Open source 14MHz 68000 accelerator

[hooverphonique]

Turning "off" the original version (i.e. slowing it down to ~OE speed) is implemented by disconnecting it from the local ram, IIRC.
What sort of speed advantage does the SE version offer without local ram?

[TheGouldFish]

Originally Posted by Mathesar

Hi GouldFish!

First things first: You are the first that I know of that has built the "SE" version.
Can you confirm that it works indeed with the DMA SCSI controller of the GVP HD8+?
That would be good to know because that is what this version was supposed to do.
As for the crashing in Operation Stealth, it could be that your 68000 does not like to be overclocked so well. You could try to source a (faster) HC part.

But congratulations on your build. Looking good

I'll have to lookup how to specificity test the DMA side of the GVP.
so far I've just been using the machine as normal.


but I've noticed a few things, if I boot it with out the GVP and run SysInfo I got a slower speed (of around 11mhz).
but running Operation Stealth from floppy disk with the GVP disconnected worked.


I've also tested a few more games and had mixed experiences.
with more of them crashing.
This might indicate issues with the DMA (I tried changing the max transfer speeds for the partitions to see if that was the issue, and it didn't affect it in anyway).


My next step will be to run running op-stealth from floppy with the drive connected.

[TheGouldFish]

Just tested it running from the floppy drive, while the HD8 is connected.
and it having booted workbench from the drive.
and it seems to run fine.

I'll look at trying to source a HC 68000.
but it does look like running games from the hdd is the issue.

[TheGouldFish]

Tried it out with a HC cpu and it ran a little better.
but it was still reporting 12.50mhz and some games still crashed.

so not sure if its the design, the DMA transfer or if I've built it wrong.

[Mathesar]

I think you build it right. the fact that a 68HC000 makes it work better makes me believe the problems are related to subtle timing issues with the DMA bus request signals.
That it works at all and that you are able to boot from the HDD means that in general things are working.

I'll think it over during the weekend :-) Maybe there are some things we can try.

[ItsTheSmell]

I had a go of building ver 2.2 yesterday and found it really difficult. do the Cx resistors have to go on a specific way? 2 of them say 222 on but the bag they were in says 22p, will they be ok?

edit: The photo on github says it should be 222 so thats that answered. Do I need to bridge the JPx joints?

[Mathesar]

Originally Posted by ItsTheSmell

I had a go of building ver 2.2 yesterday and found it really difficult. do the Cx resistors have to go on a specific way? 2 of them say 222 on but the bag they were in says 22p, will they be ok?

edit: The photo on github says it should be 222 so thats that answered. Do I need to bridge the JPx joints?

The last digit of a resistor code tells how many zeroes should be appended to the first 2 digits to get the resistance in ohms. Thus, 22 + 2 zeroes = 2200 Ohms = 2.2kOhms = "2K2". Resistors have no polarity so you can mount them either way.
and yes, you should bridge the JPx (Jumpers) joints.
If you find the build difficult it might be worthwhile to first build the version without fast RAM. The jumper settings (solder bridges) are shown on the bottom of the board. Then, if everything works you can add the fast RAM and associated components. Remember to change the jumpers then.

[ItsTheSmell]

Ok mate thanks for the info.
I have the version that allows passthrough to the zorro bus for a Supra Ram 500XP, ill build it next week if this one works. I think i need a better soldering iron; my one is from maplin and it leaves cold joints when set to maximum.

[keithamiga1200]

22p (Capacitor)
222 (Resistor). 222 is 2k2/2.2k = 2000 and 200 Ohms.(As mentioned by Mathesar)

Resistors are numbered, capacitors are coloured.

I would first clean the PCB with alcohol, then use the thick/cloudy flux, rather than the thin clear stuff. Don't worry too much about cloudy solder joints, as there's never going to be any kinetic movement on the smd components. The pin headers will require good soldering joints tho.

ps; Remember, leaded solder and non-leaded solder require different temperatures. Which might explain your dull/cold joints.

[ItsTheSmell]

So, I've treated myself to a fancy new AiXun T3A iron. I was planning on building the SE board but I can't find any pictures to follow. Nothing on GitHub. Anyone got a link to one?

[Mathesar]

Originally Posted by ItsTheSmell

So, I've treated myself to a fancy new AiXun T3A iron. I was planning on building the SE board but I can't find any pictures to follow. Nothing on GitHub. Anyone got a link to one?

There was someone in this thread who built the SE but he posted no pictures of the assembled board. I also don't have pictures.
His board worked but it didn't solve the DMA issue with SCSI side-cars like the A590.

[ItsTheSmell]

In that case is there anyway i can buy the ram chips for the standard accelerator?
My SupraRam 500XP doesnt always work with some games. I have to switch it off before booting. How compatible is your accelerator?

[Mathesar]

The full accelerator is very compatible as long as there is no DMA SCSI device hanging on the side. So, as long as the full accelerator is installed, you cannot use the supraram's harddisk at all.

[Jope]

The SupraDrive 500XP doesn't do DMA, but it does need working autoconfigure.

[ItsTheSmell]

If the accelerator works ill sell the supraram. However, i cant find anything on ebay. Help!

[Mathesar]

Originally Posted by Jope

The SupraDrive 500XP doesn't do DMA, but it does need working autoconfigure.

Then both the normal and the "SE" version of this accelerator should work.

[girv]

Is there anyone selling these ready built?

[neongod]

I've built the SE version (rev 3.0), to test it with my A590 (SCSI HD and 2MB RAM).

Here are my experiences:

- The build was fairly easy and even with beginner soldering skills, it took just a few hours.
- The card with the CPU installed fits under the metal shielding, which is nice
- JP1 -> I connected the 1-2 (int. default) option.

- Without the A590 the card works, even though with my MC68000P8 SysInfo4 shows it to be 11.6MHz and the performance is slower than on 7MHz with the A590 (0.60 vs 0.75 MIPS). Can be a timing issue?
- I'm not sure if it's just my build or my cpu is not tolerating overclocking (the more likely), but it becames unstable after a few minutes and the system freezes. I'm trying to source a 68010 or another, higher rated 68000 to give it a test.

- With the A590 plugged in however, it doesn't work. As soon as it tries to boot from the hard disk, it crashes. I guess the DMA issue is still present.

Mathesar: any idea if it's possible to fix the DMA problem somehow?


Some pics:

SE card built:



SE card with CPU installed:


SysInfo with the SE card at 11.60Mhz, no A590:



SysInfo without the SE card, 7MHz, with A590:

[neongod]

Originally Posted by neongod

with my MC68000P8 SysInfo4 shows it to be 11.6MHz and the performance is slower than on 7MHz with the A590 (0.60 vs 0.75 MIPS). Can be a timing issue?

And then I of course realized that it's the Fast Ram that's making the difference with the A590.

[Mathesar]

Hi Neongod,

Yes, the fast ram is making the difference. With fast ram it should be a bit faster than with a standard 7MHz CPU.

I have a theory why it still doesn't work with the A590. although the accelerator should properly tristate all signals related with DMA, there might still be a phase issue with the E clock. Do you have access to a scope?