Blizzard 1230-IV 128mb instability

[pipper]

Hi,

I just recently got back to my A1200 (200w AT PSU)after a long period of absence. I wanted to update the memory to 128mb (why not...) but I'm facing lots of instability.
It looks like a piece on the Blizzard is causing trouble when the card gets warm. The instabilities are severe, causing crashes all over the place...

I tried several modules:
16mb 60ns (doublesided): fully stable in 70ns setting. In 60ns gets unstable after a while when the CPU reaches around 50c. Memory testing programs don't show a problem, even in warmed up state.

128mb 60ns (doubesided): gets unstable after warmup. Memory tests do not reveal a problem, though.

128mb 50ns (doublesided): gets unstable after warmup. After warmup, memory tests show a few addresses (they are pretty consistent) that have values which differ in bit 3. These failures are revealed by 'MemCheck' ADR test where the test writes the memory address as value.
So for instance, address 0x7c0c9108 would get the value '0x7c0c9108' assigned, but when reading it back, the value '0x7c0c910c' is returned. A few memory addresses might exhibit such behavior, 0x7c0c9108 being the one that shows up consistently .

64mb 60ns (singlesided): fully stable under all conditions.

So, it looks like something on the accelerator board is causing these issues - and I'm not convinced its the memory modules that are at fault... particularly as the the problems arise only after warmup and not immediately after startup.

Does anyone have an insight on what to lookout for on the board? There are no bad solder joints that can be seen by the naked eye. Is there anything different in the addressing hardware between the single sided and double sided modules?
Schematics would help tremendously :-)


thanks!

[Estrayk]

IIRC, the Blizzards are just stable with 5v simms. 3.3v works but with problems.

[meynaf]

Double sided modules, huh ? Normally they don't fit well in a B1230. I had one and it randomly failed because of this (too thick). When putting a simm inside you should hear some click when it's in place ; if you don't then you're in trouble.

In addition some may have problems with 70ns and not with 60ns or vice versa ; 50ns boards may work or not. There it's just timing issues.

[pipper]

How would I differentiate 3.3V from 5V modules?

The modules all physically fit. The Blizzard can take double sided modules just fine. What counts is that they are no taller than 1".

[cpiac64]

net connector 150 pin

[pipper]

I cleaned the trapdoor slot and the Blizzard's trapdoor connector beforehand. The fastmem on the Blizzard should have (almost) nothing to do with the expansion slot, except for power supply.

[pipper]

I think, Estrayk hit the nail on the head. The module was sold to me as 'for Blizzard 1230, 1240, 1260'. But when I look up the chips' specs on the module, they show up as 3.3v.
They get hot pretty quickly and problems arise, once they're hot.

[pipper]

I got a 128mb 5V replacement module. The chips on the module don't get hot anymore and memory testers don't show any problem. Yet, with 128mb the Blizzard quickly gets unstable, with wild crashes all over the place. With 64mb it runs rock solid.
I'm afraid it has something to do with addressing single-sided vs double sided modules.
I haven't figured out yet what the particular difference is between 64mb modules and 128mb modules. Is it a certain address line that is working differently? Is it any of the RAS or CAS signals that may not be used with the 64mb module but is causing trouble at 128mb?

[dlfrsilver]

check also that you don't have any problem with your OS and programs.