Oh! NO! Got my MAX and it doesn't do busmastering! Argh! Haig, BBCrew! Help!

[Scytale]

This is beyond belief. I got my MAX today, and I'm having almost exactly the same problem I had with my G200 a year ago. The old G200 problem was that I couldn't get D3D busmastering to work under W9x on my P2L97S board. It worked fine with NT. I solved the problem by bringing in a BH6. End of story, months of fun.

So today I received my MAX. Launched W98, selected standard VGA, removed PowerDesk 5.13, and replaced the G200 with the MAX. [Insert reboots where appropriate].
Installed PD 5.13 again, kicked the mule, and there the story continued.

For a few seconds things went fine, but then the desktop became all garbled, and I had to reset ('repeated CAD' still worked). I went through this a couple of times. Sometimes the desktop just froze, other times it became all garbled again, including the mouse cursor.
I upped my aperture to 256 and tried running the Tech Demo: It froze down after the menu, sound still playing.
I tried Expendable. It ran fine (and looked really great), until I walked around the corner, blasted an incoming (sorta falling) truck, and tried to walk on. Then the game froze on me. I rebooted, started again, and the same thing happened at the same spot.
Similar problems with a bunch of other D3D games.

I went back to VGA, reinstalled DX61 and PD513, but no change.

Something started to itch badly, and I decided to disable busmastering. And behold, everything worked fine, but obviously deadslow.

Like before, in August 98, busmastering under NT (incl. OpenGL) works fine. The card itself seems okay. And she is a looker.

Any suggestions? I'll try ditching the sound card later today. There are no obvious hardware conflicts.

BH6, P2 333@333-450, G400 MAX (G4+MMDHA32G, Rev A), TB Montego A3D I, 3Com 3C905B-X, Diamond FirePort SCSI. W98, DX61, PD513, latest drivers of everything.

[SCompRacer]

Have you tried to run the Matrox utility "pcispy" located in the utilities folder, C:\Matrox\util (if you installed it). Run it in a DOS window with the -l switch and see if you get any conflicts, particularly with SCSI card.

I have a P II 400, ASUS P2B-LS MB, all SCSI (except for IDE DVD CD ROM), with 256 MB ECC RAM, dual Monster 3D II's, Hollywood + DVD Decoder card and SB Live. Any Open GL app crashes (Quake II, 3D Studio MAX R2), giving me blue screen/memory parity error, must reset to reboot. Forsaken refuses to run after Acclaim splash screens, get "cannot read video or audio stream."

I tried removing all cards and booting from another SCSI drive with fresh Win98 install, DirectX 6.1, 5.13 drivers and nothing but the AGP card with same results. Disabling ECC allows Open GL apps to run but has no effect on Forsaken. Disabling ECC is not a permanent option and I switched to Diret3D drivers so I can use 3D Studio MAX.

I put the SB Live card back in PCI slot 1 (which shares an IRQ with AGP slot) and it works with bus mastering enabled. I then reinstalled all the cards and I can still play the tech Demo with no trouble.

pcispy reports a conflict with "Device 6" which is my Mass Storage Controller, the Adaptec 7890 chipset (equivalent to 2940U2W card). Further states "Base 00000000 (8 o's) and Size 00000010 (six o's and 10), run with -c switch to resolve conflicts." Well it does not resolve any. Device Manager and System Information fail to confirm any conflict.

Matrox has not responded yet so I am sure it is going through the "Tech Support Gauntlet" though I suspect driver and possible SCSI memory address issues. My prior card, a Diamond Fire GL 1000 Pro, never gave me any problems in Open GL.

Hope this helps, and remember, luck is where preperation meets opportunity!

Rich O

[This message has been edited by SCompRacer (edited 08-02-99).]

[This message has been edited by SCompRacer (edited 08-02-99).]

[Scytale]

Well, reinstalling W98 actually made things worse. I deleted everything and did a fresh W98 install, then installed DX61, followed by the latest G400 PowerDesk.

No luck. Now the display freezes almost immediately when attempting to draw the desktop. Only the mouse keeps responding to movement.

I'm using all different hardware and I'm in just the same mess as one year ago. The only things in common are Windows98, DirectX, Matrox, Gxx and PowerDesk.

Argh!

[M Ragsdale]

Try using different system RAM. Also go into your bios and look at chipset features and make sure the voltages are within their specified range (hardware monitor).

Matt

[CoolBreeze]

HI all,

I had the same exact problem with my OEM SH G400 32. Everything would crash (OGL D3D) even the tech demo while loading. 2D desktop was unstable too...i would get strange color lines and garbled images. It took me a week to get this card to work right. I had to do a complete C:format (erase EVERYTHING on HDD) and a fresh install of win 98. Then the first thing i did was install the G400....then the rest of my boards and programs...now everything seems to run just fine. Here is my sys specs:

Celeron 366 not overclocked....128 MB ram...8.2gig HDD...40x cd rom...MX300 sound card...ethernet pci card for cable modem...SE440BX-2 Mobo...and the OEM SH G400 32mb.

It sux...but try to reformat yer HDD...

CoolBreeze

[Gurm]

Ok, from what I'm hearing you are definitely looking at a BIOS issue. Go through the BIOS. Set all the speeds and voltages to default. Don't overjuice the CPU, don't underjuice it.

Set everything to very reasonable numbers. Set the AGP aperture to half of your system RAM. Turn off cachine and shadowing.

Basically set the BIOS into "I'm a slow-ass 486" mode. Then see if everything works. If it does (and I think it might) then you can try turning stuff back on until it breaks.

Sometimes it's the dumbest stuff. My last system board wouldn't let me cache Video RAM. If I had Video RAM cache on, the machine would lock... but ONLY while I was in fullscreen DOS box mode. It was totally wierd.

I'm not saying THAT is your problem, I'm just illustrating how wierd BIOS issues can be.

- Gurm

P.S. Be aware of slot IRQ sharing on ABIT boards:

AGP = IRQA
PCI1 = IRQA
PCI2 = IRQB
PCI3 = IRQC
PCI4 = IRQD
PCI5 = IRQD
USB = IRQD

And disable ACPI and Power Management entirely, just leave "control by APM" on if you can.

[Gurm]

A fresh reinstall always works.



Of course, you didn't want to hear that, right? You also don't want to hear that you need to make sure there's no IRQ sharing, etc. etc.

Write again if you've already tried all of this.

- Gurm

[Scytale]

Thanks for your reply, Rich. I tried pcispy and the one thing I found was that the G400 and NIC ROM bases are both reported to reside at 00000000. Doesn't seem to make much sense...

I removed everything but the SCSI, but no improvement. No luck yet.

[Patryk2000]

Hello Guys,

I don't know what it's worth but when I first put my G400 I used to have crashes and screen freezings... until I found out that my BH6's little piece of plastic that supports the Ram was preventing the G400 from being totally "pushed" inside the AGP slot.
And up to now things r great, no more problems.

So u should check this out, just in case.
Hope this helps.

[Scytale]

Sure Gurm, I'll reinstall W98. I'll use the axe. I use it all the time.

[Scytale]

Thanks for your suggestions guys. There's no IRQ sharing going on. I removed all cards but SCSI and G400 and disabled all on-board peripherals. I did not reformat the partition but, from NT, I simply deleted everything on my W98/boot partition except for my NT boot loader.

Gurm I'll try the BIOS settings. I already tried declocking CPU and RAM, and shadowing, but there's still a lot to check.

Note that this is not really a 'plain' hardware problem but OS/driver related. NT4SP5 OpenGL works fine (when running my P2 at a modest 400 MHz, Q2 demo1 still does 25.5 fps), using busmastering. Taking all into account it looks like a timing problem when using AGP mode. Or -you may be correct Matt- mainboard current/voltages.

Okay, I'll be digging, buying, tweaking, and swapping now. Thanks again, I'll be back. I would like this card to work, it seems worth the trouble.

[M Ragsdale]

Hi Scytale,
Also make sure to try different RAM, too. I have seen this too many times. If you have two sticks in, take one out and leave the other one in, and vice-versa. Some BH6's are picky eaters when it comes to RAM. AGP cards don't put up with subpar memory, either. Be patient, this can be solved.

[Scytale]

Oh! NO! I found that everything (incl. busmastering) works without a hitch when I force AGP to 1X and select a 64MB aperture. Even the Tech Demo works fine.

This convinced me that there is a timing problem with the card and/or driver. To summarize:

- AGP2X works okay, but only when busmastering is disabled.
- AGP1X seems to work okay (under the conditions I tested).
- AGP disabled works okay.
- Busmastering works okay, but only for AGP modes < 2X. Busmastering does work (basically) for AGP2X but causes a sreen lockup/corruption at some time.

The AGP2X/busmastering problem is less apparent when I clock my system at a 66MHz FSB. It still messes up at a certain point in time, but in general it takes longer.

My best guess is that this is just a driver timing problem. Worst guess is that it could be a (hardware) design problem the G400 inherited from the G200. If Matrox could nail this one in software I think they would make a lot of people (G400 and maybe G200 users) very happy.

[M Ragsdale]

That could very well be a hardware problem, (Power supply voltage, bad ram, etc.) too. I am not too sure there is a timing problem with the drivers, other people are having no problems running 2X with Bus Mastering enabled on their ABit boards, so I would dig a little deeper. You are getting closer to finding your answer.

Matt

[Scytale]

Hi Matt, I agree that the PS may still be a problem (I never trust these things), but I think it's becoming somewhat less likely. The PS apparently has no problem generating enough juice to run AGP2X by itself. Also, it handled the G200 fine, but of course the power requirements for the MAX will probably be more demanding.

On the RAM, the problem is not resolved by being (extremely) easy on my LGS7J RAM by clocking the FSB to 66 MHz. These particular DIMMs can even take 133MHz.

Then again, of course I'm going to try your suggestions. Having a theory is one...

The fact that not everybody does experience a timing problem with the drivers is exactly what a timing problem is about. I have been a driver developer (long ago, so it seems), and let me tell you, these are obscure problems. Swap in a different component (like, in this case, RAM) and you may have an entirely different picture.

Okay I'm ready for the next step. I'm sure we're going to nail this one. Thanks.