Archived from groups: comp.sys.ibm.pc.hardware.video (
More info?)
> >No such animal as a dead pixel due to a bad video card.
>
> I saw and repaired this *exact* problem on a monochrome VGU card in a
> minicomputer CAD workstation some 15 years ago. This card had two 16KB
> bitmaps consisting of 16 x 4116 DRAMs (16Kx1). The chips cost ~$1
> each, so the major cost was my diagnostic time. BTW, the card cost
> ~$10K.
I figed my Gravis Ultrasound 512 (+512) using the same methods, back then
the memory wasn't soldered and much less hassle to switch. I salvaged the
memory from old VGA card, actually but that is besides the point.
Memory is not in fixed address in modern OS and hardware compo, because the
memory is allocated for surfaces as it is needed by the memory controller in
the card. So the address of memory that is defective could end up storing
practically anything. If it happens to store indexbuffer, and using index
out-of-range to cause the graphics processor or it's memory controller to
read or write on top of something else it relies on, such as vertex- or
fragment program (or anything I cannot think of off-the-hip) causing the
hardware to go into endless loop or otherwise crash, freeze, anything.
Such product would be unstable and untrustworthy, that is the bottom line.
Dead pixel on LCD display is more likely to be only a cosmetic defect.
Defective memory on the other hand opens a completely new can of worms. I
just had unstable system because of defective DDR dual channel memory kit,
it was wonder if the system stayed up 15 minutes at a time. GPU != CPU+MB
but it's hardly comparable to a "dead pixel" on LCD either.
Memory manufacturers don't have the luxury of not having a full warranty for
their products: the yields are too good for that with the cost of producing
a wafer. Can someone quote typical % of waste for TFT/LCD displays assuming
1280x1024 and zero dead pixel tolerance? Any manufacturer will do for
starters, I'm too handicapped and stupid to find such information myself
(okay, didn't even try to look.. if anyone has information at his fingertips
please share, thanks!). Obviously there is TOO MUCH waste for some
manufacturers liking if they have policy of X dead pixels, say, 3-5 per
display. If they had zero tolerance I bet the prices would have to be
higher, and if they keep prices too high I would assume their displays
wouldn't be so "competitive" .. I think a "defect" in their products most
consumers don't even know is there is better than obvious price defect that
would be the alternative (price defect == higher, thus non-competitive
retail price).
IMHO, I think this situation is direct result of price competition. People
are willing to pay for Good Enough as long as it is also Cheap Enough. I
could be wrong.
Or am I just stupid to come to such conclusions?