Bussing around with a G3 Beige XPC106 Memory controller upgrade

[LightBulbFun]

so I have a G3 beige mini-tower in pretty nice shape its a Rev B unit

now as many people know I love to tinker with my Computers and I knew that you could in theory overclock the bus on a G3 beige from 66.8Mhz to 83.3Mhz, so I gave this ago on my G3 beige a while ago

sadly it was not stable at all and would barely POST, Game over, back to the stock 66.8Mhz right?

Nope, Just replace the Main "North bridge" for a faster one, simples

heres what I did to get my 83.3Mhz bus speed G3 beige:

I took a spare Rev A G3 BW board removed and reballed the 100Mhz XPC106 off of it

then I took the G3 beige board, removed the 66Mhz rated XPC106, cleaned up the pads on the Logicboard and soldered on the "new" 100Mhz rated XPC106

and it worked! the G3 beige booted right up and I was able to set the bus speed to 83.3Mhz without any stability issues (of course im using PC100/133 DIMMs) (I ran a few GeekBench loops and over 120 passes in Gauge Pros memory tester without issue)

I do wonder if theres a Jumper config that would let me set the bus speed to 100Mhz, a 100Mhz bus speed G3 beige would be pretty epic.

the XPC/MPC106 also known as Grackle, is the Memory/PCI/Bus controller/"North Bridge" in: G3 beiges, G3 BWs, G4 Yikes!s, WS/PDQs, Lombards and tray loading iMac G3s. in theory the same swap I did here could be done to any of those machines alloying them to run a 83.3Mhz bus nice and stable

of course I have some pictures but "sadly" with my new equipment, the BGA soldering was going so well i forgot to stop and take some pictures during the swap... so no bare pads shot im afraid. (I got a T8280 Hot plate which combined with the TS100 soldering iron makes BGA soldering such big boards much easier )

and finally the money shot of the 100Mhz rated "North bridge" Grackle chip on the G3 beige logic-board where it says ARX100, it says ARX66 and ARX83 on 66/83Mhz rated parts, incase your wondering how to tell the speed of your XPC106 chip the chip I fitted/pictured was made 4th week 1999

 

[AphoticD]

Brilliant results again!

 

[weckart]

Was the flakiness of the B/W G3 A rev down to the controller or the firmware? Just wondering if you've swapped stabilty for performance.

 

[LightBulbFun]

Was the flakiness of the B/W G3 A rev down to the controller or the firmware? Just wondering if you've swapped stabilty for performance.

Good Question.

I was thinking about this while contemplating the swap. I think the flakiness of the Rev A G3 BW was down to its overall Logic-board design and possibly firmware, rather then 1 specific problem component. im pretty sure the Rev B BW and the Yikes! use the exact same Grackle chip but ill have to verify this (all 3 of my G3 BWs are Rev A units sadly) its also worth noting that currently im running the Grackle chip at 83Mhz rather then at the 100Mhz its rated for, so if it was on edge at 100Mhz it should be nice and comfy at 83Mhz (indeed it does run noticeably cooler then it did in the BW G3)

and so far the G3 beige has been nice and stable, in-fact I just finished using the G3 beige, to image 7.5.3 to my Macintosh SEs 40MB SCSI HDD (as my Macintosh SE is not a FDHD one and does not have any FDHD components/upgrades)

bonus shot of a 40MB SCSI Drive in Leopard

 

[weckart]

SSW 7.5.3 on an SE with 2.5MB RAM is just cruel.

You can just about run Finder and not much else. I use Zip and Jaz drives to get stuff on and off and those can also be used to run the SE on. Those 800Kb drives are a pain in the backside and it's getting hard to find 720Kb floppies cheaply to use on them.

 

[DearthnVader]

Great work, sound like it's time to clock chip too.

I forget the rating of the fixed frequency oscillator that sets the timebase on the Old World G3's, but you maybe able to push it to 100mhz by swapping it out.

However, it depends on what other than the bus frequency and memory frequency are relying on that oscillator, as things like built-in sound, ethernet, and video may not work very well if they are synced with it.

Normally the PCI bus has it's own oscillator, but it's been a long time since I had a Beige G3.

 

[LightBulbFun]

SSW 7.5.3 on an SE with 2.5MB RAM is just cruel.

You can just about run Finder and not much else. I use Zip and Jaz drives to get stuff on and off and those can also be used to run the SE on. Those 800Kb drives are a pain in the backside and it's getting hard to find 720Kb floppies cheaply to use on them.

Hey I was able to open Mac Draw II (but not claris works) however indeed 7.5.3 is quite sluggish on the ol 8Mhz 68000 CPU I will be getting 4MB of RAM for the SE at some point soonish hopefully, and get it online sadly this is my only 68K and I dont own any other SCSI gear. id like to get an external SCSI HDD at some point... luckily I have a bunch (like 30+) of TDK 1.44MB floppy disks that will happily format to 800K and seem to be holding their data at 800K. so I can at least get some stuff on and off the machine via the sneaker-net.

Great work, sound like it's time to clock chip too.

I forget the rating of the fixed frequency oscillator that sets the timebase on the Old World G3's, but you maybe able to push it to 100mhz by swapping it out.

However, it depends on what other than the bus frequency and memory frequency are relying on that oscillator, as things like built-in sound, ethernet, and video may not work very well if they are synced with it.

Normally the PCI bus has it's own oscillator, but it's been a long time since I had a Beige G3.

I dont plan to try and overclock this machine by changing any of the crystals, as you say, it will mess with the timing/speed of Everything and lead to no end of issues.

on the PCI thing, I know on the G3 beige the PCI Bus is tied somewhat to the main bus speed (for example 66.6Mhz and 83.3Mhz Bus speeds the PCI bus runs at 33.3Mhz but at 70Mhz the PCI bus runs at 35Mhz and at 75Mhz it runs at 30Mhz)

 

[LightBulbFun]

I have been doing a bit more sleuthing trying to figure out how to clock the G3 beige up to 100Mhz when I came across this:

https://www.nxp.com/products/no-lon...-evaluation-system-discontinued:YELLOWKNIFEX4 a Motorola development board/system called YellowKnife

its a very weird and interesting board, its of an ATX form factor with a fairly typical PC IO layout but is PowerPC based and has a MacOS ToolBoxRomImage and apparently is compatible with Mac OS? (they mention the TBXI is 4MB so im pretty sure it has OpenFirmware too)

spec wise its quite similar to the G3 beige (with some hints of G3 BW) and uses the same CPU Socket, although it uses an 11 pin jumper block like a G3 BW for settings rather then the 9 pin jumper block used in the G3 beige so its going to take extra sleuthing to figure out how to get the G3 beige up to 100Mhz bus

I have also been looking all over for information on the socket used in the G3 beige through to Yikes! systems, so im also glad to of finally found information on the socket including a detailed pinout (its actually 288 pins not 300 as commonly reported. its PC equivalent is Socket 3)

it would be interesting to hear if anyone else has see/come across the YellowKnife System before? I cant even seem to find pictures of it on the web sadly...

I highly recommend people click on the "documentation" tab in the link i provided, and reading the PDFs there

 

[AphoticD]

Introducing the only online image of the YellowKnife X4 Evaluation System. Restored in full glorious Beige (with yellow hints) from the B&W print in the user manual

PC Design in the 90s was so exciting. Just look at that arc in the front grill. For some reason Seaquest DSV comes to mind.

 

[amagichnich]

Just look at that arc in the front grill. For some reason Seaquest DSV comes to mind.

Reminds me of the Star Fleet emblem

 

[amagichnich]

Introducing the only online image of the YellowKnife X4 Evaluation System.

Can't get rid of the feeling that I've seen this thing before... I just don't know when and where

 

[pc297]

I do wonder if theres a Jumper config that would let me set the bus speed to 100Mhz, a 100Mhz bus speed G3 beige would be pretty epic.

Did you try to mirror the G&3 settings? The multiplier table at least seems to be backwards between the two (e.g. 3X, beige P2 P3 P4, b&w P1 P2 P3 etc), could the 100 mhz b&w setting (P8 P9) work in the beige as P7 P8?

 

[DearthnVader]

I wonder if @LightBulbFun can get to 87.5mHz bus speed?

Beige G3 Motherboard Clock Speed Jumper Settings

 

[LightBulbFun]

I wonder if @LightBulbFun can get to 87.5mHz bus speed?

Beige G3 Motherboard Clock Speed Jumper Settings

very interesting, Ill have to mess with that at some point I still wonder how to set the mythical 3x multiplier to get a full 100Mhz clock

 

[DearthnVader]

very interesting, Ill have to mess with that at some point I still wonder how to set the mythical 3x multiplier to get a full 100Mhz clock

Could be shorting PIN 5 alone?

I assume with a 33.3 mHz PCI clock.

That stands to reason, 6+5 is 2.0x 6 alone is 2.5x and five alone is the last and only choice 3.0x?????

 

[DearthnVader]

Seems to me there is one more setting on the bus, no pins shorted.

I'm going to have a launch an investigation.....

Also there are some undocumented settings for the PCI Bus speed too.....

I'd like to get that PCI bus speed up to 66Mhz, but I need to check if the Rage is 66MHz compatible with Reggie SE.....

If one of the undocumented settings for the PCI bus is 66MHz and one of the two unknown settings for the Bus speed is 1x, that will be the sweet spot, or 1.5x would be even sweeter.

Now if I can just find an XPC106APX100GB somewhere...

Looking at you @dosdude1 we can find a source for one, and you still do a little BGA rework, I've got my eye on a Rev. B logic board I can have shipped to you?

 

[DearthnVader]

Ok with no pin shorted at 5 and 6, and no pins shorted at 7, 8, or 9....

It does boot, I've got the CPU multiplier set to 3x....

Now it reports a bus speed of 83.333Mhz and a CPU speed of 250Mhz.

We just don't know what the PCI speed is now to know what the PCI Bus multiplier is right now.

We know it is not 35, 33, or 30....

Could be 41.5Mhz PCI, could really be anything I guess, I'll put the jumper back at pin 7 and see if it still boots....

 

[DearthnVader]

Ok with no pin shorted at 5 and 6, and no pins shorted at 7, 8, or 9....

It does boot, I've got the CPU multiplier set to 3x....

Now it reports a bus speed of 83.333Mhz and a CPU speed of 250Mhz.

We just don't know what the PCI speed is now to know what the PCI Bus multiplier is right now.

We know it is not 35, 33, or 30....

Could be 41.5Mhz PCI, could really be anything I guess, I'll put the jumper back at pin 7 and see if it still boots....

Odd goings on, so I set pin 7, that should be a 33.333MHz PCI bus with an unknown PCI Bus multiplier and it was a no boot...

No chime, no nothing but power on..

Then I set Pin 5 and left pin 7,8,9 open, no chime, no boot.

So I set pin 6 and left pin 5 open and now I'm right back to 83MHz bus, 33Mhz PCI and 250Mhz CPU.

Very odd, must be some closer relation to the PCI bus speed pins and the PCI multiplier than the chart shows.

I can't find my extra jumpers right now, I'll order some more so I can check the other undocumented settings of the PCI bus jumpers.

What are the undocumented possibilities of pins 7,8 and 9?

Just 7 and 9 jumpered, and no jumpers at all, right?

7&8 is 30Mhz
7 alone is 33.333MHz
8&9 is 35Mhz

Any other possibilities I'm not seeing there?

Oops, two more, 8 alone, and 9 alone.

Total of 4 possible undocumented settings.

 

[DearthnVader]

I found the 100MHz Bus setting!!!!!

She actually got to the desktop before she started glitching out GFX wise, then the window server crashed and I shut it down.

The CPU reported 500MHz but I never got to ASP to see if the Bus Speed reported 100Mhz.

I had only pin 2 set for 5x CPU, and Pin 6 set for 2.5x PCI Timing.

So I must have had a 40MHz PCI Bus, and that explains the GFX glitches pretty soon after boot and the Window Sever crashing.

The Built-in Rage is not 66MHz compatible, I did verify that with Reggie SE.

This is just a XPC106APX66GB so it's amazing if it really did get to 100Mhz and still managed to get to Tiger's desktop. Also, amazingly the XPC didn't hardly even get warm @83MHz, but that was only for 10 or 15 min. and I was not pushing the system.

I only have a MPC7400 rated at 400MHz from a Yikes in here, tho it seemed stable at 458MHz and I normally run it at 433MHz. I don't have the L2 cache enabled for these tests.



Come on @LightBulbFun drag out that old Beige with that 7410 and have a go, you can find room among the weeds somewhere.....

 

[DearthnVader]

So the Bus speed only reports 83Mhz with a 500Mhz reported CPU speed, but we know that the ASP can lie when it doesn't understand the system is overclocked.

The XPC isn't really getting that hot, but the CPU is smoking hot, so I'm pretty sure she's running at 100MHz Bus speed.

I know I have the CPU jumper #2 set only, that is 5x the Bus speed setting and we all know it.

The Rage chip isn't getting that hot, but GFX chips don't like PCI bus speeds that are not 33.333Mhz or 66.666Mhz if they are 66Mhz compatible.

40MHz PCI is bringing graphics glitches about 120 sec. into the desktop, then things start crashing.

We're going to have to see if we can get that PCI bus down and find a stable setting for 100Mhz Bus.

I did try pins 8 alone, and pin 9 alone, with pin 6 set and not set, and also pins 8&9 together with pin 6 set( that should have been 35Mhz PCI and 87.5Mhz Bus ) all of these were no chime no boot.

The GeForce 6200 PCI did not like the 40MHz Bus at all, no sync. Even tho it is 66Mhz compatible, it really expects 66Mhz or 33Mhz and is never tested in-between.

I really think the PCI bus maybe running a bit faster( 41.5Mhz ) so the systems bus could be as high as 103.5MHz.

Damn that fire that burned up our oscilloscopes.......

 

[DearthnVader]

I found the 100MHz Bus setting!!!!!

Oops, I got all excited, these little pins it's so hard to tell. I had pin 3 set on the CPU multiplier, so 6x83.3Mhz=499.8Mhz CPU.

I did not find the 100MHz bus setting.😭

 

[DearthnVader]

The Beige may use a Cypress CY2271A clock generator that is feed from a fixed frequency oscillator to provide the reference clock.

It should have a PLL, but that is likely what our jumper block is, the jumper block likely feeds an EEPROM that provides a multiplexer and the divide logic.

Likely why the MPC106 data sheets PLL(0-3) don't line up with the jumpers on the main board. We only have 3 or 5 pins depending on how you interpret the data sheet. the MPC106 Users guide likely has more info, but it's 435 pages.

We know the jumper block on the B&W/Yikes has a few more sets of pins, but there isn't much documentation I can find on setting bus speeds and reference clock(SYSCLK) on those system.

I'd be interested in setting the PCI Bus on the Beige to 66MHz, but everything that sits under mac-io is not 66MHz compatible, and we don't know if we can configure the SYSCLK multiplier to 1:1 with the PLL block and logic we have. So that may mean removing the EEPROM and imaging it to reprogram it's logic.

As far as 100Mhz bus operation I've not been able to find a PLL config for that, tho the MPC106 does support 3:1, so it should be doable, but again may require reverse engineering the EPPROM logic.

The are only a few unknown configurations for the Beige G3 SYSCLK multiplier. Both 5 and 6 open, and 5 alone bridged.

If one of those does not set the 3:1 ratio then we have to look at that EEROM and Cypress CY2271A.

The Cypress CY2271A has three clock outputs, A, B, C and we know they are PCI, Bus Clock, and CPU( or maybe just VC0 ).

The maximum speed for the VCO is 200MHz.

 

[LightBulbFun]

been looking at this myself a bit (to help someone attempt this mod but with an iMac G3 trayloader) and I think the critical stumbling block to 100Mhz on a G3 beige, is that sadly not all 4 of the MPC106 PLL config pins are brought out on the G3 beige's board, at least not to the jumper block, I do wonder if the remaining unusused 2 go to pull up/pull down resistors on the motherboard somewhere

because if we can find those points and change them accordingly then in theory all we need to do is set the MPC106 to the 3X multiplier for 33Mhz PCI bus setting and then we are laughing

here is the PLL config page from the MPC106's datasheet, note there are 4 config pins (0 1 2 3)

but note from the G3 beige overclocking guide you linked to above

only 2 jumpers are given for the actual MPC106 multiplier setting, rather then the 4 we need

I do note that these 2 jumpers on the block at PLL are 2 and 3, but what we need is 0 and 1, however I do note that for a 2x multiplier PLL config pin 0 does have to be pulled, so there must be a resistor on the motherboard doing that outside of the jumper block, so we just need to find that resistor and remove it and hope that there is also a point where we can add a resistor for PLL config pin 1


and we cant really use the G3 blue and white either here, because it also uses the 2x multiplier since its base PCI speed is 66Mhz (which then goes into a PCI-PCI bridge to get turned into 64 bit 33Mhz PCI, I do wish Apple had just given use 4 66Mhz slots instead!)

Wayback Machine

 

[DearthnVader]

been looking at this myself a bit (to help someone attempt this mod but with an iMac G3 trayloader) and I think the critical stumbling block to 100Mhz on a G3 beige, is that sadly not all 4 of the MPC106 PLL config pins are brought out on the G3 beige's board, at least not to the jumper block, I do wonder if the remaining unusused 2 go to pull up/pull down resistors on the motherboard somewhere

because if we can find those points and change them accordingly then in theory all we need to do is set the MPC106 to the 3X multiplier for 33Mhz PCI bus setting and then we are laughing

here is the PLL config page from the MPC106's datasheet, note there are 4 config pins (0 1 2 3)

View attachment 2401160
but note from the G3 beige overclocking guide you linked to above

only 2 jumpers are given for the actual MPC106 multiplier setting, rather then the 4 we need

I do note that these 2 jumpers on the block at PLL are 2 and 3, but what we need is 0 and 1, however I do note that for a 2x multiplier PLL config pin 0 does have to be pulled, so there must be a resistor on the motherboard doing that outside of the jumper block, so we just need to find that resistor and remove it and hope that there is also a point where we can add a resistor for PLL config pin 1


and we cant really use the G3 blue and white either here, because it also uses the 2x multiplier since its base PCI speed is 66Mhz (which then goes into a PCI-PCI bridge to get turned into 64 bit 33Mhz PCI, I do wish Apple had just given use 4 66Mhz slots instead!)

Wayback Machine

I think the Cypress CY2271A pull those missing two pins depending on the logic of the EEPROM it's connected to, but I could be wrong and they are pulled( programed ) by resister logic.

@joevt found the missing data to set the extra pins on the B&W/Yikes and there is a resister that can be removed to unlock 133Mhz Bus operation.

If we had a scope we could likely figure out what needs to be done for the Beige to unlock 100MHz bus operation. Sadly I lost my scopes in a fire many years ago, and really saw no reason to replace them.

I do have a friend local that could likely hook me up if I swung by his shop but I can't find any supply of the XPC106 rated at 100MHz.

Sure I could vulture one off a B&W or Yikes board, but then we lose that board. If I run across a damaged board one day it maybe worth it to satisfy our curiosity but without even a small supply of the 100Mhz rated XPC106's it really isn't an upgrade that helps whatever community of Beige users is still out there.


The other issue is the lack of much of any supply of the higher clock rate G3/G4 ZIF CPU's. Not much point in overclocking the buses if the CPU can't push data across the bus any faster.

If one day someone makes an imposer board for the ZIF Mac's it maybe worth revisiting, but as of now, it's fun to figure these things out, but not real productive.

 

[pete1]

Am following this thread with interest. I would be curious to know if it’s possible to do the 100MHz bus mod.

I had another thought, could the Heathrow chip be replaced with a Paddington chip in order to get 100BaseTX Ethernet onboard?