Upgrading PCI PowerMac 5/6x00 L2 upgrades (was) Sonnet Clocker II

[Trash80toHP_Mini]

Sounds like this might be a wedge for rework hacking a faster, lower power consumption, cooler running CPU on the slower L2 boards. If we can identify a suitable part, just how high can the multiplier be set in the extension?

Are there G3 parts with a higher clock than 500MHz?

Crazy notions:

Might a G4 conversion similar to the Pismo hack be a possibility? IIRC (probably not) that was a trade-in ProcSwap hack done by Wegener?

A GHz G4 ProcSwap for my Radius 81/110's PDS/NuBus Crescendo would just be insane.  8-o The 12" AlPartsBook would just die for the chance to organ donor transplant that lil' puppy.  

 

[trag]

Multipliers are controlled by three (? IIRC) balls/pins on the CPU.   Bias those balls in different combinations to get different clock multipliers.  Later revisions of the PPC750 assigned different multiplier values to some of those ball bias combinations.   So, for example, the pattern of ball bias that got you 8X on early PPC750 would get you 10X on the later 500MHz capable version or might get you 16X on a compatible PPC74xx (G4).

So, to answer your question jt, pick a pin compatible G4 (from reading the swap thread around here somewhere, there were such things) and then get the datasheets for the G3 and G4 and see which multipliers the corresponding ball-bias patterns yield.

G3s higher than 500MHz were generally the models that had internal cache (750cx, 750fx, 750gx) and those aren't pin compatible with the earlier G3s.   It's be cool to build a cache-slot upgrade based on the PPC750GX, but it would be a redesign, not a simple swap.

My personal interest is that I've been wondering what it would take to increase the cache from 512KB to 1MB on the cards with 512KB.    Replace the cache chips, obviously, but then after that is the cache size signaled on pins to the CPU?   Or is there a register setting?   If the latter, that extension editing utility might be useful.    If the former, there's probably a hardware solution, but depending how they built the card, it might be controllable through software (are the hypothetical pins/balls hard wired or software controlled). 

 

[Trash80toHP_Mini]

Thanks, trag, I'll have to dig around there, recollections from that thread's what sent me off on this tangent.

Dunno if it might put more or less stress on the heating/cooling budget than overclocking a standard accelerator by bus multiplication? A G4 TAM hack might be a worthwhile endeavor. Those crazies <wink> would pay good money for that kind of upgrade for their designer toys. They've driven the price of fast L2/G3s to the point that G4 rework on a cottage industry basis might be fun and make a little money for someone. It'd certainly grab some retro-computing glory.

I have higher hopes for the NuBus version though, plenty of heating/cooling margin to run a G4 in that Big Beautiful Metal Can with the Radius logo on the front panel. Might even work inside plastic boxen off underpowered PSUs from the fruit vendor too. [}]

Now that I think of it, I've got a pile of NewerTech G3 cards in NuBus/PDS to take a look at as well. Hrmmm?

 

[trag]

Well, now that I've identified a source of the X100 PDS slot connector, you could experiment with building your own PDS pass through adapter for the G3/NuBus upgrades.   Maybe reroute it so it doesn't block a NuBus slot.  

 

[Trash80toHP_Mini]

I've not got that blocked NuBus Slot problem in the Radius 81/110. LOTS of spare room for my tethered HPV/4MB card in that BIG metal can.  []

 

[Cory5412]

split from:

 

[trag]

Thank you Cory.

 

[Trash80toHP_Mini]

Yep, makes sense. Now to find a link to the ProcSwap thread that inspired this tangent.

 

[Bolle]

I have higher hopes for the NuBus version though, plenty of heating/cooling margin to run a G4 in that Big Beautiful Metal Can with the Radius logo on the front panel. 

Been there, done that