|
Post Reply
Jeff 2003-05-29 05:29
I'm having a major problem getting dual displays to work with my PC. It boots up fine, and UltraMon seems to start up without problems, but the moment any significant drawing operations take place within a window located partially or entirely on the secondary side, the computer freezes. The problem exists with and without UltraMon running; disabling all acceleration for the secondary side using display properties troubleshooting doesn't help.
The config details and attempted scenarios are listed below... any ideas where to start troubleshooting?
Configuration:
Windows 2000
5.00.2195, Service Pack 3
IBM NetVista 8303-53U w/P4 1.8ghz, 640mb PC2100 DDR (512mb + 128mb)
Integrated Intel 82845G/GL (8mb shared, 64mb aperture)
PCI bus 0, device 2, function 0
Intel driver 6.13.1.3069 (4/5/2002)
E0000000-E7FFFFFF
EFF00000-EFF7FFFF
IRQ 16
1152 x 864, 16-bit, 75hz
PCI S3 Savage/IX (Apollo XPERTPLAY 3000 w/8mb)
PCI slot 2 (PCI bus 2, device 9, function 0)
S3 driver 5.12.1.7062 (date not available)
B8000000-BFFFFFFF
03B0-03BB
03C0-03DF
000A0000-000BFFFF
1024 x 768, 16-bit, 75hz
Observed Scenarios:
PCI = bios primary, integrated = windows primary: freeze on draw in PCI screen.
integrated = bios primary, integrated = windows primary: Windows can't start PCI card (error 10).Apparently, S3 chips need BIOS-level access that only the primary card is granted under Win2k.
PCI = bios primary, pci = windows primary: freeze on draw in PCI screen. Fortunately, windows lived long enough to log in, open the properties panel, move it to the integrated screen, and make the integrated the primary again after the computer hung as usual and required a reboot.
If I mark the integrated video as "disconnected" in display properties and make the PCI video the windows primary, the same problem manifests itself -- a hard system hang the moment any significant drawing takes place.
|
Jeff 2003-05-29 06:22
Update...
Updated to newest Intel 845G driver... 6.13.10.3510 (4/15/2003). No improvement.
No sign of a newer S3 driver anywhere.
BIOS primary = PCI, PCI apparently treated as primary by windows even though it wasn't explicitly set, booted into VGA mode: the card appeared to work... but of course, only in VGA mode. No freezes in a half hour of active use, though.
same settings, but booted into normal mode: Screen corruption at "Press ctrl-alt-delete to begin", but mouse pointer didn't freeze, and system appeared to be running UNTIL I got brave and tried to move the "Welcome to Windows" window. Freeze.
disabled PCI pairity in BIOS setup. No improvement. Well, I did notice one thing... If I ran the mouse pointer over the corrupted lower portion of the "ctrl-alt-delete" window displayed at login, it refreshed the corrupted area and replaced it with what was supposed to be there. Up to a point... the moment I got brave and moved up above the lower half inch or so of the c-a-d window, hang.
went back into BIOS setup to make integrated the primary active video. Noticed that there's no apparent way to disable video bios shadowing, and wondered whether this might be one of those problems that only affect mass-market PCs having a dumbed-down BIOS...
|
Jeff 2003-05-29 07:01
Got desperate. Tried forcibly installing driver for ATI XPERT@PLAY on the 1% chance the resemblance between names was more than a coincidence. No such luck. The computer came back up in safe mode, and when prompted to find a better driver, happily chose the S3 Savage/IX as before.
Sigh.
|
ecarlson 2003-05-29 09:43
Try upgrading the motherboard BIOS.
- Eric www.InvisibleRobot.com
|
Jeff 2003-05-30 00:59
Unfortunately, no newer BIOS is available from IBM.
I also tried Knoppix 3.1 (March 31, 2003 build). Interestingly, it suffers almost the same fate as Windows 2000... mild screen corruption, and general death of the video subsystem. However, the mouse pointer remains alive, and the video subsystem's death doesn't appear to be total (closing Konqueror, for example, results in the excruciatingly slow redraw of the KDE desktop, one... icon... at... a... time....
I'm still suspicious of the BIOS's apparent failure to assign the card an IRQ, and suspect that KDE might "sort of live" because it simply times out after some long period of waiting with no IRQ firing, while Win2k might patiently wait forever for the IRQ that will never fire. Is there any way to probe KDE or Linux to discern whether or not it is, in fact, using an IRQ for the video card? Or whether it wants one, but couldn't get one?
|
ecarlson 2003-05-30 10:30
You can see what IRQ's the card is using in Windows Device Manager, and in Windows System Information.
Have you tried other PCI slots?
- Eric www.InvisibleRobot.com
|
ecarlson 2003-05-30 10:33
I don't think old video cards require IRQ's. I remember in the old days having the option to use or not use an IRQ for video.
- Eric www.InvisibleRobot.com
|
Jeff 2003-05-30 10:44
Sigh. Any more suggestions before I throw in the towel and give up on the machine?
I tried TweakBios, but its int19 reboot scheme didn't work... I guess the boot partition needs to be 2-gig-compliant or something, because it gave a "no/invalid boot drive" error (presumably because DOS6 doesn't have the slightest clue in hell what to do with a 30-gig FAT32X partition).
Here are my working theories so far, most of which I don't see any real way to confirm:
* The Savage/IX card needs an IRQ, and the BIOS isn't giving it one.
* Something is probing the video BIOS, not realizing that it's cached and not write-protected, and corrupting it in some major or minor way (or at least propagating some side effect that lives on after the video bios becomes moot and the windows driver takes over).
* The card is defective. Not entirely impossible given the nearly total lack of quality control among bargain-basement PC component manufacturers. I put the likelihood at somewhere between 5 and 20 percent.
* The card is incompatible with the PC. More specifically, the card and the PC both suck, having been intentionally engineered to the outermost limits of their components, and as a result failing to meet somewhere in the middle and play nicely together... made worse by the computer's inflexibility and refusal to budge even a little bit to accommodate the card. At least as likely as the chances the card is defective. I put it at 20-30%.
I've pretty much ruled out driver issues due to the fact that the card fails under both Windows and Knoppix (CD-bootable Debian Linux). If it worked flawlessly under Linux, I'd blame Windows in a heartbeat... but the fact that it failed under both in visibly similar ways suggests that drivers are the least of its problems.
Does anybody know of any utility that can be used to acid-test a PCI video card (maybe a S3-family card, or Savage/IX card specifically) from some bootable disk that doesn't depend upon an operating system's drivers? Like maybe one that runs from a bootable DOS or bootable console-mode Linux CD and hammers the card directly to make sure everything works right (and maybe gives insight as to what's wrong)?
|
ecarlson 2003-05-30 12:53
You should be able to download lots of old DOS based video card test utilities. They wouldn't be specific to that card though.
Have you tried the card in a different computer, or a different card in that computer?
- Eric www.InvisibleRobot.com
|
Jeff 2003-05-31 02:02
I tried moving the card from slot 2 to slot 1, partly as a strategy to force Windows to freshly reinstall the drivers in case the previous set were fatally fried (you know, those times where a few driver files get mangled or incorrectly copied, then subsequent attempts to reinstall the drivers leave the bad files untouched because Windows won't overwrite or delete them for some reason because it stubbornly believes them to be better than the new files, and Windows lacks any mechanism for unconditional "scorched earth" driver replacement.).
Just to clarify... the card is low-end, but it's not particularly old. I'd put its manufacture date somewhere around late 2001/early 2002. As far as I know, DOS utilities specific to it don't exist, and probably never will.
The big problem I see insofar as diagnostics go is that today, there's really no good way to go "around" Windows to see what's really going on hardware-wise. DOS (as in, real mode legacy operating systems and their cobbled-together descendants like FreeDOS) simply can't deal with modern PCs, because too many fundamental assumptions that were valid 10 years ago are now routinely violated by modern PCs (2-gig drive limit, ps/2 keyboard/mouse, the way serial/parallel ports work... or even their very existence in the first place, etc.).
It won't help my current problem, but I look forward to the day a year or so from now when hardware manufacturers realize that they can write diagnostic utilities to run under Linux and distribute them with their hardware on a bootable Linux CD... so even if the user runs Windows, or Windows is totally fried, he'll be able to boot into a civilized, modern environment and assess the situation from there. God forbid, if only PowerQuest made a Linux version of Partition Magic and distributed it on a bootable Linux CD...
|
Jeff 2003-05-31 02:06
I can't really try a different card in the computer, though. The stupid PC only takes low-profile PCI cards. I had a hellish time finding a cheap low-profile PCI graphics card as it was. Low-profile AGP cards are fairly cheap and abundant on eBay, but wouldn't have done much good as a second card since it would have disabled the onboard AGP video.
I haven't been able to try the card on a different computer yet, mainly because I'm reduced to my laptop for another week or two because my home computer's motherboard is away getting its bad capacitors replaced and likely won't be back for another week or so (I sent it to Epox last week).
|
ecarlson 2003-06-01 09:43
Sorry to hear you got hit with the bad capacitor problem. My primary computer's motherboard is too old to have that problem.
You have onboard video and an AGP slot? My secondary computer has onboard video, but no AGP slot. Are you sure the onboard video is on the AGP bus? Maybe you should also add an AGP video card so the motherboard will stop using shared RAM for the onboard video.
- Eric www.InvisibleRobot.com
|
Jeff 2003-06-04 00:24
Hmm... this is a totally random theory, but is it safe to assume that *ALL* PCI-based graphics cards do, and must, have their own RAM? Or is it theoretically possible for somebody to try making an ultra low-end PCI-based video card that tries to use shared system RAM as its frame buffer?
I've always assumed it to be outright impossible due to PCI bandwidth (aside from the fact that mere refreshing activity alone would saturate the bus even if it WERE adequate), but I vaguely remember reading somewhere that there were AGP-based cards (Intel i740?) that did, in fact, rely entirely upon the PC's onboard RAM for their own needs.
The point I'm working towards is that, if the PCI card is, in fact, trying to use the system's ram for its own framebuffer, the combined PCI bus saturation and shared-ram contention could be what's hanging the system. The strange thing is that the video card itself only has two chips -- a 1.25" square chip that's presumably the GPU, and a big 1970s-looking chip that's presumably the BIOS ROM. I just assumed that the RAM was part of the GPU core, but who knows... with computer manufacturers desperate to shave every cent off the manufacturing cost, god knows what they might try to pull off if they think they can get away with it...
Of course, there's always the original theory that the card is just defective. I'm hoping I'll get my home PC motherboard back within another day or two so I can test it before returning it to the vendor (the RMA window is rapidly dwindling).
|
Idealius 2003-06-04 02:47
Looks like the onboard video (Intel) just won't support it, that's no big suprise... just get two PCI cards. You can find a 16MB NVIDIA PCI Vanta for cheap these days. You might have to disable the onboard video in Device Manager for it to work properly, but that should work fine. I had to do that on an old 333 K6-2. The onboard video automatically disabled itself whenever a different video card was installed, no jumper or anything -- just used two PCI cards and everything worked fine.
|
ecarlson 2003-06-04 13:17
Or get one dual-output PCI video card, and kill the onboard video. I'm running a dual output AGP card, and it works nicely for me. My AGP card is the PNY GeForce4 Ti4200 w/64.Meg.
- Eric www.InvisibleRobot.com
|
Post Reply
|
