VOGONS


Matrox G250 Trouble

Topic actions

First post, by Yushatak

User metadata
Rank Member
Rank
Member

I invested in a Matrox G250 AGP (factory OC'd G200) and a VRAM upgrade for it (yeah I've heard it clocks it back down, dunno if I'd end up using it or not), for use in my 300Mhz Pentium box (and yeah, I mean Pentium, not P2, etc.). I set it up, working fine, but when I install drivers for the card and reboot, the screen image is a series of lines and garbled crap, except the cursor, which appears slightly garbled but drawn correctly and controllable, with a box around it for some reason. I see patterns in the garbledness that change here and there, which I think corresponds to screen changes. Not real sure what's going on here or why the drivers don't work. Ideas?

My first thought would be corruption in the on-board VRAM for the card, above the space used for 640x480 (thus it works fine in BIOS, setup, and safe mode, as well as pre-driver install) - however, even if you boot to safe mode, force the resolution to 16-bit 640x480, and reboot into normal mode, it won't work, which it would if this were the case..

I tried installing the VRAM chip out of desperation, and it did not help either - it did cause more random corrupted lines to be displayed, though. 😜 On a side note, the card still reported 8MB VRAM, so it wasn't detected, but with the drivers malfunctioning I'm not sure if it even should be? xD

Reply 1 of 18, by RichB93

User metadata
Rank Oldbie
Rank
Oldbie
Yushatak wrote:

I invested in a Matrox G250 AGP (factory OC'd G200) and a VRAM upgrade for it (yeah I've heard it clocks it back down, dunno if I'd end up using it or not), for use in my 300Mhz Pentium box (and yeah, I mean Pentium, not P2, etc.). I set it up, working fine, but when I install drivers for the card and reboot, the screen image is a series of lines and garbled crap, except the cursor, which appears slightly garbled but drawn correctly and controllable, with a box around it for some reason. I see patterns in the garbledness that change here and there, which I think corresponds to screen changes. Not real sure what's going on here or why the drivers don't work. Ideas?

My first thought would be corruption in the on-board VRAM for the card, above the space used for 640x480 (thus it works fine in BIOS, setup, and safe mode, as well as pre-driver install) - however, even if you boot to safe mode, force the resolution to 16-bit 640x480, and reboot into normal mode, it won't work, which it would if this were the case..

I tried installing the VRAM chip out of desperation, and it did not help either - it did cause more random corrupted lines to be displayed, though. 😜 On a side note, the card still reported 8MB VRAM, so it wasn't detected, but with the drivers malfunctioning I'm not sure if it even should be? xD

Bad VRAM or maybe broken AGP on your mobo.

Reply 2 of 18, by sliderider

User metadata
Rank l33t++
Rank
l33t++

How did you get a 300mhz Pentium? The only factory 300mhz Pentium was a Tillamook and that was only in like 3 models of laptops that I know of and that was on a processor card, not socket 7.

Reply 3 of 18, by Yushatak

User metadata
Rank Member
Rank
Member
sliderider wrote:

How did you get a 300mhz Pentium? The only factory 300mhz Pentium was a Tillamook and that was only in like 3 models of laptops that I know of and that was on a processor card, not socket 7.

It's a Chinese repackaged laptop chip put into a Socketed form factor. It's a Tillamook 266Mhz laptop chip turned desktop, with a gold heat spreader. 😁

I easily OC'd it to 300Mhz, and it's been running great at that speed.

It's possible the AGP is wonky on this board, this board has always been a bit flaky, but the machine worked fine with other AGP cards in the same setup.. Any way to test the G250's VRAM?

Edit: I have an idea.. the board supports AGP at 1x and 2x, so I'll try disabling 2x in the BIOS. Perhaps it's implementation wasn't tested well due to the age..?

Edit 2: Didn't help.

Reply 5 of 18, by batracio

User metadata
Rank Member
Rank
Member

Try to underclock the card with MGATweak. Use these settings for G200 stock speeds (also compatible with the VRAM upgrade):

System PLL Clock (SCLK) = 168.75 MHz

Graphics Clock (GCLK) = 84.375 MHz (SCLK/2)

Memory Clock (MCLK) = 112.5 MHz (SCLK/1.5)

Warp Clock (WCLK) = 84.375 MHz (SCLK/2)

Reply 6 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

Found some HP-specific drivers released explicitly for the G250, tried them, at first it booted to a black screen with the cursor, got my hopes up, but it hard crashed after a while of sitting there with the cursor frozen.. I rebooted and back to the same issues.

Went looking for MGATweak.. let me tell you, that is not an easy tool to find, but I did eventually locate an archived copy (most newer links led to 404 even in the archives due to external linking and the main site using a CGI download system...).

http://web.archive.org/web/20000823034336/htt … ls/Mgat120b.zip

I'll try that out..

Reply 7 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

I booted into safe mode to switch to the default driver, went back to normal mode, hooked in jump drive, set up MGATweak to start on boot so it can apply it's settings, set it to save settings and minimize to tray..

Looking at the numbers, it seems my card is underclocked to 131Mhz or so, which is odd, considering this card should be clocked to at least 180Mhz by default iirc. The multipliers were all normal except the MCLK which was set to SYS/1, meaning equal to the SCLK, which would be really fast.. I set that to 1.5, hoping it would stabalize the card.

I then switched back to the latest Matrox driver I have for it, and rebooted, making sure to wait a bit for the MGATweak to try to start and fix things.. froze at the cursor with the garbled picture.

I'm presuming this is a dynamic program, not a "tweak-the-card-CMOS" sort of thing - correct?

I would have played with the SCLK, turning it up to 168 or down even further than 131, but it appeared I couldn't modify it, only view it. All I could change was the dividers for the other clocks, and only from a predefined list.

Of note, this is version 1.20 Beta of the utility, the only version I saw referenced as late as 2002, and which had links around for me to know the filename for. If you have a later version, that might help (though this card is from '99, so it also might not - depends on what a newer version would fix/add..).

What I really would like is a VRAM tester, see if it's got a corrupted block in there. If so, then I probably need a new card. I'd like to test in another machine with AGP, but there's not a convenient one around at the moment.

Anywho no progress, but this is a cool utility.

Edit: Oh and also, I noticed it has a .sys driver for "NT" and a .vxd driver for "95" - I imagine it takes care of using those itself, and I don't need to install the VXD or anything? If not, then I wasn't "doinitrite"... D:

Edit: Definitely changed the multiplier on the MCLK, AFAIK it resets at every boot, though. I tried flashing the BIOS, now have a wonderful card that just gives me nothing. I should be able to fix it, as the flash utility automatically saves a BIOS image for ya, so if I just pop another card in and run it with that image, it'll be back, but I'm rather discouraged and am going to stop for today on this. Any further advice/ideas would be appreciated.

Reply 9 of 18, by sliderider

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

This was a great Matrox info site back in the day.
http://web.archive.org/web/20030207162604/htt … .pl/gbm/matrox/

Did you notice this?

16.12
More rumours and some explanations: despite my good relations with Matrox, my rumours do not come from any internal sources - they would have the full right to kill me if that was the case. All the revelations on T&L may be thought out of the publicly available materials. If you study some documents (even the short ones) it becomes obvious that at least some of the graphics chips have big potential capabilities. Probably one of the earliest chips with T&L capability was Rendition Verite 2100. The fact that the chip can do T&L doesn't imply that it's good to do T&L on the chip. Any MMX-compatible CPU can nicely draw shaded and textured triangles to graphics memory, but it makes no sense to do this since the graphics chip does it faster. Some graphics chips can do T&L, but in many cases it is better to leave this task to the CPU and let the chip do its own low-level dirty job faster. That's it. No mysteries. No confidential information leaks. Just a little logic. More details THERE!!!.

And this

15.12
It's a rumour time, so here are some: The current Matrox chips, starting with G200 are theoretically capable of hardware transformation and lighting (T&L). Since for many people the "hardware T&L" is a buzzword regardless of the actual results, this capability of Matrox chips should be considered important, at least from marketing viewpoint. While it doesn't make too much sense to force the G200 to do T&L, this may probably be done to G4xx line chips. It's all in drivers though, and the programmers may or may not use the capability depending on what they consider to be better/faster solution (trust them, they were wrong so many times <Alf>). So, if you have a G400-based card, you already have (unused) hardware T&L. You feel happier, dear Matrox owner, don't you?...

Does this guy even know what the hell he is talking about if he thinks Rendition Verite 2100 and Matrox G200/G400 can do T&L? It doesn't seem like a reliable source of information at all with those two glaring fallacies on that page.

Reply 10 of 18, by swaaye

User metadata
Rank l33t++
Rank
l33t++

ya I remember that talk. He was saying that the Matrox triangle setup engine, called the WARP engine, was maybe programmable enough to do T&L. Of course nothing became of that. There's probably zero chance of it being able to do that in a performant way.

Verite was partly a MIPS RISC processor tailored to graphics so was very programmable too. They always touted how they could add features. Of course if the hardware wasn't designed with a specific task in mind it's often very slow at it.

The MPACT chip was the most general processor that could do graphics as well as other things. It's really just a specialized CPU. But it sucked at graphics compared to ASICs as expected.

Reply 11 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

After a horrific bloody battle with the BIOS flashing tools that lasted (without exaggeration) 13 hours, I've got the card running and stable again.

As an extra insult to injury, I came across a page detailing the problem I was having that led me to (mis)flash the BIOS, and I dealt with that by switching a BIOS setting (there were multiple solutions to try). However, with my freshly reflashed card, I'm running a v3.3 BIOS and a custom PIN file that was made by me slapping the PIN data from one file onto the header from another. 😜

Here's my new problem: the drivers no longer see this monstrosity as a Matrox card. They don't detect it, Windows doesn't think the drivers are written for the card, etc.. I naturally tried to force them to work, by selecting the drivers manually, but there were quite a few drivers that fit my card's description (G200, G200 LE, G250, G250 LE.. I went with G250 LE), so I had to just pick.. I'm sick of screwing with the box ATM (as I've done so for thirteen odd hours now), so I'll be back with a vengeance tomorrow to try the other driver options and investigate setting the vendor and device IDs via the flashing tools..

Until then, any further ideas/assistance/comments/etc. are (as always) appreciated.

(Oh, and when this is all sorted I'll be sure to write up a whole page detailing everything I've learned, which is quite a bit..)

Reply 12 of 18, by MaxWar

User metadata
Rank Oldbie
Rank
Oldbie

This thread is over the top. Doing all this crazy stuff; out of this world frankensteined 300mhz pentium with agp + flashed matrox g250, vram replacement etc. Yet no backup rig to test the card in another mobo, 🤣. Hit the flea market buddy 😉

Reply 13 of 18, by sliderider

User metadata
Rank l33t++
Rank
l33t++
MaxWar wrote:

This thread is over the top. Doing all this crazy stuff; out of this world frankensteined 300mhz pentium with agp + flashed matrox g250, vram replacement etc. Yet no backup rig to test the card in another mobo, 🤣. Hit the flea market buddy 😉

When you're dealing with hardware this old it can be hard to find the parts to build a backup rig and expensive if you do. If you live in a country out in the boondocks like Australia or anywhere in formerly communist Europe, it can be even harder.

Reply 14 of 18, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Added is, running a Pentium @ 300Mhz requires to either overclock the AGP/PCI as no Socket 7 CPU was ever made by Intel with the correct multiplier to run @ 300Mhz @ stock AGP/PCI. Either his 300Mhz is overclocked or he found a way to get 4.5x multi and is not telling us 😜

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 15 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

I'm running at 100Mhz FSB with a 3x multiplier if memory serves. ;3
(but memory may not serve, 🤣)
100Mhz FSB + Pentium = Flying Pentium xD

However, this board supports a wide range of multipliers (2x, 2.5x, 3x, 3.5x, 4x, 4.5x, 5x, 5.5x) and FSB 66, 75, 83, 95, 100, 112, 124Mhz.. it's a beast! Also, the board supports all kinds of clock dividers to avoid OC of the buses. Furthermore, I can select the CPU voltage from 2V up to 3.2V by 0.1V increments. It's a FIC VA-503+, one of the best boards of the era. Anywho the board is designed to go upward of K6-III+ chips and all kinds of brands and varieties of other chips besides Intel - it's a SS7.

(manual for those interested, a4_2.pdf is the file with most of the relevant data: ftp://ftp.fic.com.tw/motherboard/manual/socke … va-503+a4sh.zip)

I could build a backup rig easily (sans same hardware, as I have no other of this mobo and GPU, CPU, etc), but there isn't much use since I know what the problems are now. I wanted to try the card in another machine but all the machines I have setup ATM are newer AGP and this card is one that won't fit in anything newer than a 2x (or maybe 4x) slot, no 8x..

*cracks his knuckles and prepares to power on the beast*

Edit: I rewrote the IDs (except the SYS_ID.. couldn't find a good reference for what mine should be, and couldn't get ones I got from the infs to write into the card).. Anyway, I ended up editing the INF files with a catch-all for if there's no SYS_ID and that worked fine. I kept getting horrible garbled graphics - mostly where bitmaps are drawn, less so on window controls and never on textboxes. I read around and found that installing the VIA 4-in-1 could help, and at first that did nothing, and then I reinstalled the Matrox drivers and now it gives me a 0E fatal exception every boot. I googled around for that and found little information.

One thing I do know is that I've got three bus mastering devices - the Matrox card, the VIA chipset, and my IDE controller card. In some threads I saw advice to disable bus mastering on the VIA or the Matrox, and none of these people happened to have my IDE card (Promise 133TX2). I can't figure out how to disable the VIA mastering, and I can't disable it on the IDE card (unless I find a jumper, will check that), so I'm going to toggle the Matrox mastering and switch (at least temporarily) to the mobo IDE.

Reply 16 of 18, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

I forgot about Super 7 🤣!
It means none of the busses are overclocked, which is a good thing 😀

What OS are you running btw? You mentioned little in your OP

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 17 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

98SE is the one I've been trying to get working to no avail. I may try 95, since it's more primitive in some ways, including drivers, and should have less things that can fail. Also, in 95 (maybe in 98/ME? dunno..) you can add resolutions merely by editing the registry, so if the Matrox card doesn't work out I can use whatever card I like and still push out 1080p (provided the card has sufficient VRAM).

This particular moment I'm having trouble getting it to see the optical drive, which is very odd. I had an old caddy-loading drive hooked up to my SB16 in there, which works fine with 98SE - but not with any DOS driver I could find, not the defaults from 95, 98SE, 7.10 custom (banana driver), nor the SB-IDE driver I found.. fail! Since I really wanted this machine to read DVDs anyway, in hindsight, I swapped it for a slot-loading DVD-ROM to match my slot loading 5.25" and 3.5" (yeah.. I know all 3.5" are slot loading, 🤣), as well as my ZIP-250 and LS-120.. this machine will read almost anything! 😁

Anywho, I use a Promise 133TX2 controller in there, the rarer variant of the ATA100 and ATA66 controllers they put out, but at the newer ATA133. I couldn't get DOS to see that either, which was odd. Checked the jumpers and switched to cable select, no joy. I tried the motherboard controller, still nothing.. Puzzling. Anyway I'm about to get back to that (I've not worked on it since just after my previous post).

Edit: To make everything about the machine perfectly clear:

Processor:
- Intel Pentium Tillamook MMX 266Mhz @ 300Mhz
(TDP package, unofficial Chinese conversion - yay I'm special.. xD)

Motherboard:
- FIC VA-503+ Motherboard
(one of the later revisions with both ATX and AT power support, etc., and iffy ACPI support that nobody wants to use anyway)

Memory:
- 512MB SDRAM
(2x256MB DIMMS)
- L2 cache, 512KB or 1MB
(not sure which, mobo shipped with both amounts)

Graphics:
- HP OEM Matrox G250 AGP 2x w/ 8MB & 8MB module that isn't installed ATM)
(G200A dieshrink with higher stock clocks)
- Diamond/3dfx Voodoo 2 8MB
(PCI, chained to the AGP card obviously.. would like to upgrade to dual V2s, perhaps 16MB ones, someday... :D)

Monitor:
- Dell ST2205T
(modern 1080p touchscreen IPS w/o anti-glare layer that I hate)
- IBM 8513 proper VGA
(for testing and such when I'm constantly having to switch between the two systems like in recent days)

IDE Controller:
- Promise Ultra133TX2 ATA133 IDE Controller
(PCI, 2 channels, transparent to DOS - works with other IDE controllers, but I disabled mobo one 'cuz I don't need it)

Sound:
- SB16 Value ISA
(CT2770 - this is my favorite SB16, not too large, not hard to get, takes addon boards and has an IDE header, and configures easily without those stupid TSRs and such that later ones need, and without the jumpering that older ones need ;3)

NIC:
- 3com 3c509b PCI 10/100 NIC
(I use the ISA version(s) of this card in older machines, and PCI in newer, for some reason I have like 30 of these in various forms, so it's easier to keep track of when they all use the same software! :D)

Removable Disk Drives:
- Teac FD-505 dual floppy 5.25 half-height drive
(1.2MB 5.25" & 1.44MB 3.5" in one bay ftw.. A: is 3.5", B: the 5.25")
- Pioneer slot-loading DVD drive
(on the Promise controller, first channel master)
- LS-120 IDE
(on the Promise controller, second channel master)
- ZIP-250 IDE
(on the Promise controller, first channel slave)

HDD:
- 80GB Samsung HDD
(on the Promise controller, second channel slave - zomg I broke tradition! :D - may well add more disks upward of a 750GB IDE drive (woot) once this system's fully functional)

Remaning Expansion Slots:
- One free ISA slot
- One unusable shared free ISA/PCI slot that's in use as PCI

OS:
Windows 98SE & it's original DOS - dysfunctional ATM.
(sometimes I set it up with my own DOS, like MS-DOS 8.0 unofficial)

I think I covered everything, probably way more than you'd need to assist, but I love to talk about my "baby", heh.. If you need any more info don't hesitate to ask.

By the way - do you pronounce "Tillamook" as "till-a-mook" or "ti-ya-mook"? I'm not sure of it's origin, so I dunno, myself.

[spoiler]Why don't we have this? 😜[/spoiler]

Reply 18 of 18, by Yushatak

User metadata
Rank Member
Rank
Member

Figured out the optical drive problem - apparently NOTHING you can do will make DOS (or the BIOS CDROM boot option) see an optical drive (or removable disk that can't act as an HDD, likely) on a Promise controller, sans drivers that I haven't located or may not exist. Windows can and will see these things with (or maybe even without) the driver installed, but DOS won't, so I hooked the drive up to the internal headers and re-enabled them. Installing Windows 95 for testing now..