68060 on ACA500 Amiga 500

As an experiment , i have tried and now have running a Blizzard 68060 on an ACA500 ... Amiga A500!!!!

more testing to do but I'm tired now but here it is, the 3.1 rom selection will not boot the 060 (you get a yellow screen / rainbow crash) but the 1.3 rom will and shows full 68060 and Ram and it boots from floppy....

sysinfo , demos etc later, i'll also try the scsi and 128mb

Specs

Amiga 1.3 512k , 512k OCS Rev5
blizzard 68060
ACA500 2mb fast

Also the 1.3 cannot boot in to 3.1 hard drives , so i'll format a 1.3 setup on a spare CF card and see what happens..

mike

Worlds first??? probably not :-) surely this needs a separate photo booth moment...

http://youtu.be/eNCe1KsLcO4

Nice :thumbsup:

Almost makes me want to buy the ACA500 .. wait .. should I? :unsure:

Awesome it should be able to run ScummVM ECS well with that config :lol:

If only there was a case for it...

A case.... That my next project.

which 3.1 did you try mike?

Abso-freakin'-lutey awesome!

How fast is WB1.3 on the 060?

I'm sure different tests are in the cards...but maybe a soft kicked 2.0/3.0/3.1 of various types (500/600/1200) will work?

Mmmmmm....BlizzardPPC / 060 / BlizzardVision anyone? :)

IMO it is "working" but it is most likely really slow.. (compared to "real" 68040/060)

AFAIK 68040+ cache line fill burst fetches are not compatible with 16-bit data bus. (ACA500 and A500) and it "works" because 1.3 won't enable any 68040/060 caches or features. 3.x (perhaps even 2.x) enables instruction cache automatically (=crash).

I'm looking forward to read more about this.

What about using a real 2.x/3.x ROM on the mobo?

Quote:

---

Originally Posted by Toni.Wilen

IMO it is "working" but it is most likely really slow.. (compared to "real" 68040/060)

AFAIK 68040+ cache line fill burst fetches are not compatible with 16-bit data bus. (ACA500 and A500) and it "works" because 1.3 won't enable any 68040/060 caches or features. 3.x (perhaps even 2.x) enables instruction cache automatically (=crash).

---

yes, the 500 isn't exactly optimised for it :lol: Cool nonetheless :)

Quote:

---

Originally Posted by mjnurney

Worlds first??? surely this needs a separate photo booth moment...

---

Not really, sorry :) I ran my Blizzard1260 on the ACA more than a month ago, but not being able to use kickstart 3.1 is a problem.

BTW: running a BlizzardPPC won't work, so running AmigaOS4.1 in the Amiga 500 is a no go for now... (that would have been a first!)

---------- Post added at 09:40 ---------- Previous post was at 09:40 ----------

Quote:

---

Originally Posted by Astral

Mmmmmm....BlizzardPPC / 060 / BlizzardVision anyone? :)

---

Doesn't work, sorry :(

Do you think there's potential for a 68060 library that would work on the 16bit bus then (disabling the cache?)

I'm sure it would still be advantageous to have at least the Ram and some CPU boost, would it still be faster then an 030?

Can you get a MIP's output in SysInfo Mike or does that guru?

But KS 1.3 uses no Caches, so this 060 solution is not really fast...

I guess it should be possible to create ACA500 specific MMU tables but it would not be compatible with any existing 680x0.library or any program that requires MMU, possibly incompatible with whdload too.

Is it possible for you to do a bustest ? I would like to see the chipram access with it... Thanks.

EDIT : AOS 2.x required... Forget about that

I think this calls for a....

:thumbsup: :thumbsup:

:D

true speed will be an issue and its ultimately pointless but hey nice to see WB 1.3 with an 060 :-)

I'm sure Jens could get 3.1 running on the 500 if he wanted to....

which version of 3.1 was you using mike?

it fails booting from the 3.1 rom on the aca500 roy

Forgive me for being a bit dumb but can you re-flash the kickstarts on the ACA500?

@Mike

Have you tried booting the 060 under 3.1 without RAM (I have a suspicion ;))

Jens said no 32bit card will work on 16bit A500 properly. Fast 030 is a max.

030 cards are 32 bit.

Hi Keith

same result, yellow screen and instant reboot - cycle

Could try turning some of the option(s) on/off in the blizzard card menu on a A1200 then re-test on A500.

Quote:

---

Originally Posted by Tahoe

Quote:

---

Originally Posted by mjnurney

Worlds first??? surely this needs a separate photo booth moment...

---

Not really, sorry :) I ran my Blizzard1260 on the ACA more than a month ago, but not being able to use kickstart 3.1 is a problem.

BTW: running a BlizzardPPC won't work, so running AmigaOS4.1 in the Amiga 500 is a no go for now... (that would have been a first!)

---------- Post added at 09:40 ---------- Previous post was at 09:40 ----------

Quote:

---

Originally Posted by Astral

Mmmmmm....BlizzardPPC / 060 / BlizzardVision anyone? :)

---

Doesn't work, sorry :(

---

For now :) Until EVERYTHING has been tried (initial thoughts of custom firmware, various settings, modified kickstarts, hacks etc) I don't think that "it doesn't work" is a result to accept :p This IS Amiga, and there is some VERY talented people in the scene...so let's just see...

I have one on the way also, when it's here I will do some tests.

The thing is, if you have an ACA 1220 or 1232 on their, it's more than enough.
Let's face it, what is the point of having 060 or PPC in any Amiga ?

OS4.1 and few PPC games and demo's is the only thing, with the new Cyrus coming, I think that would be a better choice if you really want to get into that.

And for demos, most 060 demos are AGA, so you'll lack the most important thing.
Zeus68k will be a lot better if it really has AGA.

Very interesting, I never thought that an 68060 can run on a A500 but dreams come true sometimes ;) thank you jens that you made the ACA500!

from jens on 1K.org

.. and the statement brings us to a really simple solution, which is unfortunately not occurred to me: you can turn off the CPU in "transparent translation registers" of both data and instruction caches as well as the 060 CPU. The TTR eke the Amiga a nearly unnoticed existence, because they always relate to a 16 MB block.

For the ACA500 but this is ideal because the * entire * lower 16MB block (just the entire A500) is from the perspective of the 68060 non-cacheable, ergo it does not hurt to disable the caches on the way depend on your location. Fastmem is outside, so grab because the caches again, and the B1260 * might * work. Subjunctive, because a matching 68040 library is missing, also retains this setting.

Groundwork for this is already done, coming in the next version of Flash.

Jens

http://www.a1k.org/forum/showthread.php?t=41442&page=3


Here is a beta version for Blizzard owners . The 68040/68060-Karten work, but the library must not be used under any circumstances , otherwise again , the caches are turned on in the lower 16M , and that pops immediately.

You can display the processor to the CPU command , normally turn the caches (which are active only on the local Fastmem the B1260 ), but the Blizzard SCSI module will NOT work , because running out of DMA controller of a fully functioning MMU - and is not there without a 68060.library replacement.

In the ACA500 menu a little more info text is now added to come - but with F10 in the expert menu , or use the TAB key call up the info screen : It is now also checked whether the $ f8 - range over which the ACA500 usually their Residents fades , is covered by the 1200 Turbo card (" shadowed "). In that case , this means that the function of Maprom Blizzard is active, and menu options can not be changed until the next power cycle . We have not yet succeeded in a " Maprom " which is turned on in the Blizzard once off again (hints welcome).

Other changes:

- ICache is now turned on the ROM Unzip , so the unpacking with 32- bit Turbo card is noticeably faster

- If B1260 is found, the unpack routine is placed in the Fastmem this card , which makes the unpacking process rapidly

- Maprom function of ACA12xx card is automatically used so that ACAtune is not needed (which anyway does not work with the ACA500 )

- $ C0 memory is automatically switched on at ACA1230 , so that Execbase is in any case as soon as possible Fastmem - again, no more need for ACAtune

This is an interim release . Tests should be done only if you have created a rescue disk. It might have crept in new errors that make it necessary to go back to an older version back !

The resident- command " acamaprom " also gets another update, so the Maprom function of the Blizzard can be activated ( ie you can load kickstart files) .

I've already tested the new firmware - the 1260 works now, but a new 060library is needed.

thanks Mike.

me too , just done it. Does the Blizzard 060 libs work?

---------- Post added at 17:46 ---------- Previous post was at 17:44 ----------

Flash update: files on next post

flash pics and file.

Copy unzipped file to your CF card and run the flash up date tool from the advanced menu.

( create a flash recovery disk first)

Okay

So the new firmware allows booting into 3.1 Workbench but it then reboots as though the 060 libs are wrong or something else is.. if i boot with no startup sequence then i can get into workbench but obviously SYSinfo crashes the system ...

on a clean CF WB install , it loads up okay and runs very quick. however sys info or sys speed still crashes it.

mike

Great progress, so a new 68060 library would do the trick. That's almost an excuse to buy another 1260 :D

How does the Maprom work if you want to load your own custom Rom?

at the moment any Blizzard tool crashes the system. anything that seems not too taxing is working okay.

a vanilla 3.1 install is the only to get my one working , so no WHDload or demos at the mo...

i'll continue playing

Quote:

---

Originally Posted by mjnurney

thanks Mike.

me too , just done it. Does the Blizzard 060 libs work?

---------- Post added at 17:46 ---------- Previous post was at 17:44 ----------

Flash update: files on next post

---

No, sadly we need a reworked 060lib for the combination ACA500/1260. :(

damn ... you any good at coding?

Also the SCSI doesn't crash the ACA500 now but I'm getting the wrong ram allocation..

mike.

i think that should 96mb ...but i could have the wrong ximm in (i did , ram allocation working great)

It runs fine as long as you don't run any programs that touch MMU or TTRs (transparent translation registers) or it will immediately hang the system because they restore original and incompatible cache settings for lowest 16M of address space. Also FPU is disabled and can't be used.

68040/060 library update or patch is needed that keeps lowest 16M totally uncacheable (for both data and instruction caches).

btw, "C:CPU cache" can be used to enable data caches (noticeable speed increase) but it will break SCSI expansion because it is DMA and DMA controllers require proper MMU tables with 68040 or 68060 CPUs.

thanks Toni i will try as I'm still more or less sober

Quote:

---

Originally Posted by Toni.Wilen

It runs fine as long as you don't run any programs that touch MMU or TTRs (transparent translation registers) or it will immediately hang the system because they restore original and incompatible cache settings for lowest 16M of address space. Also FPU is disabled and can't be used.

68040/060 library update or patch is needed that keeps lowest 16M totally uncacheable (for both data and instruction caches).

btw, "C:CPU cache" can be used to enable data caches (noticeable speed increase) but it will break SCSI expansion because it is DMA and DMA controllers require proper MMU tables with 68040 or 68060 CPUs.

---

Can Thor's lib package a solution for our problem?http://aminet.net/package/util/sys/Mu680x0Libs

im not sure ....

you still got that lc060 mike?

probably wouldnt work,but what the hell..

Quote:

---

Originally Posted by AMike

Can Thor's lib package a solution for our problem?http://aminet.net/package/util/sys/Mu680x0Libs

---

Unfortunately not. I tried to add env-arc:mmu-configuration file containing "SetCacheMode from 0x00000000 size 0x01000000 cacheinhibit" which should disable caches from lowest 16M but it still hangs.

I guess this only affects data caching, not instruction caching.

nice try , shame it didn't work :-(

Quote:

---

Originally Posted by Toni.Wilen

Quote:

---

Originally Posted by AMike

Can Thor's lib package a solution for our problem?http://aminet.net/package/util/sys/Mu680x0Libs

---

Unfortunately not. I tried to add env-arc:mmu-configuration file containing "SetCacheMode from 0x00000000 size 0x01000000 cacheinhibit" which should disable caches from lowest 16M but it still hangs.

I guess this only affects data caching, not instruction caching.

---

:(:( Thanks for testing.

sysinfo on the 060 with the ACA

a quick pic before it reboots..

I found a solution, it is quite ugly hack but it does work, 68060.library works, FPU works. Sysinfo won't crash anymore and results are slightly faster than previously.

Replace your libs:68060.library with this one: http://www.winuae.net/files/b/ACA500...d1260_Hack.zip
(DO NOT USE if you don't use it with ACA500!)

It is Phase5 68060.library patched to not clear ITT0 and DTT0 registers. (Transparent translation registers that ACA500 uses to mark lower 16M as uncacheable for both data and instruction), these registers have priority over MMU tables.

Note that most likely any debugging program (like enforcer) won't anymore work with chip ram (or any ram inside first 16M), also this hack won't work if any program touches ITT0 or DTT0.

fantastic , copying it over now...

standby :-)

YES!!!! your a hero my friend :-)

working , although it says , no 68060 upgrade now! when it boots lol

but here we are....

Sysinfo working
which amiga still crashes tho
info shows cpu68040

youtube vids uploading now.

http://youtu.be/Kam226mb0PI