Hi Enrique!
D'oh! Too much desktop upgrading planning! The ball grid package versus the
flat pack is something to watch for, but this may prove to workable or even
beneficial if there is enough room to put in a *small* daughterboard along
with a quad flat soldered adapter to a BGA socket. The daughterboard would
contain the CPU along with the voltage regulation to change the 3.3V Vdd
inputs to 2.5V, and change the multiplier from 3x to 8x, or possibly
jumpered so it is user configurable. This neatly sidesteps any need to probe
the motherboard and find out what the heck is going on with the input lines.
The idea would look like this (I'm too lazy to format for width, so use your
imagination that it all lines up nicely):
______Heat exchanger___ (silver foil? attaches to frame)
==========740======= (BGA)
*********socket**********
--------Adapterboard-------------- (voltage reg, sticks out a bit 1 side,
mult)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! (socket pins (ball grid) to fit the
socket below)
***BGA/QFP adapter***** (soldered into where QFP pack went)
############################## (logic board)
This introduces a few potential problems. First, the fact that there may not
be any headroom for such a daughterboard may be a show stopper (The quad
flat is a little taller than a the ball grid package, but add an adapter and
socket...). Second, I don't know if there is such a thing as a BGA->QFP
adapter (21mm square vs. 32mm square, so there is room width-wise). Having
one made/tooled could get very costly. Lastly, although the voltage
regulation circuit on an adapter board is a neat way to side step altering
the Vdd lines on the motherboard, it adds more bulk to an already cramped
space; but the real problem here lies in the production of such a
device/board, which changes the project from an individual garage/basement
hack to something a little more serious. Anyone familiar with board design?
I'm unfamiliar with the costs of producing something like this. As I recall
the original duo series used the frame as a heat sink (I owned a 230 and a
270c before my ill fated 5300), does the 2300 still do this? Perhaps some
room can be gained by cutting off the heat sink plate and using a silver
foil transfer to pass the heat. It'd be nice to have someone post some
pictures of the internals, as the apple dev notes don't have any mechanical
diagrams to accomodate the hackers.
Some more thoughts on this approach. Even if there isn't enough room
heightwise, one could try using just the BGA to QFP adapter and alter the
motherboard Vdd lines to 2.5V (eliminating the adapterboard), so the project
can still work, and minimize the height. It's just not as tidy from a
product point of view. The best part of a BGA package is that it isn't
soldered like the QFP, so if one so chose they could conceivably
upgrade/downgrade the package to a higher speed CPU if they chose, or even
plopped in a 133MHz (PID6) 3.3V 603e if they wanted to (and overclock it),
if they found that the project isn't working well. One could even attempt to
put in the C500/C600 G3 ZIF upgrade card or try a 750 with a backside cache
upgrade...
And as far as the 6500/300, I can't say as I don't have one, but I'd bet
it's the same as the 6400/xx0 series, which means that they are all QFP
packages. The reason I say this is because these machines need to be G3
upgraded by the ROM SIMM, so the CPU is not removable. Otherwise, all those
3rd party upgrades would be via the socket, like the C-class UMAX/Pulsar
series, which can be upgraded both ways (socket and SIMM).
Hope I've provided some food for thought...
Tony
PS. I've posted this to the duolist thread as you've raised a very good
point about the BGA problem with the G3 series. I hope you don't mind.
----- Original Message -----
From: Enrique Montiel <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Sunday, August 01, 1999 10:34 AM
Subject: 2300/300 crusade -- 2300/G3 crusade instead!
> Hi Tony! This is Enrique.
>
> Ahhh, another kick butt kung-fu crusader! Thanks so much for paying
> attention and becoming involved. I noticed yesterday that the 740 and the
> 603e are pin compatible. But my question to you is: How do you suppose we
go
> from a CBGA package to a CQFP package?
>
> And, do you know if the 6500/300 PowerMac series came in CQFP packaging?
>
> > Just following this thread here, and I'd like to point out a couple of
> > things, as I was looking into this a couple of years ago.
> >
> > 1. The maximum bus speed of the duo is 33MHz, anything over that is not
> > guaranteed to work on every machine out there, as that would overclock
the
> > bus. To achieve 300MHz, the bus would have to be run at 37.5MHz, as
noted in
> > point #4. I'm not sure which subsystems would be affected by
overclocking
> > the bus, there may be serial port communication/localtalk problems.
> So 37.5 is if we're going to go G3 right? - In my writing to you here I
> will be considering both 740 and 603e - I'm not a tech, bare with me.
>
> This is a message from The Dynamic Duo author Philip Smith
>
> Nothing much should happen really. I would put
> some heat sink goo on it though. Maybe even build
> a custom heat sink. Remember the 6500 and 5500
> are on a bare board and have 50Mhz buses. Right
> now I am extremely busy. We are about to make a
> national presence I beleive. Hopefully, by the
> time MacWorld san fran comes around we will have
> REALLY big news as far as Apple goes.
> --- Enrique Montiel <[EMAIL PROTECTED]>
> wrote:
> > Philip,
> >
> > I wanted to ask you what you thought would
> > happen if the bus speed on a 2300
> > is overclocked from 33.33mhz to 50mhz? What are
> > the negative consequences,
> > if any, that would arise?
> A suggestion I made was to take a 200mhz 603e (they are available in CQFP)
> and overclock the bus to 50mhz to get 300mhz. The maximum multiplier on
the
> 603e is 6x.
>
> >
> > 2. The 603e and the 740 are pin compatible. At higher speeds of the 603,
a
> > 2.5V input is required for the core, which is the same requirement as
the
> > 740. The maximum input recommendation is 2.7V to the core and no more.
The
> > 750 is not pin compatible with the 603e.
> What do you know about adjusting the core? The 2300 is very similar to the
> 5300 that you were gathering data for.
> >
> > 3. The 740 is a higher performance implementation of the PPC
architecture
> > than the 603e.
> Perceivable benefits of the 740 over the 603e you would speculate?
> >
> > 4. The maximum multiplier of the 740/750 is 8x currently. Rumours of 10x
> > implementation are appearing for the 750 (755?), but nothing is known or
> > guaranteed.
> Max multiplier of the 603e is 6x. One of the other crusaders has been
asking
> me if I knew how to set the clock multiplier. Do you know how to do this?
> (For me: What's the difference between the clock multiplier and bus speed)
>
> >
> > 5. The best performance you could achieve would be 33MHz x 8 = 266MHz.
> This is with 740 correct? 33mhz x 6 = 200mhz for 603e
> But what's the base chip speed? This is where I'm confused. I'm missing
one
> tiny bit of information: a basic point of reference to quantify from.
> >
> > 6. There would be no backside cache, which is the main performance boost
in
> > the G3 line.
> This is true and we'd like to consider after our success here to find a
way
> to have L2 on the DuoDock
> >
> > 7. Motorola and IBM have recommended that the input lines be on a
separate
> > plane than the output lines and the bus lines. Whether Apple has
followed
> > these recommendations is unknown.
> How may we find this out? The Dynamic Duo Philip's cousin worked on the
> development of the Duo line. But it was pre 2300.
> >
> > 8. The multiplier sense pins are driven high-low-low-low (1000) for a 3x
> > multiplier in the original 603e at 100MHz. The same lines to are used in
the
> > 740 and need to be driven high-high-low-low (1100) for a 8x multiplier.
Thus
> > pin #1 (of 0-3) only needs to be driven high.
> This I know nothing about. I'd like to understand it. But do I need to
> understand it?
> >
> > 9. The maximum power consumption of the 603e at 100MHz is in the 3W
range,
> > the 740 at 266MHz is in the 6-7W range. The average power consumption is
> > about 40%-60% of maximum. Driving the processor at 3.3V is not
recommended
> > as it will, in most likelihood, severely damage the chip, double the
power
> > consumption, and decrease battery life. In an over-volted case, a heat
sink
> > and fan would be necessary to cool the chip or it will fry itself.
> That's twice as much. Wow! If it's underpowered how will it damage the
chip?
> When you write: over-volted do you mean at 6-7W or higher? Could we get
away
> with a great heat sink and a breathing tube?
>
> Doing the mod with a 200mhz to 300mhz 603e it has been suggested to use
> tungsten as the heat sink.
> >
> > 10. The performance of the chip, lacking a backside cache, would be
severely
> > compromised. The main memory bus will run at 33MHz and the bit width (32
> > bit?) would be insufficient to feed such a fast CPU adequately.
> So would it be bottlenecked or would it break?
> >
> >
> > As you can see, none of these problems are insurmountable, with the
> > exception of #2 and #7. This would involve the most detective work to
> > discover which pins are the core input lines (Vdd) and regulate them to
2.5V
> > from the 3.3V. In the cramped quarters of the 2300, this may prove
> > impossible.
> Are these pins on the actual chip? How small can a regulator be made?
> >
> A key issue is the different packaging. I have possession of 4 300mhz 603e
> chips but they are in CBGA packaging. The 2300 is in CQFP. I've asked if
> there is some form of converter that can be used. The CBGA chips I have
are
> so tiny that I believe they would still fit into the 2300 if the converter
> would be small enough.
>
> Yours,
> Enrique Montiel
>
>
> Creative minds create solutions!
----------
Duo/2400 List, The friendliest place on the Net!
A listserv for users and fans of Mac subportables.
FAQ at <http://www.themacintoshguy.com/lists/DuoListFAQ.shtml>
Be sure to visit Mac2400! <http://www.sineware.com/mac2400>
Dr. Bott LLC | Get a LapBottom with new �GRIPPIER� gripper pads!
| http://www.drbott.com/prod/lapbot.html $23.00
Midwest Mac Parts <http://www.midwestmac.com> | After-market parts
for Macs. Fix your ticking dock for $59 w/ trade-in of PS. 888-356-1104
PowerBook Guy is | Click here!
Everything PowerBook! | http://www.powerbookguy.com
