It appears that our two projects do certainly intersect. Anything you learn can be applied to future plans for the BlackBird project, and vice-versa.

Pretty much anything PowerBook in 603e is exactly the same thing as I see it until you cross the great divide into the PCI architecture. Anything before PCI is PCMCIA only.

Share this post


Link to post
Share on other sites
On 1/29/2019 at 12:24 AM, Paralel said:

I had no idea the 2300 was so similar in architecture. I can completely see what you are talking about. It appears that our two projects do certainly intersect. Anything you learn can be applied to future plans for the BlackBird project, and vice-versa. 

I finally figured as much. Major overlap's exactly why I've kept up a running commentary about my unofficial project here in your topic. We're pretty much working different sections of the same jigsaw puzzle.

 

Your Plan A is a (very cool) practical experiment with alternate avenues available in case of failure. You have a fairly clear path toward building a very complex new board, if and when after either success or failure at first pass. I'm researching methods for doing the least complicated experiment possible next to your Plan A, which is building a 1400 Processor Card. That's the type of simple electron plumbing I can do, no logic involved. Just solder the copper pipe network together correctly, add the proper gate valve for power and it should be good to go.

 

The really fun stuff starts when we begin to pull the 2300c architecture apart in order to miniaturize it for fitment into the Blackbird's Faraday cage. :approve:

 

p.s. to flesh out the post I made above when I was very tired: PCMCIA/CardBus and the Blackbirds

PCMCIA - "The original standard was built around an 'enhanced' 16-bit ISA bus platform."

CardBus - "CardBus is effectively a 32-bit, 33 MHz PCI bus in the PC Card form factor."

http://www.bixnet.com/pccaandex.html

 

Edited by Trash80toHP_Mini

Share this post


Link to post
Share on other sites
3 hours ago, Trash80toHP_Mini said:

p.s. to flesh out the post I made above when I was very tired: PCMCIA/CardBus and the Blackbirds

PCMCIA - "The original standard was built around an 'enhanced' 16-bit ISA bus platform."

CardBus - "CardBus is effectively a 32-bit, 33 MHz PCI bus in the PC Card form factor."

http://www.bixnet.com/pccaandex.html

 

Yeah, unfortunately, since it works as a PCI bus, it seems as though it would be very difficult to back-port it to any of the Powerbooks that use the '030 bus style, since that is more or less NuBus type (No such thing as an '030 bus that can handle PCI, at least as far as I am aware)

Edited by Paralel

Share this post


Link to post
Share on other sites
6 hours ago, Paralel said:

No such thing as an '030 bus that can handle PCI, at least as far as I am aware)

Anything is possible if you're crazy enough, as evidenced by the fact that PCI slot busboards for the Amiga 1200 exist. But, yes, like most of the truly far out Amiga expansions that's basically the exception that proves the rule.

Share this post


Link to post
Share on other sites
7 hours ago, Paralel said:

Yeah, unfortunately, since it works as a PCI bus, it seems as though it would be very difficult to back-port it to any of the Powerbooks that use the '030 bus style, since that is more or less NuBus type (No such thing as an '030 bus that can handle PCI, at least as far as I am aware)

actually there are 68k PCI bridges eg Qspan by moto there are a few others too already lookin at them for another project of mine

Edited by kev7112001

Share this post


Link to post
Share on other sites

https://lists.purdue.edu/pipermail/cytometry/1999-March/012466.html

 

PCI card to NuBus Slots is a thing, have that in house. Slots are on a passive backplane. The NuBus chipset on the PCI card could be deleted from its schematic. The remainder would be a raw 68k bus that could be torqued into a 68030 PDS to PCI adapter. If someone wanted to translate the PCI side of the card to CardBus you'd wind up with a CardBus to 68030 PDS adapter for upgrading a PCI architecture PowerBook to use SE/30 and IIsi cards. Drivers for doing that via the NuBus chipset are in house as well. PDS and NuBus cards are handled the same way in the NuBus architectire Slot Manager setup so that ought to work I would think.

 

You'd need to flip-flop the drivers to go bi-directional, but much of the work there has been done. I'll take a look at the interface card later if I have a chance and post the 68K<->PCI bridge chip's ID.

 

Edited by Trash80toHP_Mini
Apparently I can no longer typel ;-/

Share this post


Link to post
Share on other sites

in my other project we are trying to make a faster PCI to 030bus since at the moment the mediator board we use goes through the old slower Zorro bus which is a big loss atm since we only get a max 13MB/s bandwidth shared with a Voodoo 3

youtube has vids of our cards playing Quake 3 and RTCW etc etc

ther is another PCI bridge that just became fully opensource and just FYI Zorro bus is basically a 030 bus

Edited by kev7112001

Share this post


Link to post
Share on other sites

Unfortunately, it looks like they did custom bridge logic.

 

PCI-NuBus-Bridge.thumb.JPG.8f29141210f0f8189c167dbdba0b374b.JPG

 

Card's from 1995, what are the dates on your available bridge solutions? Might the Mac drivers be compatible? The bridge itself should be a black box, no?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×