TF536 Source Code?

[terriblefire]

Can I just point out that using an 030 with Kick 1.3 is a really bad idea. The OS has no awareness of Cache and makes some assumptions that the TF536 is not designed to support. Its highly likely you have lots of issues even after you fix this. Especially with Zorro cards.

I thought the A3000/030 shipped with Kickstart 1.3 and Workbench 1.3?

I'm pretty sure OS 2.0 was developed for the A3000? it may have shipped like that but its still a bad idea as the A3000 has ZIII and 1.3 has no ZIII support.

[Alex Oughton]

The A3000 did indeed ship before 2.0 was available. Early versions use a "super kickstart" which allow you to boot 1.3 and 2.0 from disk. Trying to use the A3000's 1.3 was one of the first things I attempted here, but I couldn't get past the incompatible scsi.device which is included in the "bonus code" appended to the image which also updates expansion.library.

Also I bought this A2000 with a different 030 accelerator in it (from GVP), and it had never been upgraded beyond 1.3. The manual states no need for upgrading to a newer kickstart, and the sellers (the original owners I met at a VCF swap-meet) didn't report any issues. So at least that specific accelerator didn't have any issues on 1.3.

But of course this doesn't account for any differences between the GVP accelerator and the TF536. I will similarly note when I publish this that instability could be experienced due to the untested nature of it, and this is more of a curiosity than a suggested setup. After all I only took this on just to see if it could be done.

[terriblefire]

The A3000 did indeed ship before 2.0 was available. Early versions use a "super kickstart" which allow you to boot 1.3 and 2.0 from disk. Trying to use the A3000's 1.3 was one of the first things I attempted here, but I couldn't get past the incompatible scsi.device which is included in the "bonus code" appended to the image which also updates expansion.library.

Also I bought this A2000 with a different 030 accelerator in it (from GVP), and it had never been upgraded beyond 1.3. The manual states no need for upgrading to a newer kickstart, and the sellers (the original owners I met at a VCF swap-meet) didn't report any issues. So at least that specific accelerator didn't have any issues on 1.3.

Remember those old GVP 030s are designed to inhibit cache on more of the memory map. The TF536 assumes the OS knows what its doing so doesnt inhibit cache as much.

But of course this doesn't account for any differences between the GVP accelerator and the TF536. I will similarly note when I publish this that instability could be experienced due to the untested nature of it, and this is more of a curiosity than a suggested setup. After all I only took this on just to see if it could be done.

Did you think about removing the expansion.library from that kickstart and putting in the vanilla 1.3?

[Alex Oughton]

Remember those old GVP 030s are designed to inhibit cache on more of the memory map. The TF536 assumes the OS knows what its doing so doesnt inhibit cache as much.

Ah, interesting. Well if things are really screwy under 1.3 I may try enabling the cache disable pin on the card (I use a physical switch on my KS ROM to go to 1.3, so I will tie them together).

Did you think about removing the expansion.library from that kickstart and putting in the vanilla 1.3?

Unfortunately I haven't been able to do so. The tools I found to extract libraries from ROM images don't understand the "bonus code" file to be able to try that.

So far my limited testing has been promising. I actually got the hard disk to read just now (running "binddrivers" after running my tool is the only needed step). Under "sysinfo" the reported speed is the same as when I have cache disabled in 3.2's early-boot menu. Perhaps this indicates the cache isn't enabled and things will be OK...

I still have this running as a post-boot tool (which would go in startup-sequence). The biggest downside of that is that everything up to that point is loaded in chip RAM (with all the performance implications that brings). So I'm still working on turning this into something placed in the KS ROM itself. The simplest idea would be to compile a version which only fills the SHUTUP register. I will then look at injecting that into the KS 1.3 ROM in the same way I have seen people inject the GAYLE-enabling code. This would run prior to expansion.library so that it can enable my Z2 RAM and use that instead of chip RAM wherever appropriate. It should also enable AutoBoot from the SCSI card.

[terriblefire]

Try this for an experiment. I've not tested it but the boot rom checks the kickstart version and shuts up the TF if its less than 37.

tf536r2_2022_12_11_76b5c72_alpha.zip

this is not a general firmware i'll post a new one if i get this working. If anyone on 3.1 wants to test its working ok still please feed back. If its not working the symptoms will be no fastram on 3.1.

[go0se]

On 3.1 (40.063) it boots to ROM splash or IDE if connected. 64MB Fast is recognised under WB and it passes the first few rounds of ATK memory test (before the memory test was abandoned).

On 2.0 (37.175) it fires up to the ROM splash as normal.
On 1.3 it fires up to the ROM splash as normal.

I assume by shutting it up if it version detects < 37 that is not configuring Z3 fast RAM, so the 536 is essentially operating as an 030 CPU only?

[Maximilian]

What about Kickstart 1.4 Alpha 18?
http://theamigamuseum.com/amiga-kicksta ... -1-4-beta/

[terriblefire]

I assume by shutting it up if it version detects < 37 that is not configuring Z3 fast RAM, so the 536 is essentially operating as an 030 CPU only?

the ram is still available but needs added manually

[Alex Oughton]

Try this for an experiment...

Just flashed it. Looks good, thank you!

Behavior under 1.3:

As expected. No Z3 RAM, but Z2 cards (RAM and SCSI) are working as normal. AutoBoot even being attempted (and would likely work to completion if I had something 1.3-compatible on my disk right now).

Behavior under 3.2:

Normal. Z3 RAM and Z2 cards available.


Over the last few days I've been narrowing-down the cause of some instability I have been experiencing under 1.3. Since you had mentioned likely problems with the cache, I tried disabling that. No improvement. Tried disabling MMU. No improvement. Tried not enabling the Z3 RAM, and now appears stable.

So, sending SHUTUP to the Z3 RAM and not manually adding the RAM seems to be the way to go. At least so far it's looking good for me, but the instability is sometimes intermittent, so more testing will be needed to be sure.

What about Kickstart 1.4 Alpha 18?
http://theamigamuseum.com/amiga-kicksta ... -1-4-beta/

Previously tried that as a first step to boot the A3000's Super Kickstart. I did manage to remove enough A3000-specific modules to get it to attempt a boot from Super Kickstart, but then it would get stuck due to incompatible modules in that.

[terriblefire]

If that firmware works for all then we're home. It should do everything needed to allow kick 1.3 machines to boot and configure the Z2 chain.

EDIT: If you disable all the fastram then your machine is running at 7Mhz. So adding the 030 is pointless.