VOGONS


First post, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

I have been expanding my video card collection recently and have come across a frustrating reoccurrence within my testing. Sometimes cards which are perfectly functional on one system will refuse to work on another. But this isn't just random; the boards that have problems are always consistent with which cards they reject.

For example, I have a set of "problem" cards that when paired with either my ASUS P2B v1.10 Slot 1 Intel 440BX or my ASUS P2E-M Slot 1 440EX motherboard will produce BIOS error beeps indicating that no video card is connected. This is not necessarily the age of the chipset however. All of these cards work fine in an even older Socket 7 430TX motherboard.

The range of cards I've found this problem to exist spans many years. Cards which I've identified the problem include an NVIDIA Rival TNT2 m64, GeForce FX 5200, GeForce GT610, and a recently manufactured ATI Rage 128 XL, as pictured below. I have encountered a few others but this is a short list of cards I can remember offhand.

This not necessarily limited to PCI cards as I have one example of an AGP card failing to work on these boards but working fine in another even when it should be an identical model. I have also experienced similar issues with PCIe cards failing to work in one motherboard but working fine in another.

Whether the same result is due to the same problem among PCI, AGP, and PCIe video cards, it's very frustrating whenever I encounter this issue but one thing is for certain. Don't throw away a card without testing it on other systems.

Does anyone have any experience to share on this subject or any knowledge about what could be causing this problem?

IMG_20210922_002825_resize_29.jpg
Filename
IMG_20210922_002825_resize_29.jpg
File size
484.84 KiB
Views
630 views
File license
Public domain
IMG_20210922_002848_resize_98.jpg
Filename
IMG_20210922_002848_resize_98.jpg
File size
625.12 KiB
Views
630 views
File license
Public domain
IMG_20210922_002812_resize_89.jpg
Filename
IMG_20210922_002812_resize_89.jpg
File size
477.08 KiB
Views
630 views
File license
Public domain
IMG_20210922_002836_resize_38.jpg
Filename
IMG_20210922_002836_resize_38.jpg
File size
551.52 KiB
Views
630 views
File license
Public domain

Reply 1 of 24, by framebuffer

User metadata
Rank Newbie
Rank
Newbie

I guess that there is not one single reason but several (and they can also be combined randomly), like MB/VGA BIOS limitations/issues/features or "hidden" hardware problems, like power delivery/draw and so on
Also, back in 90s and early 2000 hw compatibility was not a given thing in general

https://framebuffer.io

Reply 2 of 24, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

The 440EX and 440BX that I have this issue on are both Award BIOS. I don't have a problem on a 430TX with Intel BIOS or a VIA Apollo board.

My sample size is very small but I wonder if this is related.

The GT610 is also the first card I found with this issue which also has a fan on it. When system tries to boot the CPU fan will spin but not the fan on the graphics card. Maybe there is some issue with the card not being able to get the voltage it needs.

Last edited by Kahenraz on 2021-09-23, 16:43. Edited 1 time in total.

Reply 3 of 24, by 386SX

User metadata
Rank l33t
Rank
l33t

Imho some of these modern PCI cards, not to mention the GT610, were built for some of the last decade mainboards/chipsets/bios and not not nearly intended for older ones. So as said above, the problems might be more than one and for example the same power draw from the PCI bus might be a problem. I would not even try such card on some early PCI mainboards for sure, who knows what might happens even on the mainboard stress considering the single GT610 basically might ask all the PCIs power wattage. And so on for the older cards and previous mainboards specifications. Also these cards has different bridge chips to interface the PCI bus with AGP or PCI-E GPUs. On the mainboards there's another bridge to have a "compatible" PCI connection. This add another variable to the result. One thing is a video card having a native PCI oriented video chip, another is using these bridged versions not originally intended for it. And as tested lately with all sort of strange results even depending on the o.s tested.
And of the one posted only the Rage XL I think should be intended for the PCI bus too, but in these modern PCB layouts who knows what's changed from a time correct Rage PCI card to know if something might make it less compatible than the original PCB layout tested by ATi in the past on multiple time correct mainboards. Even original cards were released on multiple revisions to correct or improve things, I suppose these modern PCI cards that are nice to see them sold again, anyway might not have the same compatibility as a priority.

Reply 4 of 24, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Welcome to the 90s where the standards were all over the place and adhering to standards was not a hardware quality you could take for granted, but a point of extra praise in reviews.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 5 of 24, by 386SX

User metadata
Rank l33t
Rank
l33t

Back in the 90's it sure was a complex situations but beside any nostalgic retro hardware passion, I'd say I might prefer the multi companies multi standards multi products situations of those times than the modern completely standardized with mostly no differences (and mostly few consumer choices) components situation. At least there was something interesting to see new efforts from so many companies trying to create something new even when not having a long lifetime support. Let's think to how many video chips and cards came out in the 95-2000 period even the lowest end ones. Nowdays it'd look like a dream having such availability in components choice. 😉

Reply 6 of 24, by zapbuzz

User metadata
Rank Member
Rank
Member

I was lucky to get an Nvidia GT card it has molex power so doesn't rely souly on weird motherboard voltages for correct power at high speed. Also due to the age of these pc's typically allows more stable function.
Where the discrete bus powered cards don't work above 2x (mostly AGP 4X to 8x speed) I recommend try the GT card variety they may be harder to get because are favoured for gaming and in constant use.

Reply 7 of 24, by BitWrangler

User metadata
Rank l33t
Rank
l33t
appiah4 wrote on 2021-09-23, 10:09:

Welcome to the 90s where the standards were all over the place and adhering to standards was not a hardware quality you could take for granted, but a point of extra praise in reviews.

Heh though they'd typically give a low key nod to it being compatible with standards, but then go long and loud about how it was a whole 10% slower than the competition.

2017: Basement full of ancient PC stuff, starting to go through it. 2021: Still starting, heh, many setbacks. So what's this BitWrangler guy's deal ??? >>> Taming the pile, specs to target?

Reply 8 of 24, by appiah4

User metadata
Rank l33t++
Rank
l33t++
BitWrangler wrote on 2021-09-23, 16:28:
appiah4 wrote on 2021-09-23, 10:09:

Welcome to the 90s where the standards were all over the place and adhering to standards was not a hardware quality you could take for granted, but a point of extra praise in reviews.

Heh though they'd typically give a low key nod to it being compatible with standards, but then go long and loud about how it was a whole 10% slower than the competition.

So true. It was amazing how people flocked to out of spec and completely unstable motherboards just because they were 10% faster (when they worked) and complained about blue screens afterwards..

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 9 of 24, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

Note that this list includes a GeForce TNT2 M64 which is period-correct.

I included the other cards as an example of the range of cards which exhibits this issue. So far it only seems to affect Award BIOSes but the sample size is still very small.

Reply 10 of 24, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

I think the root of the problem in your case is how fast the card can be initialized by BIOS. Some of these cards just fail the timing and thus motherboard can't see them on boot. You can test it with two video cards, one of which you know is working. If both are recognized successfully in Windows, then it's BIOS problem.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 12 of 24, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

My ASRock Core2 Dual VSTA motherboard doesn't like Geforce 3 cards for some reason. I haven't yet tested any Geforce 4 Ti cards yet, but the Geforce 4 MX cards work fine.

Could have something to do with the AGP bus being overclocked but 3 different brands and models all exhibited the same exact issue. POST screen is fine, then the screen goes blank when loading into Windows XP though the screen still says it is active... an NEC 19" 1600x1200 CRT.

It is really weird because older and newer nVidia cards work just fine. Have tested from TNT2 Vanta 16 all the way up to Geforce FX 5950U.

Put those same cards in my PIAGP Pentium 4 system and they boot and test perfectly fine in XP though I do pretty much expect that the PIAGP industrial system will probably not have problems with any AGP card due to the industrial systems having a much more stringent testing process.

Yamaha YMF modified setupds and drivers
Yamaha XG resource repository - updated November 27, 2018
Yamaha YMF7x4 Guide
AW744L II - YMF744 - AOpen Cobra Sound Card - Install SB-Link Header

Reply 13 of 24, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

I have more research to do on this but someone suggested that this may be a result of these boards not providing +3.3V on the PCI bus. This is consistent with a few limited tests I've done but I haven't performed verified this on all of my boards.

For motherboards which do not have +3.3V to the PCI bus, it may be possible to add jumpers on the back of the board from the AGP slot. This would provide filtering that would not otherwise be available if connected directly to the +3.3V rail of the ATX connector.

I have very little knowledge of electronic circuits and my experience is primarily in repair. Would a modification like this have the risk of damaging traces from too much current? Does anyone know of a safe way to wire something like this?

I've also been told that +3.3V was not implemented on many of these earlier boards as it wasn't a requirement of the specification but one way to identify it is to look for the color-coded ports on the back of the board such as the purple and cyan PS/2 connectors. This should be around the time that this voltage was a required specification, I believe.

Reply 14 of 24, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

may be a result of these boards not providing +3.3V on the PCI bus

Not an issue on ATX 440BX motherboard.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 15 of 24, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

Neither of my two 440BX motherboards provide +3.3V on the PCI bus including my ASUS P2B revision 1.10. Neither do my 440EX or 440LX boards.

I wasn't looking for this until now but I had to switch to a VIA board specifically for this voltage.

What 440BX boards can you confirm with a POST code diagnostics card have this voltage on the PCI bus?

Reply 16 of 24, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

ASUS P2B revision 1.10. Neither do my 440EX or 440LX boards.

All ASUS, I assume? BTW TNT2 m64 on your photo should work just fine in +5V PCI slot.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 17 of 24, by Kahenraz

User metadata
Rank Oldbie
Rank
Oldbie

It does not. It is the most interesting one to fail.

We already know that the newly minted Rage XL cards from China have a similar problem. I have a genuine ATI Rage XL coming in the mail I want to verify against to see if it's a change in design or whether it's a fault that was directly copied. The card is old enough that it should work without the 3.3V.

Reply 18 of 24, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

It does not. It is the most interesting one to fail.

Then it's not +3.3v problem, but BIOS problem.

The card is old enough that it should work without the 3.3V.

If it doesn't have buffer chips - chances are slim.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 19 of 24, by pentiumspeed

User metadata
Rank Oldbie
Rank
Oldbie

Nearly all consumer motherboards that use PCI slots are 5V. Due to the notch location that plastic slot is located.

The server boards usually had 3.3V PCI slots especially if it has PCI-X.

That said, I find that many PCI video cards will not work varies with some boards, usually earlier ones has more issues. Better ones is not necessarily true. At first, I bought Micronics HX based board and did not work with many video cards.
Then bought another motherboard based on TX chipset and even both are PCI 2.1 revision, TX had better support in bios for later PCI video cards.
Try and see.

Cheers,

Great Northern aka Canada.