Running X windows on a 16MHz 386sx

Kaleb Keithley kaleb at thyme.jpl.nasa.gov
Wed Oct 17 01:50:20 AEST 1990


In article root at dlb.uucp (0000-Admin(0000)) writes:
>In article tim at maths.tcd.ie (Timothy Murphy) writes:
>>In bill at unixland.uucp (Bill Heiser) writes:
>>
>>>I doubt that a 16/20 MHz 386SX will be sufficient to run Xwindows,
>>>although it may be OK for normal apps.
>>
>>I read recently a claim that a 386SX with lots of memory
>>made a perfectly adequate Unix box (which I took to include X).
>>
>>It would be nice to hear from someone that has actually done this,
>>in the real world.
>
>I would be very surprised if reasonable performance resulted from such a setup.
>I am convinced that X could run, given 8 Meg of RAM, but it would be annoyingly
>slow.  At 4 Meg of RAM, there would probably be severe swapping problems.

I've done it in the real world, using ESIX 3.2.C X11R3 on my 16mhz SX with 
both 4 and 8 meg.  Four meg works, it's not pretty; every time you drag the
cursor between windows the drive access light comes on as the thing starts
to swap.  With 8 meg it doesn't spend as much time swapping.  I think the 
real bottleneck tho' is the VGA display on the 8mhz AT bus.  After a year of 
living with the SX, I just upgraded to a 486 and, while all around faster,
the display is still the weak link (it is a 16bit VGA.)  What intrigues me 
the most is that the display performance for OS/2 is orders of magnitude
better; the X server writers need to get a lot smarter about handling the
VGA, especially bit blitting.

BTW, I've heard that ESIX 3.2.D is substantially better in many areas
including the X Server, I'm waiting for my update to arrive.

-- 
Kaleb Keithley                      Jet Propulsion Labs
kaleb at thyme.jpl.nasa.gov

causing trouble again.



More information about the Comp.unix.sysv386 mailing list