Re: In search of bad 4164, 41256 DRAM

On 13/10/2019 22:46, David Roberts wrote:

> One part of the test is an addressing error test. Another part is a
> 'stuck at' test. The remainder of the tests generally test for
> interaction between bits.

As the causes of dram failure are complex, the tests mostly brute force
as many different accesses as possible.

Bonding wires can break and that can introduce easily detectible errors,
but dram was never perfect & you could get a random flipped bit
(especially if you used contaminated raw materials like intel).

http://the-eye.eu/public/Books/Electronic%20Archive/Soft_Errors_in_Modern_Electronic_Systems.pdf

The memory that commodore used was carefully chosen on price to have the
bare minimum of reliability that would be required for a cheap home
computer system. A flipped bit after a thousand hours wouldn't worry
them at all.

As the dram ages that unreliability just gets worse. We on the other
hand have gotten more picky.