Re: TED badlines, how do they work?

From: Segher Boessenkool <segher_at_kernel.crashing.org>
Date: Sun, 4 Sep 2011 19:40:57 +0200
Message-Id: <0CA5B4F3-F198-4FDF-8EDD-125EED9DADCB@kernel.crashing.org>
>> It turns out the (internal) clock drivers and a few other things  
>> use the
>> 12V
>> supply on the 6569 (those use the normal supply on the 8565).  
>> Those can
>> sink
>> a lot of current I bet ;-)
>
> Makes me wonder why they did that. Needed +12V for the dotclock and  
> color clock to work properly?

Higher voltage makes faster circuits; for clocks and other big fan-out
signals that can be important, so that edges stay steep.

I finally figured out one other weird thing: near the RAS and CAS pin
drivers there is a huge FET, with quite a bit of logic from all over
the place leading to it -- but it doesn't drive anything.  On the 6569,
you can see where it would have connected: the enable signal for the
address pins.  And that makes sense for the logic too, it seems it  
enables
this driver when it is actually fetching anything (not a "3fff" cycle).
I guess on the 6566 it didn't drive the address bus when it wasn't  
actually
fetching anything; on the 6569 though, the logic for RAS/CAS etc. is
already complicated enough, so it was decided to just drive address
always

> The power comsumption on +12V helps explain why the 6569 runs so hot.

400mW...  That's not _that_ much is it?  You can do about 2W without
heatsink (depends a lot on your package of course).

> Without a heatsink you'll burn your fingers on it. Still, it  
> doesn't seem to mind. TED and the 8501 don't get as hot by a long  
> shot but are still more prone to die. :(

Would be interesting to know how it dies.  You can often see it under
a microscope, hint hint hint :-)

Heat alone isn't such a big deal, but it certainly can be a contributing
factor to untimely chip death.  There must be something worse as well
though.


Segher


       Message was sent through the cbm-hackers mailing list
Received on 2011-09-04 18:00:20

Archive generated by hypermail 2.2.0.