16 May
2007
16 May
'07
9:28 a.m.
On Tue, 2007-05-15 at 13:24 -0700, Chuck Guzis wrote:
In particular, I recall playing around with a memory
test that
gradually increased the interval between refresh until memory started
failing. I began to wonder if there was something wrong with my test
when some NEC uPD416 DRAMs went 15 seconds without refresh and
*still* retained data. Nothing wrong with the test--just good DRAM
design.
I remember with the ZX Spectrums, if you powered them off and on again,
quite often you could still see the screen contents quite clearly before
the RAM test flattened it. Some RAMs seemed better than others for this
- some would keep the image almost perfect for maybe five seconds, some
wouldn't do it at all. I found you could spot which RAMs were likely to
fail by seeing which pixel dropped out the most - work out where the
black stripe was when it was working, and if the low 16K failed later
that would be the first one to change!
Gordon