TUA266 with sound and AGP video problems (long)
Paul
it as a RAM tester. This board seemed perfect, as it has SDR and
DDR slots on it, and can take large capacity memory DIMMs.
Recently, I found some free software, that would only run under
WIN2000 or WINXP, so I decided to build up a machine around this
board, so I could use the software.
The first thing I noticed, after loading BIOS 1004_10, WIN2000
and upgrading to SP2, was that the sound would lag or break up
with a static-like sound, every time the disk drive was accessed
while sound was playing. If I play a streaming movie from a web
site, there is disk activity once a second, that upsets the sound.
The disk activity is the WIN2000 file cache (not paging, as I
tested disabling paging to the hard drive by using a RAM disk
for the pagefile).
Doing a search on deja.com (google groups), suggested that the
problem could be due to the PCI Latency. I tried changing the
sound card (changed from a SoundBlaster to a Hercules card with
a CMI8738) and this helped a bit. I replaced the Realtek Ethernet
card with a 3COM 3C509C, and that didn't make too much difference.
I even tried using a Promise IDE card for the boot drive, but
that seemed to make the problem worse. (I used SiSoft Sandra to
see what the latency setting was for each PCI card, and selected
cards with lower latency timer values. The PCI drivers don't have
options to tune this parameter.)
I also changed the PCI Latency setting in the BIOS, from 32 to a much
lower value. This made the machine very sluggish, so I returned the
setting to 32 again. At this point, I am living with the sound
problem, as it only affects Windows system sounds or sounds played
while viewing movies. Sound in games seems unaffected.
My second problem was with the video card. I had an ATI Expert 98
AGP laying around, which I used in the machine at first. It would
run for a while (maybe a couple of hours), and then freeze up the
display.
So, now I'm trying a Radeon 8500LE AGP card. To use this, I've
acquired and loaded the following:
DirectX 8.1
Ali AGP 1.90 driver
Ali AGP Utility 1.4 (used to try changing AGP settings)
ATI Catalyst drivers, versions 2.2, 2.3, and 2.4
(The 2.4 version has both drivers and the Control Panel software
together. The others require downloading the Control Panel
installer separately.)
With no AGP driver installed, WIN2000 reverts to using ALIM1541.sys
to get basic functionality. I can use non-VGA resolutions this way,
but still experience occasional freezes.
With respect to the Ali AGP drivers, there are several versions.
Version 1.74 is installed, if you download the integrated driver V1.04
from the ASUS website. This was the first driver that allowed me to
get past a black screen, as the WIN2000 desktop comes up at boot time.
I've since gotten to the point, where I'm using the latest version
of everything, but there are still problems. The ATI SmartGart option
is very hard to control, because it doesn't seem to respect the AGP
Bios setting (1X,2X,4X). It always tries to go to 4X. Also, the
Ali AGP utility is a waste of time, as there are actually very few
entries in the registry that it will actually change, and the "manual"
settings page has most options greyed out. (So much for people
praising Ali for providing this utility!)
At this point, both DirectX and OpenGL seem to be working. Using
DXDIAG from DirectX 8.1, AGP textures are enabled, as long as an ALi
AGP driver is loaded. For testing, I've been using the madonion.com
3DMark2000 and 3DMark2001se, to try to exercise the AGP functions on
the Radeon. Initially, I could only get part way through the first
helicopter display test in 3DMark2000, before the display would hang
or the program would just exit.
My biggest improvement came, when I started back into the BIOS settings
again. Having tried everything else, I decided to change the PCI
latency setting in the BIOS from 32 to 24. After I did this, I was
able to run all the way through 3DMark2000 and 3DMark2001se.
However, the 8500 is still far from stable. Another test I use is
to run Quake2 in OpenGL mode, then run "massive1.dm2" or "crusher.dm2"
in timedemo mode. I still cannot get all the way through this test,
without Quake2 freezing (and the sound stays in a loop).
If you look at the "architecture" of the chipset for the TUA266, it
has a M1651T northbridge and a M1535D+ southbridge. These are connected
together by a 32 bit, 33MHz PCI bus. Since the AGP port on any
motherboard, uses both PCI and AGP protocols at the same time, it seems
that this chipset has some fundamental flaw with the PCI bus.
Can anyone suggest any changes to my BIOS settings, to improve the
behavior I've got? Here are the relevant BIOS settings, from the
beta BIOS 1004.10 I am currently running. I wish real info on what
these actually did, was available!
System Acceleration Mode [ Disabled ]
Note - Usenet postings suggest this is a slight overclock setting ?
Graphics Aperture size [ 128MB ]
Tried 32 and 64, but it just changes the time until a freeze occurs.
AGP Capability [ 4X ]
Doesn't matter, as ATI SmartGart software will override anyway.
Video memory Cache Mode [ UC ]
(Uncached)
Memory Hole at 15M-16M [ Disabled ]
PCI 2.1 Latency Compliant [ Disabled ]
This doesn't change the effectiveness of the latency timer setting.
Delay transaction [ Disabled ]
This has something to do with the PCI-ISA bridge in the southbridge.
Delay transaction decouples ISA transfer complete, from holding up
the PCI bus. I think that is what I read...
AGP Fast Write [ Enabled ]
ATI SmartGart in Catalyst seems to ignore this, as Fast Write claims
to be turned off.
PCI Enhancement Mode [ Enabled ]
No idea what this one does.
PCI/VGA Palette Snoop [ Disabled ]
PCI Latency timer [ 24 ]
I don't know if this limits the bus mastership time of north/south
bridge devices, or stops the other cards from transferring data after
this number of cycles...
Hardware config
TUA266 board rev 1.01 (BIOS 1002 bumped to beta BIOS 1004.10)
Maxtor 6L060J3 60GB drive, partitioned into a C and D drive, FAT32, Win2000.
LG CRD-8522B 52X CDROM, on separate IDE cable.
Radeon 8500LE 128MB in AGP slot
3COM 3C905C-TX 100BT Ethernet in PCI Slot 2
Hercules Gamesurround Muse 5.1 in PCI Slot 4
3x512MB SDRAM
Tualatin 1100/100MHz processor
Thanks in advance for any advice -
Paul
Highspeed
ASUS 8170DDR GF4 MX440, with SDRAM. But I have tried DDR, SDRAM, different
HDD, AGP cards, everything but the sound card. Although I did have the
problem with the sound card while using a Creative PCI128 card. I have had
the board replaced once already and no help. Now replacing a second time.
Lars
"Paul" <nos...@needed.com> wrote in message
news:nospam-3010...@192.168.1.177...
Paul
<cr...@crap.com> wrote:
> I have the same board and the same flaw! I am running Celeron 1000/100,
> ASUS 8170DDR GF4 MX440, with SDRAM. But I have tried DDR, SDRAM, different
> HDD, AGP cards, everything but the sound card. Although I did have the
> problem with the sound card while using a Creative PCI128 card. I have had
> the board replaced once already and no help. Now replacing a second time.
>
> Lars
>
Just an update.
Today, I tried -
PCI 2.1 Latency Compliant [ Enable ]
Delay Transaction [ Enable ]
PCI Latency Timer [ 24 or 16, rather than default 32 ]
Enabling "Delay Transaction" fixed the "Windows sound problem while
accessing disk" issue I described yesterday. However, while I was
using the Adminstration Tools, my disk drive got in a funny state,
where the drive light stayed on for 20 to 30 seconds at a time. My
theory here, is that the IDE controller is having trouble reading
some disk sectors, and is retrying the same read, over and over
again.
Rather than take a chance, I have disabled Delay Transaction again.
Doing a chkdsk c: /f revealed no problems with the disk structure.
(I don't have time to do a /r check right now.)
Checking the history, Delay Transaction caused data corruption on
Via chipsets. It is unclear to me, whether Ali has the same problem
or not, with the M1535D+ Southbridge used on the TUA266.
With regard to my video problem, I still have a problem with my Quake2
test. Maybe people on this list, who have a copy of Quake2 (patched to
226), can try this and tell me if this test works for them.
1) Get a copy of "massive1.dm2" and "crusher.dm2". These can be found
at http://www.reactorcritical.com/download.shtml
2) Make a folder in baseq2 called "demos". Place the two .dm2 files
into the "demos" folder.
3) In the baseq2 folder, create a file called demoloop.cfg, using
notepad. The file should contain the following lines.
timedemo 1
developer 1
alias loop1 "demomap massive1.dm2;wait;set nextserver loop2"
alias loop2 "demomap crusher.dm2;wait;set nextserver loop1"
loop1
4) In the main Quake2 directory, place a one-liner file "demoloop.bat"
It should contain the following line -
quake2.exe +exec demoloop.cfg %1 %2 %3 %4 %5 %6 %7 %8 %9
Now, you are ready. Double-click on the demoloop.bat file, to start
the test. What should happen, is the massive and crusher demos should
run at full speed, over and over again. After two or three passes, I
get one of two types of failure.
1) Screen freezes. The last second or so of sound loops over and over
again. System requires a reset to recover. No keyboard input is
accepted.
2) Radeon 8500LE video mode seems to get corrupted. The demo continues
to play. Sound is normal and judging by the average intensity of the
screen, the individual pixels are still being rendered. It just
seems like my screen is no longer running at 1152x870. I can
type a tilde and enter "quit" and escape from Quake2. But,
I don't have any easy way, to get the video mode to reset.
Anyway, I'd appreciate it if someone with a more stable computer than
mine, could try this, and tell me whether this is a Quake2 problem
or a hardware problem. If I was betting money, I'd say my hardware is
at fault.
Paul