Re: radeon r300

On Tue, Jan 20, 2026 at 3:40=E2=80=AFAM Michael <[email protected]> wr=
ote:
>
> Hello,
>
> On Mon, 19 Jan 2026 22:56:10 -0700
> Kevin Bowling <[email protected]> wrote:
>
> > On Mon, Jan 19, 2026 at 3:47=E2=80=AFPM Michael <[email protected]=
> wrote:
> > >
> > > Hello,
> > >
> > > On Fri, 16 Jan 2026 18:59:35 -0700
> > > Kevin Bowling <[email protected]> wrote:
> > >
> > > > With the modesetting default of AccelMethod "glamor", X will not st=
art with:
> > > > glamor requires at least 128 instructions (64 reported)
> > > >
> > > > So it seems the older EXA radeon might be preferable, but something=
 is
> > > > going wrong in the kernel.  Has anyone looked into this drm stuff
> > > > recently?
> > >
> > > Do you actually care about DRM? If all you want is 2D accelerated X
> > > with xrender and a fast console you may be better off with radeonfb a=
nd
> > > the old, pre-KMS radeon driver. This is the default on macppc and
> > > sparc64 but hasn't been tested much on anything else.
> > > Will need
> > > options         RADEONFB_BIOS_INIT
> > > if your card's connectors aren't the usual VGA and DVI.
>
> > I am just experimenting with a sentimental laptop and seeing what it
> > can do.  The device under test is a ThinkPad T42p.  I have a sister
> > system, an X40, with fairly less specs (low voltage Dothan CPU, Intel
> > Extreme graphics while the T42p is the maximum config of that era)
> > running XP and it can plausibly browse the web using a vintage Firefox
> > derivative like Mypal browser.  I am trying to see if I can get the
> > firefox52 in pkgsrc to work at similar or better speed.  I think it
> > needs full hw accel or opengl to have a chance to perform.
>
> I found palemoon quite usable on a 1.6GHz powerbook which has a
> mobility 9600 with 64MB, with xrender enabled, and 2GB RAM. Not sure
> what they mean by 'full hw accel' these days but I bet most of it is
> either xrender / alpha blending or video decoding. The old radeon
> driver will do at least xrender.
> Firefox52 was always rather sluggish by comparison.

Hi Michael,

Continuing on through the ages, the next device under test is a lower
spec X61 (Intel C2D T7250 with 965GM).  The machine came loaded with
Windows 10 and it was much more sluggish than the X40 running XP, so
that quickly went and we put NetBSD current amd64 on it.  Testing the
intel and modesetting drivers, there doesn't seem to be much of a
difference between the two but I've had better stability with
modesetting especially on newer hardware.  Firefox 145.0.2 from the
pkgsrc 11.0 binaries.  about:config reveals poor fitment to the
graphics hardware, WebRender is running in software mode due to the
limited OpenGL level.

So I think it is probably reasonable to run the T42 with KMS,
modesetting, and no accel for now, as its OpenGL level is similarly
downlevel and Firefox would need at least r600 to see WebRender
benefits.

palemoon's UI is especially much snappier on these old machines!

> This is what it looks like:
> [     1.000000] radeonfb0 at pci0 dev 16 function 0: ATI Technologies Rad=
eon Mobility 9600/9700 (M10/11) NP (rev. 0x00)
> [     1.000000] trying to read disabled BIOS...
> [     1.000000] radeonfb0: No video BIOS, using default clocks
> [     1.000000] radeonfb0: refclk =3D 27.000 MHz, refdiv =3D 12 minpll =
=3D 125000, maxpll =3D 400000
> [     1.000000] radeonfb0: 64 MB aperture at 0xb8000000, 64 KB registers =
at 0xb0000000
> [     1.000000] radeonfb0: display 0: initial virtual resolution 1280x854=
 at 8 bpp
> [     1.000000] radeonfb0: using 32 MB per display
> [     1.000000] radeonfb0: port 0: physical 1280x854 60Hz
> [     1.000000] radeonfb0: port 1: physical 1280x854 60Hz
> [     1.000000] wsdisplay0 at radeonfb0 kbdmux 1: console (fb, vt100 emul=
ation)
> ...
>
> [  2419.385] (II) RADEON(0): TOTO SAYS 00000000b0000000
> [  2419.385] (II) RADEON(0): MMIO registers at 0x00000000b0000000: size 6=
4KB
> [  2419.385] (II) RADEON(0): PCI bus 0 card 16 func 0
> [  2419.386] (=3D=3D) RADEON(0): Depth 24, (--) framebuffer bpp 32
> [  2419.386] (II) RADEON(0): Pixel depth =3D 24 bits stored in 4 bytes (3=
2 bpp pixmaps)
> [  2419.386] (=3D=3D) RADEON(0): Default visual is TrueColor
> [  2419.386] (II) RADEON(0): VGAAccess option set to FALSE, VGA module lo=
ad skipped
> [  2419.387] (=3D=3D) RADEON(0): RGB weight 888
> [  2419.387] (II) RADEON(0): Using 8 bits per RGB (8 bit DAC)
> [  2419.387] (--) RADEON(0): Chipset: "ATI Radeon Mobility 9600/9700 (M10=
/M11) NP (AGP)" (ChipID =3D 0x4e50)
> [  2419.387] (--) RADEON(0): Linear framebuffer at 0x00000000b8000000
> [  2419.387] (II) RADEON(0): AGP card detected
> [  2419.464] (II) RADEON(0): Attempting to read un-POSTed bios
> [  2419.539] (WW) RADEON(0): Unrecognized BIOS signature, BIOS data will =
not be used
> [  2419.540] (II) RADEON(0): Generation 2 PCI interface, using max access=
ible memory
> ...
> [  2420.302] (II) RADEON(0): Render acceleration enabled for R300/R400/R5=
00 type cards.
> [  2420.302] (II) RADEON(0): Setting EXA maxPitchBytes
> [  2420.302] (II) RADEON(0): num quad-pipes is 1
> [  2420.303] (II) EXA(0): Offscreen pixmap area of 60522496 bytes
> [  2420.303] (II) EXA(0): Driver registered support for the following ope=
rations:
> [  2420.304] (II)         Solid
> [  2420.304] (II)         Copy
> [  2420.304] (II)         Composite (RENDER acceleration)
> [  2420.304] (II) RADEON(0): Acceleration enabled
> [  2420.304] (=3D=3D) RADEON(0): DPMS enabled
> [  2420.304] (=3D=3D) RADEON(0): Silken mouse enabled
> ...
> [  2420.312] (II) RADEON(0): Set up overlay video
> [  2420.313] (II) RADEON(0): Set up textured video
> [  2420.314] (II) RADEON(0): [XvMC] Associated with Radeon Textured Video=
.
> [  2420.314] (II) RADEON(0): [XvMC] Extension initialized.
>
> have fun
> Michael