From: roell@informatik.tu-muenchen.de (Thomas Roell) Subject: Re: Warning about Diamond Speedstar 24 Date: 1 Aug 1992 08:51:34 GMT
Well, sometimes, but only sometimes I'm sad about people that connot
read and understand. Thus here the final clarification about the DS24
and HighColor:
The highcolor switch was a bad (and unofficial) hack. The people that
did it didn't really understand the issues, but got a workable
solution for the problem. Basically newer ET4000 boards faced the fact
that only 8 different dot-clocks are to fee for all modes. The
HighColor RAMDACs made that obviouse. Thus they used a trick. All
frequencies that are normally used are doubled. For the normal modes a
bit in the ET4000 is set and the frequencies will be divided by two.
This was great, since the HCs wanted the double speed for the 16bpp
modes. Anyway, X386 forced this divide_by_two bit to be 0 (i.e. no
divide), and assumed that clock 0 was a 25.175 MHz clock. As written
above in reality there was a 50 MHz clock there. All what this
'HighColor' keyword did was doubling the clocks that X386 detected
(instead of fixing the real problem).
The ultra new boards like the DS24 have a new problem. Guess what ?
Yeah, 16 clocks are not enough. This time they made it right. They
choose a programmable PLL syntesizer. clock0 is a 25.175MHz, clock 1
is a 28.322MHz frequency. Clock 2 is reserved for an external input
(via feature connector) and Clock 3 is freely programmable. The only
problem is that you have to know which PLL is there, and how to
programm him. That's all.
Hope this clears up things and will avoid unecessary discussions and
some black magic ;-)
- Thomas