XF86_Accel(1) XF86_Accel(1)
NAME
XF86_Accel - accelerated X Window System servers for UNIX
on x86 and Alpha platforms with an S3, Mach8, Mach32,
Mach64, P9000, AGX, ET4000/W32, ET6000, I128, TGA, or
8514/A accelerator board
SYNOPSIS
XF86_S3 [:displaynumber] [ option ] ...
XF86_S3V [:displaynumber] [ option ] ...
XF86_Mach8 [:displaynumber] [ option ] ...
XF86_Mach32 [:displaynumber] [ option ] ...
XF86_Mach64 [:displaynumber] [ option ] ...
XF86_P9000 [:displaynumber] [ option ] ...
XF86_AGX [:displaynumber] [ option ] ...
XF86_W32 [:displaynumber] [ option ] ...
XF86_I128 [:displaynumber] [ option ] ...
XF86_TGA [:displaynumber] [ option ] ...
XF86_8514 [:displaynumber] [ option ] ...
DESCRIPTION
XF86_S3 is an 8-bit PseudoColor, 16-bit TrueColor and
32-bit TrueColor server for S3 graphic accelerator boards.
Note, 16-bit and 32-bit operation is not supported on all
S3 accelerator boards. Refer to README.S3 for details of
which boards are supported at which depths.
XF86_S3V is an 8-bit PseudoColor and 16-bit TrueColor
server for S3 ViRGE and ViRGE/VX graphic accelerator
boards.
XF86_Mach8 is an 8-bit PseudoColor server for ATI Mach8
graphic accelerator boards.
XF86_Mach32 is an 8-bit PseudoColor and 16-bit TrueColor
server for ATI Mach32 graphic accelerator boards. Note,
16-bit operation is not supported on all Mach32 accelera
tor boards. Refer to README.Mach32 for details of which
RAMDACs are supported at which depths.
XF86_Mach64 is an 8-bit PseudoColor, 16-bit TrueColor, and
32-bit TrueColor server for ATI Mach64 graphic accelerator
boards. Note, 16-bit and 32-bit operation is not sup
ported for all RAMDACs. Refer to README.Mach64 for
details of which RAMDACs are supported at which depths.
XFree86 Version 3.3.5 1
XF86_Accel(1) XF86_Accel(1)
XF86_P9000 is an 8-bit PseudoColor, 16-bit TrueColor, and
32-bit TrueColor server for Weitek Power 9000 (P9000)
graphic accelerator boards.
XF86_AGX is an 8-bit PseudoColor and 16-bit TrueColor
server for AGX/XGA graphic accelerator boards.
XF86_W32 is an 8-bit PseudoColor server for ET4000/W32,
ET4000/W32i, ET4000/W32p and ET6000 graphic accelerator
boards. For 16, 24 and 32 bits per pixel support on the
ET6000, try the XF86_SVGA server.
XF86_I128 is an server for Number Nine Imagine 128 graphic
accelerator boards.
XF86_TGA is an server for Digital's 21030 (TGA) chipset.
This server is available only for Alpha platforms.
XF86_8514 is an 8-bit PseudoColor server for 8514/A
graphic accelerator boards.
These are derived from the X386 server provided with
X11R5, and from the X8514 server developed by Kevin E.
Martin <martin@cs.unc.edu>.
CONFIGURATIONS
The servers support the following chipsets:
XF86_S3: 86C911, 86C924, 86C801, 86C805, 86C805i,
86C928, 86C928-P, 86C732 (Trio32),
86C764 (Trio64), 86C765 (Trio64V+),
86C864, 86C868, 86C964, 86C968
XF86_S3V: 86C325 (ViRGE), 86C375 (ViRGE/DX),
86C385 (ViRGE/GX), 86C988 (ViRGE/VX)
XF86_Mach8: ATI Mach8, ATI Mach32
XF86_Mach32: ATI Mach32
XF86_Mach64: ATI Mach64 (GX, CT, ET, VT, VT3, VT4,
LT, GT (3D Rage), 3D Rage II, 3D Rage
IIC, 3D Rage II+DVD, Rage Pro)
XF86_P9000: Diamond Viper VLB, Diamond Viper PCI,
Orchid P9000, and some clones (Weitek
P9000)
XF86_AGX: AGX-010, AGX-014, AGX-015, AGX-016,
XGA-1, XGA-2
XF86_W32: ET4000/W32, ET4000/W32i, ET3000/W32p,
ET6000
XFree86 Version 3.3.5 2
XF86_Accel(1) XF86_Accel(1)
XF86_I128: Number Nine Imagine 128 series 1 and 2
XF86_TGA: DEC 21030
XF86_8514: IBM 8514/A and true clones
For S3 virtual resolutions up to (approximately) 1152x800
are supported, using (up to) 1Mb of display memory (the
S3 uses an internal width of 1280 for some older chips,
hence 1Mb can't support 1152x900). Physical resolutions
up to 1280x1024 (1600x1280 on some cards) are possible
using 2Mb or more of display memory (virtual resolution is
dependent solely on the amount of memory installed, with
the maximum virtual width being 2048, and max virtual
height is 4096).
Similar resolutions are supported on the Mach64. Refer to
README.Mach64 for configuration details.
Similar resolutions are supported on the Mach32. For the
Mach32, the maximum virtual width is 1536, and the maximum
virtual height is 1280.
For Mach8, the maximum virtual width is 1024.
For 8514 the maximum resolution is 1024x768.
For the AGX chips, maximum resolution depends upon the
chip revision and amount of available display memory.
Refer to README.agx for configuration details.
For the P9000, the virtual and physical resolutions must
be the same. With sufficient memory, resolutions up to
1280x1024 are supported.
All the servers which support 24 bit visuals currently do
so using a 32 bit per pixel configuration where 8 bits in
every 32 bits is unused. This needs to be taken into
account when calculating the maximum virtual display size
that can be supported at this depth.
OPTIONS
In addition to the normal server options described in the
Xserver(1) manual page, these servers accept some more
command line switches, as described in the XFree86(1) man
page.
The Mach64, Mach32, S3, S3V, P9000 and AGX servers support
more than 8 bit color. The Mach32, S3V and AGX servers
support 16 bit TrueColor and the Mach64, S3, I128, and
P9000 servers support 16 and 32 bit TrueColor. The 32 bit
TrueColor mode only uses 24 bits per pixel for color
information (giving you 16 million colors). These modes
may be used by specifying the -bpp option as specified in
XFree86 Version 3.3.5 3
XF86_Accel(1) XF86_Accel(1)
the XFree86(1) manpage.
SETUP
XFree86 uses a configuration file called XF86Config for
its initial setup. See the XF86Config(4/5) manpage for
general details. Here only the parts specific to the
XF86_S3, XF86_S3V, XF86_Mach8, XF86_Mach32, XF86_Mach64,
XF86_P9000, XF86_AGX, XF86_W32, XF86_TGA, XF86_I128, and
XF86_8514 servers are explained.
Entries for the Device section in the XF86Config file
include:
Chipset "name"
specifies a chipset so the correct driver can be
used. Possible chipsets are:
XF86_S3:
s3_generic (for a standard IO driven
server)
mmio_928 (for a memory mapped IO driven
server on 86C928, 86C732,
86C764, 86C864, 86C868, 86C964
and 86C968 boards)
newmmio (for a newer memory mapped IO
driven server on 86C765,
86C868, and 86C968 boards)
XF86_S3V:
s3_virge
XF86_Mach8:
mach8 (to force the Mach8 server to
run on Mach32 boards)
XF86_Mach32:
mach32
XF86_Mach64:
mach64
XF86_P9000:
vipervlb (for the Diamond Viper VLB)
viperpci (for the Diamond Viper PCI)
XFree86 Version 3.3.5 4
XF86_Accel(1) XF86_Accel(1)
orchidp9000 (for the Orchid P9000 and many
generic P9000-based boards)
XF86_AGX:
agx-016
agx-015
agx-014
agx-010
xga-2
xga-1 (note: Only the agx-016,
agx-015, agx-014 and XGA-2 have
been tested. Refer to
README.agx before attempting to
use.)
XF86_W32:
et4000w32
et4000w32i
et4000w32i_rev_b
et4000w32i_rev_c
et4000w32p_rev_a
et4000w32p_rev_b
et4000w32p_rev_c
et4000w32p_rev_d
et6000
XF86_I128:
i128
XF86_TGA:
tga
XF86_8514:
ibm8514
XFree86 Version 3.3.5 5
XF86_Accel(1) XF86_Accel(1)
Clocks clock ...
For boards with non-programmable clock chips, the
clocks can be specified here (see XF86Con
fig(4/5)). The P9000 server no longer requires a
Clocks line. It will work the same way as other
servers with a programmable clock chip (i.e., use
the clocks as specified in the Modes). Note,
clocks over 110 Mhz are not recommended or sup
ported by the P9000 server. The Mach64 server
also does not require a Clocks line since the
clocks are normally read directly from the video
card's BIOS or programmed as required. For the
Mach64 server, the clocks given in the XF86Config
file are ignored unless the "no_bios_clocks"
option is given (see below).
ClockChip "clockchip-type"
For boards with programmable clock chips (except
with the P9000 and AGX servers), the name of the
clock chip is given. The only supported values
for the W32 server are "icd2061a", "dcs2834",
"ics5341" and "stg1703". Possible values for the
S3 server include "icd2061a", "ics9161a",
"dcs2834", "sc11412", "s3gendac", "s3_sdac",
"ti3025", "ti3026", "ti3030", "ics2595",
"ics5300", "ics5342", "ch8391", "stg1703",
"att20c408", "att20c409", "att20c499", and
"ibm_rgb5xx". Possible values for the Mach64
server include "ati18818", "ics2595", "stg1703",
"ch8398", "ibm_rgb514" and "att20c408". The only
possible clockchip value for the ET6000 is
"et6000".
Ramdac "ramdac-type"
This specifies the type of RAMDAC used on the
board. Only the S3, AGX, Mach32, Mach64, and W32
servers use this.
normal - (S3, AGX, W32) Card does not have one of
the other RAMDACs mentioned here. This option is
only required for the S3 server if the server
incorrectly detects one of those other RAMDACs.
The AGX server does not yet auto-detect RAMDACs,
this is the default if no RAMDAC is specified.
Use this option for W32 server if it incorrectly
identifies your RAMDAC or if RAMDAC detection
fails entirely.
att20c490 - (S3, AGX, Mach32) Card has an AT&T
20C490 or AT&T 20C491 RAMDAC. When the dac_8_bit
option is specified, these RAMDACs may be operated
in 8 bit per RGB mode. It also allows 16bpp oper
ation with 801/805/928 boards. True AT&T 20C490
RAMDACs should be auto-detected by the S3 server.
XFree86 Version 3.3.5 6
XF86_Accel(1) XF86_Accel(1)
This RAMDAC must be specified explicitly in other
cases. Note that 8 bit per RGB mode does not
appear to work with the Winbond 82C490 RAMDACs
(which SuperProbe identifies as AT&T 20C492).
16bpp works fine with the Winbond 82C490. The
Diamond SS2410 RAMDAC is reported to be compatible
when operating in 15bpp mode (not 16bpp). The
Chrontel 8391 appears to be compatible in all
modes.
sc15025 - (S3, AGX) Card has a Sierra SC15025 or
SC15026 RAMDAC. The S3 server has code to auto-
detect this RAMDAC.
sc11482 - (S3) Card has a Sierra SC11482, SC11483
or SC11484 RAMDAC. The S3 server has code to
auto-detect this RAMDAC.
sc11485 - (S3) Card has a Sierra SC11485, SC11487
or SC11489 RAMDAC. The S3 server will detect
these RAMDACs as a sc11482, so this option must be
specified to take advantage of extra features
(they support 16bpp, 15bpp and 8bpp while the oth
ers only support 15bpp and 8bpp).
bt485 - (S3) Card has a BrookTree Bt485 or Bt9485
RAMDAC. This must be specified if the server
fails to detect it.
att20c505 - (S3) Card has an AT&T 20C505 RAMDAC.
This must be specified either if the server fails
to detect the 20C505, or if the card has a Bt485
RAMDAC and there are problems using clocks higher
than 67.5Mhz.
att20c498 - (S3) Card has an AT&T 20C498 or 21C498
RAMDAC. This must be specified if the server
fails to detect it.
att22c498 - (S3) Card has an AT&T 22C498 RAMDAC.
This must be specified if the server fails to
detect it.
ibm_rgb514 - (S3) Card has an IBM RGB514 RAMDAC.
This must be specified if the server fails to
detect it.
ibm_rgb524 - (S3) Card has an IBM RGB524 RAMDAC.
This must be specified if the server fails to
detect it.
ibm_rgb525 - (S3) Card has an IBM RGB525 RAMDAC.
This must be specified if the server fails to
detect it.
XFree86 Version 3.3.5 7
XF86_Accel(1) XF86_Accel(1)
ibm_rgb526 - (S3) Card has an IBM RGB526 RAMDAC.
This must be specified if the server fails to
detect it.
ibm_rgb528 - (S3) Card has an IBM RGB528 RAMDAC.
This must be specified if the server fails to
detect it.
stg1700 - (S3) Card has an STG1700 RAMDAC. This
must be specified if the server fails to detect
it.
stg1702 - (Mach64) Card has an STG1702 RAMDAC.
This must be specified if the server fails to
detect it.
stg1703 - (S3, W32, Mach64) Card has an STG1703
RAMDAC. This must be specified if the server
fails to detect it. Using the W32 server you MUST
explicitly set the STG1703 as ClockChip to be able
to use the programming capabilities.
s3gendac - (S3) Card has an S3 86C708 GENDAC.
This RAMDAC does not support 8 bit per RGB mode
(don't specify the dac_8_bit option). It allows
16bpp operation with 801/805 boards. There is
currently no auto-detection for this RAMDAC.
s3_sdac - (S3) Card has an S3 86C716 SDAC RAMDAC.
This must be specified if the server fails to
detect it.
ics5300 - (S3) Card has an ICS5300 RAMDAC. This
must be specified if the server fails to detect it
(the server will recognise this as an S3 GENDAC
which is OK).
ics5341 - (W32) Card has an ICS5341 RAMDAC. This
must be specified if the server fails to detect
it. For pixel clocks higher than 86MHz the server
uses pixel multiplexing which seems to fail in a
small band around 90MHz on most boards. While the
ICS5341 RAMDAC is usually recognized as RAMDAC you
MUST set it as ClockChip to be able to use the
programming capabilities.
ics5342 - (S3) Card has an ICS5342 RAMDAC. This
must be specified if the server fails to detect it
(the server will recognise this as an S3 SDAC
which is OK).
ti3020 - (S3) Card has a TI ViewPoint Ti3020 RAM
DAC. This must be specified if the server fails
to detect the Ti3020. Note that pixel
XFree86 Version 3.3.5 8
XF86_Accel(1) XF86_Accel(1)
multiplexing will be used for this RAMDAC if any
mode requires a dot clock higher than 70MHz.
ti3025 - (S3) Card has a TI ViewPoint Ti3025 RAM
DAC. This must be specified if the server fails
to detect the Ti3025.
ti3026 - (S3) Card has a TI ViewPoint Ti3026 RAM
DAC. This must be specified if the server fails
to detect the Ti3026.
ti3030 - (S3) Card has a TI ViewPoint Ti3030 RAM
DAC. This must be specified if the server fails
to detect the Ti3030.
bt481 - (AGX, Mach32) Card has a BrookTree Bt481
RAMDAC.
bt482 - (AGX) Card has a BrookTree Bt482 RAMDAC.
herc_dual_dac - (AGX) Card (Hercules Graphite Pro)
has both the 84-pin (Bt485 or AT&T20C505) and
44-pin (Bt481 or Bt482) RAMDACs installed.
herc_small_dac - (AGX) Card (Hercules Graphite
Pro) has only the 44-pin (Bt481 or Bt482) RAMDAC
installed.
ati68875 - (Mach64, Mach32) Card has an ATI 68875
RAMDAC. This must be specified if the server
fails to detect it.
tlc34075 - (Mach64, Mach32) Card has a TI 34075
RAMDAC. This must be specified if the server
fails to detect it.
ati68860 - (Mach64) Card has an ATI 68860 RAMDAC.
This must be specified if the server fails to
detect it.
ati68860b - (Mach64) Card has an ATI 68860b RAM
DAC. This must be specified if the server fails
to detect it.
ati68860c - (Mach64) Card has an ATI 68860c RAM
DAC. This must be specified if the server fails
to detect it.
ch8398 - (Mach64) Card has an Chrontel 8398 RAM
DAC. This must be specified if the server fails
to detect it.
att20c408 - (Mach64) Card has an AT&T 20C408 RAM
DAC. This must be specified if the server fails
XFree86 Version 3.3.5 9
XF86_Accel(1) XF86_Accel(1)
to detect it.
intenal - (Mach64) Card has an internal (i.e.,
built-in) RAMDAC. This must be specified if the
server fails to detect it.
IOBase ioaddress
specified the base address for extended IO regis
ters. This is only used by the AGX server, and by
the P9000 server for the Viper PCI. For details
of how to use it, refer to README.agx and
README.P9000.
MemBase memaddress
specifies the hard-wired part of the linear frame
buffer base address. This option is only used by
the P9000, S3, Mach64, Mach32, and TGA servers
(and only when using a linear framebuffer). For
the S3 server, the hard-wired part is the high 10
bits of the 32-bit address (ie memaddress is
masked with 0xFFC00000). Note: this should not be
required for the 864 and later chips where the
entire framebuffer address is software-selectable,
or for PCI cards. Also, note that the in versions
prior to 3.1.1, the S3 server used only the top 6
bits of memaddress, and ored it with 0x3C00000.
To get the same behaviour, or 0x3C00000 with the
value given previously. For the Mach32 server,
the mask is 0xF8000000 (except for PCI cards,
where the membase setting is ignored).
This option must be specified with the P9000
server. With local bus Diamond Vipers the value
of memaddress can be either 0x80000000,
0x20000000, or 0xA0000000. The default is
0x80000000. Any value should work as long as it
does not conflict with another device already at
that address. For the Viper PCI, refer to
README.P9000. For the Orchid P9000, the base
address may be 0xC0000000, 0xD0000000 or
0xE0000000, and must correspond the the board's
jumper setting.
Note: The S3 server will normally probe for this
address automatically. Setting this option over
rides that probe. This is not normally recom
mended because the failure of the server's probe
usually indicates problems in using the linear
framebuffer.
Note: The Mach64 server requires the memory aper
ture. For ISA bus video cards, this means that
the aperture must be enabled and the aperture
address must be set to a value less than 16Mb
XFree86 Version 3.3.5 10
XF86_Accel(1) XF86_Accel(1)
(which means that, on ISA systems only, to use the
Mach64 server you must have 12Mb of main memory or
less). Normally the Mach64 server will use pre-
defined values for this address, but setting this
option will override the pre-defined address.
The Mach32 server should not require the use of
this option under normal circumstances.
COPBase baseaddress
This sets the coprocessor base address for the AGX
server. Refer to README.agx for details.
Instance instance
This sets the XGA instance number for the AGX
server. Refer to README.agx for details.
S3MClk memclk
This allows the video card's memory clock value to
be specified. This is only used for 805i, 864 and
Trio32/64 cards, and the value should not normally
be given here for cards with an S3 Gendac or
Trio64). This entry doesn't change the card's
memory clock, but it is used to calculate the DRAM
timing parameters. For further details refer to
README.S3.
S3MNAdjust M N
This allows some memory timing parameters to be
adjusted for DRAM cards. This entry is not nor
mally required.
S3RefClk refclk
This allows the PLL reference clock to be speci
fied. This may be required for some cards that
use the IBM RGB5xx RAMDACs. The value is in MHz.
For further details refer to README.S3.
Option flags may be specified in either the Device section
or the Display subsection of the XF86Config file.
Option "optionstring"
allows the user to select certain options provided
by the drivers. Currently the following strings
are recognized:
nomemaccess - (S3) disable direct access to video
memory. This option is ignored for the 864 and
964 chips.
noaccel - (AGX, P9000) disable hardware accelera
tion for the P9000, and disables the font cache
with the AGX.
XFree86 Version 3.3.5 11
XF86_Accel(1) XF86_Accel(1)
vram_128 - (AGX, P9000) when memory probe fails,
use if you have 128Kx8 VRAMs.
vram_256 - (AGX, P9000) when memory probe fails,
use if you don't have 128Kx8 VRAMs.
nolinear - (S3 and Mach32) disable use of a lin
ear-mapped framebuffer.
ti3020_curs - (S3) Enables the Ti3020's internal
HW cursor. (Default)
no_ti3020_curs - (S3) Disables the Ti3020's inter
nal HW cursor.
sw_cursor - (S3, Mach32, Mach64, P9000, AGX) Dis
able the hardware cursor.
dac_8_bit - (S3, Mach32, Mach64, AGX) Enables
8-bit per RGB. Currently only supported with the
Ti3020/5/6, Bt485, AT&T 20C505, AT&T 20C490/1,
Sierra SC15025/6, AT&T 20C498 and STG1700/3, IBM
RGB5xx (S3 server), Bt481 and Bt482 (AGX server),
ATI68875/TLC34075/Bt885 (Mach32 server), ATI68875,
TLC34075, ATI68860, ATI68880, STG1702, and STG1703
(Mach64 server) RAMDACs. This is now set by
default in the S3 server when one of the above
RAMDACs other than the AT&T 20C490/1 is used. Is
also the default for the AGX server, except for
the default VGA "normal" RAMDAC.
dac_6_bit - (S3, AGX) Force 6-bit per RGB in cases
where 8-bit mode would automatically be enabled.
sync_on_green - (S3, P9000) Enables generation of
sync on the green signal on cards with Bt485, AT&T
20C505, Ti3020/5/6 or IBM RGB5xx RAMDACs. Note:
Although these RAMDACs support sync on green, it
won't work on many cards because of the way they
are designed.
power_saver - (S3, Mach64) This option enables the
server to use the power saving features of VESA
DPMS compatible monitors. The suspend level is
currently only supported for the Mach64 and for
the 732, 764, 864, 868, 964, 968 S3 chips. Refer
to the XF86Config(4/5) manual page for details of
how to set the timeouts for the different levels
of operation. This option is experimental.
intel_gx - (Mach32) Sets the hard-wired offset for
the linear framebuffer correctly for the Intel GX
Pro cards. This option is equivalent to setting
the membase to 0x78000000.
XFree86 Version 3.3.5 12
XF86_Accel(1) XF86_Accel(1)
ast_mach32 - (Mach32) Sets special handling for
the Mach32 cards sold soldered-in on some AST
motherboards.
spea_mercury - (S3) Enables pixel multiplex sup
port for SPEA Mercury cards (928 + Bt485 RAMDAC).
For these cards, pixel multiplexing is required in
order to use dot clocks higher than 67.5 MHz and
to access more than 1MB of video memory. Pixel
multiplexing is currently supported only for non-
interlaced modes, and modes with a physical width
no smaller than 1024.
stb_pegasus - (S3) Enables pixel multiplex support
for STB Pegasus cards (928 + Bt485 RAMDAC). For
these cards, pixel multiplexing is required in
order to use dot clocks higher than 67.5 MHz.
Pixel multiplexing is currently supported only for
non-interlaced modes, and modes with a physical
width no smaller than 1024.
number_nine - (S3) Enables pixel multiplex support
for Number Nine GXe level 10, 11, 12 cards (928 +
Bt485 RAMDAC). For these cards, pixel multiplex
ing is required in order to use dot clocks higher
than 85 MHz. Pixel multiplexing is currently sup
ported only for non-interlaced modes, and modes
with a physical width no smaller than 800. This
option is also required for some other Number Nine
cards (eg, GXE64 and GXE64pro).
diamond - (S3) This option may be required for
some Diamond cards (in particular, the 964/968
VRAM cards).
elsa_w1000pro - (S3) Enables support for the ELSA
Winner 1000 PRO. This option is not usually
required because the board can be auto-detected.
elsa_w1000isa - (S3) Enables support for the ELSA
Winner 1000 ISA. This option is not usually
required because the board can be auto-detected.
elsa_w2000pro - (S3) Enables support for the ELSA
Winner 2000 PRO. This option is not usually
required because the board can be auto-detected.
pci_hack - (S3) Enables a workaround for problems
seen with some PCI 928 cards on machines with a
buggy SMC UART.
s3_964_bt485_vclk - (S3) Enables a workaround for
possible problems on cards using the 964 and
Bt485.
XFree86 Version 3.3.5 13
XF86_Accel(1) XF86_Accel(1)
genoa, stb, hercules or number_nine, - (S3) These
options may used to select different defaults for
the blank delay settings for untested cards with
IBM RGB5xx RAMDACs to avoid pixel wrapping prob
lems.
slow_vram - (S3) Adjusts the VRAM timings for
cards using slow VRAM. This is required for some
Diamond Stealth 64 VRAM and Hercules Terminator 64
cards.
fast_vram - (S3) Adjusts the VRAM timings for
faster VRAM access. There will be display errors
and pixel garbage if your card can't support it.
slow_dram - (W32) Adjusts the DRAM refresh for
cards with slow DRAM. Try this if your monitor
goes into power save mode with the W32 server and
older W32 cards.
slow_dram_refresh - (S3) Adjusts the DRAM refresh
for cards with slow DRAM to avoid lines of cor
rupted pixels when switching modes.
pci_burst_on - (W32) Turns on the PCI burst for
the W32p chipset. Use this if your picture looks
distorted and your mouse leaves trails behind with
burst disabled.
pci_burst_off - (W32) Turns off the PCI burst for
the W32p chipset. Use this if your picture looks
distorted and your mouse leaves trails behind with
burst enabled.
w32_interleave_on - (W32) Turns on the memory
interleave for the W32i and W32p chipset. Try
this if your server runs stable with it.
w32_interleave_off - (W32) Turns off the memory
interleave for the W32i and W32p chipset. Try
this if your picture looks distorted or you don't
get a picture at all.
no_block_write - (Mach64) Disables the block write
mode on certain types of VRAM-based Mach64 and on
Rage Pro-based cards. If noise or shadows appear
on the screen, this option should remove them.
block_write - (Mach64) Enables the block write
mode on certain types of VRAM-based Mach64 and on
Rage Pro-based cards. Normally the Mach64 server
will automatically determine if the card can han
dle block write mode, but this option will over
ride the probe result.
XFree86 Version 3.3.5 14
XF86_Accel(1) XF86_Accel(1)
no_bios_clocks - (Mach64) The Mach64 server nor
mally reads the clocks from the BIOS. This option
overrides the BIOS clocks and forces the server to
use the clocks given in the XF86Config file.
no_program_clocks - (Mach64) The Mach64 server
will automatically detect the clock chip and pro
grams it directly from the video modes given.
This option disables the clock chip programming
and forces the use of the pre-programmed clocks
either read from the BIOS or given on the Clocks
line in the XF86Config file.
clkdiv2 - for all accelerated chipsets using a set
of discrete clocks (i.e. not using a programmable
ClockChip or a ClockProg, like older S3 cards and
most ET4000W32 cards). With this option enabled,
the accelerated driver can also use all the clocks
mentionned in the clocks line divided by 2, pre
senting the server with twice as many clocks to
choose from, especially in the low- and mid-range.
This is useful for creating very low resolution
modes like 320x200, because the lowest available
clock on many cards (25.175 MHz) is too high to
create a standard 320x200 mode. A few SVGA chips
don't support this option, causing a distorted
screen (S3-805 rev C and P are known to have this
problem).
There are also numerous tuning options for the AGX
server. Refer to README.agx for details.
Note that XFree86 has some internal capabilities to deter
mine what hardware it is running on. Thus normally the
keywords chipset, clocks, and videoram don't have to be
specified. But there may be occasions when this autode
tection mechanism fails, (for example, too high of load on
the machine when you start the server). For cases like
this, one should first run the server on an unloaded
machine, look at the results of the autodetection (that
are printed out during server startup) and then explicitly
specify these parameters in the configuration file. It is
recommended that all parameters, especially Clock values,
be specified in the XF86Config file.
FILES
<XRoot>/bin/XF86_S3 The 8, 16, and 32-bit color
X server for S3
<XRoot>/bin/XF86_S3V The 8, and 16-bit color X
server for S3 ViRGE
<XRoot>/bin/XF86_Mach8 The 8-bit color X server for
Mach8
XFree86 Version 3.3.5 15
XF86_Accel(1) XF86_Accel(1)
<XRoot>/bin/XF86_Mach32 The 8, and 16-bit color X
server for Mach32
<XRoot>/bin/XF86_Mach64 The 8, 16, and 32-bit color
X server for Mach64
<XRoot>/bin/XF86_P9000 The 8, 16, and 32-bit color
X server for the P9000
<XRoot>/bin/XF86_AGX The 8, and 16-bit color X
server for AGX and XGA
<XRoot>/bin/XF86_W32 The 8-bit color X server for
ET4000/W32 and ET6000
<XRoot>/bin/XF86_I128 The 8, 16 and 32-bit color X
server for Imagine 128
<XRoot>/bin/XF86_TGA The 8-bit color X server for
DEC TGA
<XRoot>/bin/XF86_8514 The 8-bit color X server for
IBM 8514 and true compati
bles
/etc/XF86Config Server configuration file
<XRoot>/lib/X11/XF86Config Server configuration file
(secondary location)
<XRoot>/lib/X11/doc/README.agx
Extra documentation for the
AGX server
<XRoot>/lib/X11/doc/README.P9000
Extra documentation for the
P9000 server
<XRoot>/lib/X11/doc/README.S3 Extra documentation for the
S3 server
<XRoot>/lib/X11/doc/README.W32
Extra documentation for the
W32 server
Note: <XRoot> refers to the root of the X11 install tree.
SEE ALSO
X(1), Xserver(1), XFree86(1), XF86Config(4/5), xvid
tune(1), xdm(1), xf86config(1), xinit(1)
AUTHORS
In addition to the authors of XFree86 the following people
contributed major work to this server:
XFree86 Version 3.3.5 16
XF86_Accel(1) XF86_Accel(1)
Kevin E. Martin, martin@cs.unc.edu
Jon Tombs, tombs@XFree86.org
Rik Faith, faith@cs.unc.edu
Did the overall work on the base accelerated
servers.
David Dawes, dawes@XFree86.org
Dirk Hohndel, hohndel@XFree86.org
David Wexelblat, dwex@XFree86.org
Merged their work into XFree86.
Jon Tombs, tombs@XFree86.org
David Wexelblat, dwex@XFree86.org
David Dawes, dawes@XFree86.org
Amancio Hasty, hasty@netcom.com
Robin Cutshaw, robin@XFree86.org
Norbert Distler, Norbert.Distler@physik.tu-muenchen.de
Leonard N. Zubkoff, lnz@dandelion.com
Harald Koenig, koenig@tat.physik.uni-tuebingen.de
Bernhard Bender, br@elsa.mhs.compuserve.com
Hans Nasten, nasten@everyware.se
Dirk Hohndel, hohndel@XFree86.org
Joe Moss, joe@morton.rain.com
Development and improvement of the S3 specific
code.
Kevin E. Martin, martin@cs.unc.edu
Rik Faith, faith@cs.unc.edu
Tiago Gons, tiago@comosjn.hobby.nl
Hans Nasten, nasten@everyware.se
Scott Laird, scott@laird.com
Development and improvement of the Mach8 and
8514/A specific code.
Kevin E. Martin, martin@cs.unc.edu
Rik Faith, faith@cs.unc.edu
Mike Bernson, mike@mbsun.mlb.org
Mark Weaver, Mark_Weaver@brown.edu
Craig Groeschel, craig@metrolink.com
Bryan Feir, jenora@istar.ca
Development and improvement of the Mach32 specific
code.
Kevin E. Martin, martin@cs.unc.edu
Development of the Mach64 specific code.
Erik Nygren, nygren@mit.edu
Harry Langenbacher, harry@brain.jpl.nasa.gov
Chris Mason, clmtch@osfmail.isc.rit.edu
Henrik Harmsen, harmsen@eritel.se
Development and improvement of the P9000 specific
code.
Henry Worth, henry.worth@amail.amdahl.com
XFree86 Version 3.3.5 17
XF86_Accel(1) XF86_Accel(1)
Development of the AGX specific code.
Glenn Lai, glenn@cs.utexas.edu
Dirk Hohndel, hohndel@XFree86.org
Koen Gadeyne, koen.gadeyne@barco.com
Development of the ET4000/W32 and ET6000 specific
code.
See also the XFree86(1) manual page.
BUGS
Some S3 cards with Bt485 RAMDACs are currently restricted
to dot-clocks less than 85MHz.
The P9000 server may still have problems with cards other
than the Diamond Viper VLB. There may still be problems
with VGA mode restoration, but these should almost never
occur. Using physical resolutions different from the vir
tual resolution is not supported and is not possible with
the P9000. Use at dot-clocks greater than 110 MHz is not
recommended and not supported. Diamond claims that 135
MHz is the maximum clock speed, but some of their bt485's
are not rated that high. If you do not have a 135 MHz
bt485 on your Viper, contact Diamond tech support and they
will send you an RMA number to replace the board. Accel
eration is being added in slowly. At the present, only
CopyArea and MoveWindow and DrawLine are implemented.
Other accelerated features are being tested and may be
available in the next release. There seems to be a prob
lem with olvwm when used with xdm and VT switching. The
cursor will be messed up when you return to a VT if the
cursor changed while you were in the VT.
The ET6000 server is quite new, and therefor not very
thoroughly tested. Accelerated support is present, but
doesn't use the full potential of the ET6000 chip (yet).
CONTACT INFO
XFree86 source is available from the FTP server
ftp.XFree86.Org and mirrors. Send email to
XFree86@XFree86.Org for details.
XFree86 Version 3.3.5 18
Man(1) output converted with
man2html