Very Computer

Some of you may have already encourtered the following problem :

on my SunBlade 2000 and SunUltra 60 workstations, equiped with
20" CRT displays, I experience sporadic vibrations of the whole
image on the screen. The vibrations consist in horizontal
deviations of the odd lines relatively to the even lines.

For examples, and 6 x 6 pixel area of the screen that should
look like :

aaaaaa
bbbbbb
aaaaaa
bbbbbb
aaaaaa
bbbbbb

(where "a" represents a pixel on an odd screen line and "b"
an pixel on an even one)

rapidly (with a > 10 Hz frequency) varies between :

 aaaaaa
bbbbbb
 aaaaaa
bbbbbb
 aaaaaa
bbbbbb

and

aaaaaa
 bbbbbb
aaaaaa
 bbbbbb
aaaaaa
 bbbbbb

I would like to insist on the fact that this problem happens
on the whole screen at a time (the 6x6 example was just to save space
on this message).

By trial and error, I discovered that I could reproduce this
flickering by writing a simple program that allocates chuncks
of memory of a fixed size. An implementation in pseudo-C would be :

while (1)
    malloc(1000 * sizeof(float));

When I launch that program, the flickering really prevents
anyone from working on the machine, as the amplitude of the
horizontal bias from the two extreme positions shown above
nearly reaches 1 millimeter.

I have launched my program on several Sun Ultra 60 stations,
and this problem is perfectly reproductible.

So, I would like to know if some of you have already been
through this, and if you found an explanation. Sun would
not admit that it's a bug of their motherboards. But I would
think it is.

Thank you for your answers !

Jery P.

Quote:>on my SunBlade 2000 and SunUltra 60 workstations, equiped with
>20" CRT displays, I experience sporadic vibrations of the whole
>image on the screen. The vibrations consist in horizontal
>deviations of the odd lines relatively to the even lines.

        specific hardware configuration of the U60/SunBlade 2000 systems
                - framebuffer type
                - make model of monitor

        specific software configuration
                - OS revision & relevant patches (core OS, X server,
                  framebuffer)
                - framebuffer configuration (fbconfig -prconf)

I've certainly send both my old U60 and current SB2000 through their
paces and never seen anything remotely like what you are seeing.

Casper