Re: VIC-II DRAM refresh

On Fri, Oct 14, 2016 at 12:21 PM, Gerrit Heitsch
<gerrit@laosinh.s.bawue.de> wrote:

>>>
>>> I'd change the failed chip, so far only the MT4264 have shown that 'if
>>> one
>>> fails, others will follow' pattern when it comes to RAMs.

this has not (yet) happened to me. I changed 1 ram chip in year 2001
or 2002 and that C64 it's still working fine. I remember it because it
was the same collegue that sent me some few C64 last year to repair
and there wasn't that one that I already fixed in 2001 (but I didn't
record the brand of the failed chip, nor I do have pictures of it),
however there was a new C64 with 2 x MT failed out of the 8.
I now have the one of the video with 1 x MT failed. I will use it from
time to time, so if I live enough, I can see if some of the other chip
fails and how much it takes :)


> I don't keep records that detailed, but I'm also active on forum64.de and
> for the last years if there was a question about a C64 and it was bad RAM,
> the wast majority were MT4264. As I said before, they don't seem to age very
> well.

I agree that the vast majority of failed chips are MTs, but my other
question is, how many MTs vs other brands were actually installed by
Commodore factories to begin with? I have 2 boards with MT4164, so 1
failed out of 16 chips so far and 2 boards one with Fujitsu (none
failed) and one with Samsung (none failed) chips (not counting the
repairs for others, since I didn't record the good chips, only a
couple of failed MTs).
Too bad I didn't record the failures in the '80s.
Fujitsu chips (by memory) seems the second most used brand in C64s but
that's me noticing them since the epoxy is darker and shiny and they
look nicer than other brands., I've seen a few boards with Samsung
too.

>
>> If we see that 7 chips out of 8 works fine after so
>> many years, what we conclude?
>
>
> Not much since the MT4264 have been working fine for years and only lately
> started failing and when they fail it's a consistant pattern of not the
> whole chip failing but a number of cells.

that's almost always how DRAM chip fail. Maybe the bit capacitor
insulation gets leaky, would be interesting to know, but decapping a
chip and analyzing it is out of my possibilities.

>> I've had one failure on a regular 7406 that I can remember... pretty
>> reliable against the 7707 failure records :)
>
>
> I had a few 7406 where the driver that has the 180 Ohm Pullup went bad.
> Which means the 74LS257 don't come out of tristate and the CPU doesn't get
> to talk to RAM or VIC. They were all made by National Semiconductor.

Same, NS 7416 (low voltage version of the 7406, but perfectly in specs
on a C64) is the one failed that I recorded.
By the way, if statistics were to be taken seriously, we should never
leave any MOS ROM into CBM equipment :)
My collection of failed chips has quite a few of them, only CIAs and
generic TTL equivalents are more than them :)
Frank

       Message was sent through the cbm-hackers mailing list