Super Gold Card (Clone)

[Dave]

Because of that system and CPU clock may be different from 24 MHz.

Have you tried any other clock speeds? What worked?

[Cristian]

Excellent job Tetroid! Thank you very much!
Now I'm so curious about the experimental board...

[kordics]

I'm glad I will get one of your SCG clones! I'm also interested to know few more details about the "blue" SCG card. Does it mean that It can run faster than 24 MHz? If so, how much faster?

[tetroid]

Because of that system and CPU clock may be different from 24 MHz.

Have you tried any other clock speeds? What worked?

I tried several clock generator.

25 MHz - works
32 MHz - started and crashed after several seconds.
20 MHz - works
16 MHz - started but dont work properly, have video issues in HiRes mode.

I have no generators from 25,xxx to 30 MHz, very interesting to check it.

I think I naad to make the new thread in Hardware about my SGC investigations. such as clock generators, old and new EP1810 chips and CSYNCL problem.

[Derek_Stewart]

I notice the blue SGC has a slightly different layout to the red and black ones. Is it an earlier or later revision? What caused the change?

The blue board - experimental.
The differences:
2 clock : 24 MHz oscillator for floppy disk controller and other - system clock generator. Because of that system and CPU clock may be different from 24 MHz.

Hi,

Does the QL Network work on the experimentals boards.

Some of the old Miracle SGC cards could increased in clock but the Network woukd work due timing issues.

[tetroid]

I notice the blue SGC has a slightly different layout to the red and black ones. Is it an earlier or later revision? What caused the change?

The blue board - experimental.
The differences:
2 clock : 24 MHz oscillator for floppy disk controller and other - system clock generator. Because of that system and CPU clock may be different from 24 MHz.

Hi,

Does the QL Network work on the experimentals boards.

Some of the old Miracle SGC cards could increased in clock but the Network woukd work due timing issues.

I am sorry, I didn't try this.

[Dave]

*grins*

Do you have 5x7mm standard oscillator SMD pads under that DIP component?

What I think you will find is that the CPU will scale nicely until the INGOT falls over - it was already very close to its limit. Putting a decent heatsink on the INGOT and 68EC020 would maybe buy another 2MHz of headroom.

In my experience, you should be able to keep it unconditionally stable up to 28 or 29 MHz in all cases, IF you have good quality EP1810LC-25T.

However, performance will not scale linearly. There will be a magical place where the CPU starts waiting for the memory. I think that is around 28MHz also, depending on the speed of the memory.

Another way to get a bit of speed is to replace the 68EC020 with a 68EC030. It has the same bus, and should work with the INGOT with no changes. HOWEVER, one of the problems with the INGOT is that it uses trickery, deception and magic to work. That whole "copy the ROM into RAM, patch it, write protect it then relaunch the OS from that patched copy in fast RAM" thing means that all DRAM handling goes through the INGOT. The INGOT also does DRAM refresh, based off the CSYNCH or VSYNC signal (I forget which)... Which is something that could have easily been offloaded to release a LOT of LUTs and a couple of pins. There's just no need at all for any DRAM handling to be managed by INGOT once we're passed the QL's original 1MB address map.

It does strike me that half the reason for INGOTs existence at all is the legalities of having to do the ROM copy and patch. If they had simply licensed Minerva and included it pre-patched the double-boot and etc would have been unnecessary, and the INGOT would have been a LOT simpler.

That would be a good place to start for a future card.

But then, we have to remember that this was done in the late 80s and early 90s, with very limited development resources, a tiny budget, and the skills of only a very small group of people. The fact it happened at all is amazing to me. I am very grateful.

[tetroid]

*grins*

Do you have 5x7mm standard oscillator SMD pads under that DIP component?

No, only pads for big SMD oscillator.

[mk79]

It does strike me that half the reason for INGOTs existence at all is the legalities of having to do the ROM copy and patch. If they had simply licensed Minerva and included it pre-patched the double-boot and etc would have been unnecessary, and the INGOT would have been a LOT simpler.

How would you have attached ROMs to the 32-bit bus of the 68020 that are as fast as the RAM? Also you would have lost SMSQ/E compatibility. The SGC didn't need a double boot by the way.

Marcel

[Dave]

How would you have attached ROMs to the 32-bit bus of the 68020 that are as fast as the RAM? Also you would have lost SMSQ/E compatibility.

There's lots of 5V 32-bit flash that's very fast. Any 40 ns flash will keep up. There's a 64Mx16 flash which, if two were used, it could be used for the OS and 127MB of fast storage. That flash would work to 29MHz.

SMSQ/E is a living operating system that is currently enthusiastically maintained. It would be able to adapt. With flash, it could be natively booted without any other OS involved. It's just a matter of knowing what the memory map looks like to know where SMSQ/E needs to be mapped in, and the pointers to it.

But that's speculative, since this obviously isn't the case, and we're talking about imaginary hardware.