[Sunhelp] Sol8 on SS5-110

[James Lockwood]

On Thu, 1 Jun 2000, Christopher J Ceska wrote:

> bottom line, the SX is faster than any SBUS buffer at any
> resolution/color depth.  There is documentation from sun and third
> parties that back this up.  SBUS == slow.

The SX is not faster in 24bpp mode than a cg6 is in 8bpp mode.  Xstone
results and empirical testing of my dual-headed SX/TGX bear this out.

The theoretical maximum bandwidth to the SX is 4x greater than that to a
(32-bit) sbus based framebuffer, but SX has to move 4x as much data in
24bpp mode. Add to this the fact that more data must be pumped in from
main RAM and it is likely to be rougher on the caches and you have an
overall slowdown. The SX is also impacted negatively by mbus latancy since
it is noncacheable, this can be seen by comparing raw bitblitting
performance of a SS20/50 and a SS20/61 (the 50 will win).

That said, the SX is definately the fastest 24bpp framebuffer for
pre-Ultras.  Comparing it against 8bpp cg6 cards isn't really fair.

> anyone using cde that isnt forced to by a university or college should
> be zapped by a tazer.

CDE isn't just a wm.  As a wm it is marginal, when you count the other
integrated features it still has some use.  I'd be much happier if Sun
upgraded to CDE 2, though.  The improvement in the integrated printing
facilities alone would be worth it.

-James