> What about soldering in a 7448 in current CPU cards ? It's a pin for pin compatible chip.

7448 support may need an updated SmartFirmware.

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=4095&forum=11&start=16

Quote:

Beyond Andreas' answer, with your idea we are only going to be able to upgrade existing cards, not create new ones.
Also, what sets the voltage and multiplier on the these cards?
I don't know of any documentation for that.

Want a more useful project?
We need a donated cpu card and an adequate bounty, then I can pay a Chinese firm to back trace all the connections on the board.
With that and a good guess at the function of the buried layers we could start to re-create the card.

With that and the available schematics of the main board, the Pegasos II can be re-created and even upgraded.

I would really love to just lower the AGP voltage.

Quote:


Well, a recreated Pegasos II will still suffer from being non-RoHS and parts not avaialable any longer. It's a dead end.
If you want to do your own machine, better go for a simple new board.

Quote:

It would never be simple.
And the Pegasos design has one feature I really like, a separate cpu card.
And RoHS compatibility isn't an issue in the US for small runs (although most of the board could easily be converted to this).
The real disadvantage is the AGP slot (instead of a modern PCI-E slot).

On new designs, I would love something simple that could use a T1042 or T2081 processor.
1.4 or 1.8 GHz four core e5500 or four core (eight virtual core) e6500 goodness.
And it would be cheaper than an X5000.

[ Edited by Jim 12.02.2014 - 21:04 ]

> what sets the voltage and multiplier on the these cards? I don't know of any
> documentation for that.

The core voltage of the MC7447RX1000LB on the Pegasos II CPU card is 1.3 V. This voltage is compatible with the 7448, which can run at up to its maximum nominal clock rate of 1.7 GHz with that.
For the multiplier settings there's inofficial documentation up to at least 1.467 GHz:

* multiplier setting
* localization of SMD bridges
* example position of the SMD bridges

Pin compatibilities from 7447 (to 7447A) to 7448 are not quite 100% for every use case, which might impose implications on such an endeavour (replacing 7447 with 7448):

https://www.nxp.com/docs/en/application-note/AN2656.pdf

In case the target clock rate is below 1.5 GHz, maybe it would be better to use the 7447A (maximum nominal clock rate of 1.42 GHz @ 1.3 V) instead of the 7448. For this it would be handy to know if there were any changes done for the existing 1.4 GHz 7447A Pegasos II CPU cards (beyond the obvious thing of changing the multiplier from 7.5 to 10.5, that is).

Quote:

Would it? Someone would still have to build it. And every failed board is money wasted.

Quote:

Only making that statement based on parts costs, obviously development and production costs would affect the overall cost.

BTW - Way to go Andreas.
I have ever seen that info before.
So, if we could re-create the board, we could produce at least a 1.42 GHz cpu card.

Anyone got a cpu card to donate for the trace back?
That is be costly as it is ridiculously time consuming.

> Way to go Andreas. I have ever seen that info before. So, if we could
> re-create the board, we could produce at least a 1.42 GHz cpu card.

With the 1.42 GHz 7447A it'd be 1.40 GHz (multiplier of 10.5), like with the existing but very rare 7447A cards, or 1.47 GHz (multiplier of 11). And with a 1.7 GHz 7448 I guess it wouldn't be a problem to find out the remaining PLL configs for multipliers of 11.5 (1.53 GHz), 12 (1.60 GHz), 12.5 (1.67 GHz) and even 13 (1.73 GHz).

Quote:

That could be a real advantage to PegII owners.
Imagine the performance jump if a 1.7 7448 was used.

> On new designs, I would love something simple that could use a T1042 or T2081 processor.

The T2081 is pin-compatible with the T1 series but lacks SATA unfortunately.

> it would be cheaper than an X5000.

Would the price difference be more than the price difference between the T1042/T2081 chips and the P50x0 chips? If yes, why so?

> That could be a real advantage to PegII owners. Imagine the performance jump
> if a 1.7 7448 was used.

Note that the 7448 would still be crippled by the 133 MHz bus of the MV64361 (which could run at up to 183 MHz but is configured to 133 MHz in the Pegasos II).

Quote:

I don't anticipate an as complex a design as the X5000.
Just an MATX or FLEX board with a couple of expansion slots.
Also, I'd seriously consider a separate cpu card (to allow for upgrading the T1042 or possibly even using the T1021) and to reduce the number of mainboard layers.

If necessary we can add an SATA controller for the T2081.

[ Edited by Jim 12.02.2014 - 22:57 ]

Quote:

No more crippled than it is in a G4 Power mac accelerator, and they fly.

> I'd seriously consider a separate cpu card (to allow for upgrading the T1042 or possibly
> even using the T1021)

...or the T1022 I guess ;-)

> and to reduce the number of mainboard layers.

Less layers make for a cheaper design, but a separate CPU card is more expensive than the CPU soldered on the mainboard. I'm not sure if the result would be positive or negative on balance.

> If necessary we can add an SATA controller for the T2081.

...adding to the BOM and complexity again.

Quote:

Well...yeah, got that one wrong.

Quote:

Yes, but the separate cpu card is needed to minimize the number of layers and offers an advantage when configuring the system.

Quote:

Yes, but a PCI-E based SATA controller could be fairly cheap.

Overall, with the lowered cpu prices it could add up to a decent savings.

But in any case, I may buy an A-eon system anyway.
Its designers have my respect for their competence.

The downside is that this would probbly raise the price of the Pegasos boards, or upside depending on how you view it...

New CPU cards ?! Now come on guys, I was trying to be realistic :) I love the Peg2 too (have a miserable G3 though..) but the board is too outdated to justify new accelerators, unless a new motherboard with few upgrades (i.e. similar chipsets allowing for similar design) was made, which is obviously not likely (who's gonna pay for development costs ?).
The idea here was just that replacing the CPU with a 7448 could actually be feasible. The bigger cache probably compensates for the slower bus BTW...

"And with a 1.7 GHz 7448 I guess it wouldn't be a problem to find out the remaining PLL configs for multipliers of 11.5 (1.53 GHz), 12 (1.60 GHz), 12.5 (1.67 GHz) and even 13 (1.73 GHz)."

Would the board even allow 13x ? If yes great, it could be a start.

There is still the problem with the outdated AGP bus though.. And the firmware doesn't initialize hidden devices behind a bridge..
H264 hardware acceleration would make such a difference in video.

I know from reading the thread that a new firmware was in the makes, would it be that hard to get ?

BTW I don't even have OS4 or even MOS installed on mine ATM, but this would benefit both.. (MOS users could sell their machines easier too...)

[ Edited by Jose 13.02.2014 - 11:13 ]

> The idea here was just that replacing the CPU with a 7448 could actually be
> feasible. The bigger cache probably compensates for the slower bus BTW...

The bus wouldn't be slower with the 7448 than it is with the 7447 now. My consideration was just to note that the speed gain would be less than the clock rate increase alone would imply as the *relative* bus speed compared to the CPU clock rate would decrease.

>> with a 1.7 GHz 7448 I guess it wouldn't be a problem to find out the remaining
>> PLL configs for multipliers of 11.5 (1.53 GHz), 12 (1.60 GHz), 12.5 (1.67 GHz)
>> and even 13 (1.73 GHz).

> Would the board even allow 13x ?

I honestly don't know. 11x seems to be the maximum that was actually tested and reported by users. 5 binary digits can represent 32 different numbers, and with 0.5 steps and some naive thinking this could amount to a 2.07 GHz range from 0x through 15.5x, or from 5x through 20.5x, or whatever. This is pure speculation, though.

> I know from reading the thread that a new firmware was in the makes, would
> it be that hard to get ?

From what transpired over the years I'd say a definite yes :-)

One thing that puzzles me is that some machines were sold as low power servers, including for the small enterprise, what if the enterprise decides to add an add on card with a PCI bridge on it ?
So who knows, maybe there are machines out there with an updated firmware..

Quote:

There were probably some at Genesi, and maybe at some of their parnters/customers, but i doubt so. And in any case, you won't have access to it. The Pegasos had its time, you can do nothing but to speculate now.

Quote:

If the cpu cards can be re-created, then the entire design will be known and it could be re-implemented.
Yes, its dated, and a G4 outperforms it (let alone a G5), but neither of those can run both OS4 and MorphOS.

>>> some machines were sold as low power servers, including for the small enterprise,
>>> what if the enterprise decides to add an add on card with a PCI bridge on it ?
>>> So who knows, maybe there are machines out there with an updated firmware..

>> There were probably some at Genesi, and maybe at some of their parnters/customers,
>> but i doubt so. And in any case, you won't have access to it. The Pegasos had its time,
>> you can do nothing but to speculate now.

> If the cpu cards can be re-created, then the entire design will be known and it could be
> re-implemented.

How would this help with coming by an updated SmartFirmware that remedies the current limitations like the unability to see behind a PCI bridge?

> a G4 outperforms it

The Pegasos II *has* a G4 (unless it's a G3 one) :-)

Andreas, as you probably have guessed by now, that in no way addresses the SmartFirmware issue (which as far as I can tell can not be resolved).
And you have ignore my meager point (not that it worries me as I have no interest in building more Pegs).

If I could bankroll something like that, you know full well I'd consider something more up to date.

Quote:

If so, it was probably the 7447A based 1.4GHz model.

Quote:

Indeed.

The same goes for PPC desktop in general, I'd say...

Quote:

No performace _jump_ ->
https://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=4498&start=35

[ Editiert durch Zylesea 15.02.2014 - 00:15 ]