Sonnet 7200 PCI programming

[jbenam]

Originally Posted by olav

It has become obvious that these now are scarce items.. (!):

http://www.amibay.com/showthread.php...Crescendo-7200

Damn, the scalpers are already at work, eh? Looks like I won't be getting PPC on my A4000 using this either Oh well.

[olav]

How is the situation on A1200?
We need to get Elbox to open up their PCI-library.. How hard has this been tried?

[strim]

Originally Posted by olav

How is the situation on A1200?

The sonnet.library is working but buggy, partly it is a problem with pci.library interface...

We need to get Elbox to open up their PCI-library.. How hard has this been tried?

First you'd need to get them to reply any mail or answer the phone. Or generally give a f***.

Slightly OT, but can the Sonnet card be oc'd?

[Hedeon]

Originally Posted by olav

I got hold of it. Thanks.

Did you get one or all of them? Think there were 2.

Originally Posted by olav

Now, I just need to get the mediator. Currently only the Mediator 1200TX Black Edition seems available, directly from Elbox.

A1200 is currently not supported. I plan to return to the A1200 soon (my Blizzard is in repair/upgrade at the moment) but during the last tests I experienced some problems with the sliding Z2 window preventing the PPC correctly executing 68K functions. (The picture with the dual PPC output is from a program which is 68K but with PPC function calls; those seem to work fine).

[Hedeon]

Regarding the post on Amibay:

Just for people information, anyone wanting to use this in there Amiga, there is some more work than just plug and play. You need to modify either the card or the mediator to add a 3.3v line and you need to find suitable memory that it doesn't spew out, Kingston 608/8 or 608/16 work but a lot of other that I had didn't work also the drivers are experimental and not all warp is features implemented. you also need to be able to build/compile some source code.

But other than that it's all easy

Regarding building compiling: The programs need to be patched, not recompiled. The movies made showing the Sonnet at work are patched executables. However, vbcc is being adjusted / has been adjusted to support recompilation for the Sonnet.

Regarding memory there is a point. All 16 and 64MB 5V 168 pins EDO and FPM I tried, worked. All 128MB failed (memory corrupts over time). But there are different Sonnet revisions (A, B, etc.). They (Sonnet) claim with the latest revision to support even 256MB DIMMs. (On the Amiga, only a max of 256 MB is supported at the moment so the Z3 window does not have to move with access outside of the window).

Currently, most essential WarpOS functions are implemented. Whether they are working flawlessly is another question. Exotic features regarding the MMU are not and probably won't be implemented (like user adjustable memory protection).

[Hedeon]

Oh, and the pci.library (up to 9.11 tested) will just lock up when you add the Sonnet to your system. It also needs a patch.

If you put the Sonnet in a 5V system, it will actually show up (occasionally) during pciinfo (not locking up the Amiga) but the PCI registers cannot be manipulated (= the reason why it doesn't lock up).

[olav]

Originally Posted by Hedeon

Did you get one or all of them? Think there were 2.



A1200 is currently not supported. I plan to return to the A1200 soon (my Blizzard is in repair/upgrade at the moment) but during the last tests I experienced some problems with the sliding Z2 window preventing the PPC correctly executing 68K functions. (The picture with the dual PPC output is from a program which is 68K but with PPC function calls; those seem to work fine).

Good news, and I hope to be able to contribute to A1200 support once I have my system assembled etc.
I got hold of a couple of units, 400MHz, 1MB cache, G3s. Rev. B. From here: http://www.pasocomclub.co.jp/list/mac_cpu.html
Will continue to be on the look-out here in Tokyo.

Where do the sonnet.library developers communicate mainly, and how many are you?

[strim]

Where do the sonnet.library developers communicate mainly, and how many are you?

Just the two of us - Hedeon (Dennis) and me. But to be honest Hedeon is doing all the hard work - I am just helping with menial stuff . Plus grelbfarlk is doing some testing as seen earlier in this thread.

We use a private CVS server to manage the source code, but it automatically is exported to GitHub. And we also use GitHub to track issues. We also have a wiki there. Oldschool email for everything else.

[S0ulA55a551n]

Do I take that all the software to run on this will need to be patched ? Or will there be the possibility in future that stuff written for warpos will just run ?

[strim]

Originally Posted by S0ulA55a551n

Do I take that all the software to run on this will need to be patched ? Or will there be the possibility in future that stuff written for warpos will just run ?

Currently patching is done manually, but in the future a run-time patcher could be implemented for better user experience (i.e. "just run").

[S0ulA55a551n]

Wonder what the barrier that stopped elbox on this project. Must have been something. Run time patching sounds sweet

[strim]

Originally Posted by S0ulA55a551n

Wonder what the barrier that stopped elbox on this project.

I'm afraid we can only guess...

Some people think that Elbox announced Shark only to sell more Mediators and that they never had a working prototype. What they were showing to Amiga users was modded Sonnet, but discussions with Elbox people (back in the times when they still cared and talked with the comminuty) mentions a completely different bridge used to interface between PCI bus and PPC CPU (an IBM CPC).

My personal guess is that even if they started working on something, they didn't get very far, just never had the guts to admit that the project was cancelled.

[grelbfarlk]

I got things running again, I got into the amiga patch hole and just started digging myself out. AHI 5x apparently was a problem as well as RAM-Handler 44.24, some P96 variables and the 128MB DIMMs I was using successfully for a while have stopped working along the way.

[strim]

Originally Posted by grelbfarlk

Here's the pinout of the U8 spot.
<-PCI Slot
1 2 3 4 5 Tab
3.3 5V 3.3V Unknown 3.3V 3.3V

I've had some power supply problems so I haven't been able to test until parts show up Monday. I may just order an LM1084-3.3V for testing.

First thing, pin numbering should be from C21 towards PCI slot.

This pinout does not look correct to me. It's missing GND signal. There should also be GND on the Tab in my opinion... If you have 3.3V on the tab and it is working, it might be by coincidence.

I guess the (middle) pin 3 should be GND too. Pin 4 would typically be Vout on D2PAK5 type of regulator. See MCP1862 or NCP4632 datasheet for typical pinout.

[olav]

Is it, as of yet, support for using multiple sonnets? (Doing research in relation to parallel computing, this is an interest of mine..Amiga parallel computer. :P ).
I mean, with the Mediator we do have the freedom to connect multiple devices.
As each device would run its own "scheduler" (warpOS), it should be possible. In theory.

[Acill]

Very cool project. Elbox should have kept it up. They also had that coldfire board announced that never took off. It was also shown.

[grelbfarlk]

Originally Posted by strim

First thing, pin numbering should be from C21 towards PCI slot.

This pinout does not look correct to me. It's missing GND signal. There should also be GND on the Tab in my opinion... If you have 3.3V on the tab and it is working, it might be by coincidence.

I guess the (middle) pin 3 should be GND too. Pin 4 would typically be Vout on D2PAK5 type of regulator. See MCP1862 or NCP4632 datasheet for typical pinout.

Yes I looked for something that would match the pinout of what I measured and didn't find anything yet. Thinking the pin I marked as unknown must be ground I hooked it up to the ground pin of the VRM. BAD IDEA, 3.3V line read nearly 4V. Which is why the wire connects to the ground on C21. On the backside of the board from U8 are pads for a couple resistors and caps, suspiciously like what you would see for the ADJ pin to use a non fixed VRM, but what do I know (not much). Even just using a 3 pin VRM of the SMT variety should work fine, haven't got around to ordering one yet.

[strim]

In the meantime... we've just passed 500 commits. This project started over a year ago.

[olav]

From the Elbox website:

"The Radeon Mediator drivers are designed for all MEDIATOR models. However, as most of the Radeon 9200 PCI series cards require the 3.3V auxiliary PCI power supply, they will work out of the box only in the Mediator 1200 TX and Mediator 1200 SX busboards with the ATX power supply.

PCI Radeon cards may also be used in other Mediator models, but it will require installing the Radeon card in the special PCI adapter which provides an auxiliary 3.3V power supply. Elbox is planning to produce a PCI 3.3V power adapter riser. It will include the bracket for the low profile Mobility Radeon 9200 card model available in Elbox Online Store."

I suppose this could also be applied to the Sonnet..
Does the Mediator 3000Di and 4000Di 3V require an ATX psu to supply the 3V?