exxos's Atari, Amiga & retro forum

@Badwolf So that locks up on the same test with caches turned off as well. I just found a different CPU so will try that just in case something simple for once...

EDIT:

Nope. It did get a little further the test though..

Maybe burst mode on the RAM needs the caches to function correctly ? @terriblefire

EDIT2:

Nevermind, I think it's because ST RAM was running slightly to fast..


EDIT3:

I wonder if the gain only been like 50% is just literally because a lot of it is graphics related routines, in that the CPU would be slowing down continuously to update screen memory... And soldering alt-ram and ROM itself are like 800%, the CPU has to be slowing down considerably to do something, EG accessing ST-RAM a lot. So wonder if at some point I ought to have some other tests to benchmark alt-ram with some math functions or something...

Sorry, I had to go and do something else for a bit. I had one of those 'it was working perfectly and then it blew up' moments. I'd just got the entire card working beautifully and had just changed out the CPU for a 50MHz version for some benchmarking and... my brand new £20 CPLD goes. Half the pins shorted to ground.

F*** it!

Perhaps that unnecessary rework this morning pushed it over the edge, or perhaps there's some partial contact scenario that's fatal to them? The Falcon expansion connector sometimes needs reseating a few times.

Dunno. Third one I've had fail that way.

exxos wrote: 02 Apr 2021 14:31 Nevermind, I think it's because ST RAM was running slightly to fast..

That's cured the locking up during VDI draw without cache, you mean?

BW

Badwolf wrote: 02 Apr 2021 18:54 Dunno. Third one I've had fail that way.

Assuming you are using a ground plane and enough decoupling caps ? Also some of these chips tend to blow up if you have a bad power connection on one of the pins as well :roll:

Badwolf wrote: 02 Apr 2021 18:54 That's cured the locking up during VDI draw without cache, you mean?

Yep! :)

Now I am trying to work out why the board only works with a fan blowing on it :lol: :roll: So far with the help of freezer spray, it looks like it's the PLD itself :roll: though I have been contemplating moving down to the 144xl As they offer high-speed grades...

@exxos you can get the 95288s as 200mhz if I remember correctly.

PhilC wrote: 02 Apr 2021 19:34 @exxos you can get the 95288s as 200mhz if I remember correctly.

The ones thats on is 10ns.. but the price jumps up to like £40 a pop IIRC... The 5ns 144 is the same price as the 10ns 288 or something like that.

@exxos but will it fit in a 144?

PhilC wrote: 02 Apr 2021 19:44 @exxos but will it fit in a 144?

52% currently, so If I tidy up my code and delete some crap out of it then maybe...

It can be done...

Thats pretty cool as the 144 is so much cheaper

@PhilC Didn't the original TF536 stuff fit in a 144 for the ST?
There was some mention of keeping the 288 in case there was a "man in the middle" fix for something (possibly IDE with TOS? :shrug:)