Check the connectivity between the two white ports that are connected via ribbon cable.

Have you tried VGA passive adapter on DVI port ?

This card seems to me to just digitize whatever would go out on the VGA port. Maybe neither of those monitors accepts digital in at 75Hz? Boot/BIOS screens, DOS "text mode", Win9x splash would be 75 and might now show up because of that.

Can you connect through VGA, boot Windows, then check whether it outputs at 60Hz? It already should, but best to make sure. In case it doesn't, set it to 60, then try again through DVI, giving it enough time to boot to the Windows desktop.

I have several cards with that SIL chip. For DVI digital output if it gets EDID that it doesn't like there won't be an image. I solved this by using DVI EDID emulator between card and display (StarTech VSEDIDDVI). When connecting to capture card (DataPath Vision-E1S) I had to tweak default EDID to max out at resolution that chip supports. So yeah, it should work but it is picky about what monitor reports in EDID.

Thanks for the input everyone! I appreciate it.

Tree Wyrm wrote on 2024-05-05, 20:38:

I have several cards with that SIL chip. For DVI digital output if it gets EDID that it doesn't like there won't be an image. I solved this by using DVI EDID emulator between card and display (StarTech VSEDIDDVI). When connecting to capture card (DataPath Vision-E1S) I had to tweak default EDID to max out at resolution that chip supports. So yeah, it should work but it is picky about what monitor reports in EDID.

Sadly, that's what I was suspecting. At $80 I don't think the EDID emulator is a feasible option for this card. I was really hoping these would make for easy drop-in DVI-ready cards to use in any old AGP system, but it seems you'd need a fairly old monitor anyway.

I do have some 1280x1024 monitors in storage so I'll pull one of those out tomorrow and try it. I'm sure it'll work.

... personally I hate the existence of 1280x1024 as a standard resolution since it isn't 4:3 and yet computers seem to pretend that it is... but, well... whatever... :p

As a side note:
EDIT: Rivatuner seems to be reporting clocks incorrectly. TNTCLK detects it properly as 150\183.

I tried to see if I could get these to run at 150Mhz since the core certainly should be able to, but I can't get Rivatuner to apply the overclock settings. No matter what I do, after I test the clock and then apply\OK the new speed it just reverts everything back to 100\183. I have googled it but understandably there is very little information still online regarding overclocking pre-geforce cards with Rivatuner. There are tons of settings in the "Power User" section of the program and I have tried changing some that were mentioned in old posts I found, but nothing is letting me increase the clock speed of these cards.

Anyone know what's going on here? Rivatuner was always my go-to overclocking program back in the post-3dfx days before MSI Afterburner and EVGA Precision (all based on the same code base of course).

Bumping this old thread because a weird idea came into my brain, and I can't seem to locate the information I need to sufficiently stomp out that idea. I'll try to be brief.

Situation:

• The TNT2 Ultra is likely the fastest\newest card that has full driver support for Windows 3.11 while also powering through DOS games and handling 3D accelerated pre-2000 games quite well.
• Is there enough use for such a card to warrant swapping the Silicon Image DVI transmitter chips with newer pin-compatible ones that can handle more modern 1080P digital displays?

The Sii154 chips on these cards are limited to 1280x1024 and will show no display if the monitor's native res is higher than that, regardless of what resolution you're trying run. The slightly newer Sii164 will handle up to 1600x1200 and is pin-compatible as far as I can tell, but that wouldn't really open up a whole lot more options.

Sadly, it seems that Silicon Image no longer provided public datasheets of any of the digital video transmitters they made after this point... which is super frustrating!!

After a fair amount of digging, it seems that the Sii1930CTU comes in a TQFP-64 package and may be aimed at replacing these older chips, while supporting 1920x1080 displays. I cannot verify that they are pin compatible though, because there are no datasheets. The chips also seem to be extremely hard to find. The only lead I have been able to find is that the 1st Gen Apple TV (from 2007) used this chip, so they can likely be found in those.

Am I crazy? I see this as an interesting challenge, and I have more than one of these cards to tinker with. Potentially being able to hook one of these up to a 1080P monitor or TV with a crisp digital connection and run a fully accelerated Windows 3.11 at native resolution just seems neat. I don't believe anything else available can do that... I just don't know if it is a thing worth doing. 1080P at 60Hz is already decently clear on monitors that still have VGA ports, as long as a good cable is used.

If there is absolutely zero interest in such a card, I probably won't go out of my way to mess with this. A simple datasheet or any information that would show that the Sii154CT64 and Sii1930CTU aren't pin-compatible would also end this craziness.

I just can't shake the idea of making this card exist if it could be as simple is desolding the chip with a hot air gun and putting a different one in its place.

Halp!

What happens when you try FullHD resolution, for instance in Windows 3.1x with these drivers: Re: Windows 3.11 - nVidia TNT2 built-in GPU chip compatibility

Sil164 works at 1920x1080@60Hz. You can also use TFP410 instead, pin compatible. IIRC, the only difference is the HPD pin behaviour, opposite to Sil164. Irrelevant if the GPU monitors the DVI HPD signal instead of the TMDS encoder provided one.

sdz wrote on 2024-11-16, 11:00:

Sil164 works at 1920x1080@60Hz. You can also use TFP410 instead, pin compatible. IIRC, the only difference is the HPD pin behaviour, opposite to Sil164. Irrelevant if the GPU monitors the DVI HPD signal instead of the TMDS encoder provided one.

Are you sure about the Sii164 working at 1920x1080? I have found posts saying it doesn't.

SGI V3 GeForce 256 and DVI
Re: SGI V3 GeForce 256 and DVI

... but then this post says it should do 1920x1080, but not 1920x1200, which isn't a big deal.
DVI resolution limits of various old GPUs

I would be totally content with those capabilities. I doubt it's possible to get anything more flexible that would fit in the same pinout (1080@120hz, 1440p, 4k, etc.).

Still not sure if this is worth doing or not though.

EDIT: Wow, looks like the TFP410 should be even more flexible, with official WUXGA (1920x1200) support. Honestly, I will probably go with those if I choose to attempt this.

I wonder if the TI chips have a more graceful handling of situations where the native res from the display is higher than expected. If it was possible to hook a TNT2 Ultra up to any modern display via a DVI to HDMI adapter, that would be awesome.

Yes, I'm sure Sil164 works at 1920x1080@60. It supports 165MHz pixel clock, same as the TFP410, which also supports 1920x1080@60.
I also used both, and got 1080p out of them.

"I would be totally content with those capabilities. I doubt it's possible to get anything more flexible that would fit in the same pinout (1080@120hz, 1440p, 4k, etc.)."
I doubt that many of the old GPUs can even output that on the digital RGB bus.

"I wonder if the TI chips have a more graceful handling of situations where the native res from the display is higher than expected."
What do you mean? The Sil164 or TFP410 TMDS encoder have no clue what display you have connected to them. They are also not a scaler. What comes in on the parallel RGB bus comes out as DVI, nothing more.

sdz wrote on 2024-11-16, 18:51:

"I would be totally content with those capabilities. I doubt it's possible to get anything more flexible that would fit in the same pinout (1080@120hz, 1440p, 4k, etc.)."
I doubt that many of the old GPUs can even output that on the digital RGB bus.

I don't really need them to run at those resolutions. They just have to function when connected to such displays. See below.

sdz wrote on 2024-11-16, 18:51:

"I wonder if the TI chips have a more graceful handling of situations where the native res from the display is higher than expected."
What do you mean? The Sil164 or TFP410 TMDS encoder have no clue what display you have connected to them. They are also not a scaler. What comes in on the parallel RGB bus comes out as DVI, nothing more.

This thread and some of the others linked earlier were made specifically to discuss an issue with older encoders where they would show no signal when connected to some displays that report a native res (via EDID) higher than the encoder supports. It's possible that not all card + display combinations have issues with this, but it seems common enough to warrant a work-around or upgrade for the cards that have the issue.

This post explains it a bit:
Re: No output from DVI ports on any ELSA Erazor III 32MB TNT2 Pro
This thread also discusses it:
SGI V3 GeForce 256 and DVI

Sorry, I didn't read the whole thread. If the display reports a native resolution via EDID higher than that the video card supports on the DVI output, and a supported resoltution is set, and it results in a black screen, then replacing the TMDS encoder won't do anything. This is a GPU/VBIOS/driver issue, and not related to the TMDS encoder at all.
So it's either hacking the VBIOS/driver or using an EDID emulator.

sdz wrote on 2024-11-16, 22:07:

Sorry, I didn't read the whole thread. If the display reports a native resolution via EDID higher than that the video card supports on the DVI output, and a supported resoltution is set, and it results in a black screen, then replacing the TMDS encoder won't do anything. This is a GPU/VBIOS/driver issue, and not related to the TMDS encoder at all.
So it's either hacking the VBIOS/driver or using an EDID emulator.

It will probably be over my head, but can you provide any information (datasheet, forum thread, etc.) that explains why this is the case? Does the TMDS encoder not interact with EDID at all?

You can find a datasheet here:
https://www.ti.com/product/TFP410
The TMDS encoders do not interract in any way with EDID. RGB data + sync in, DVI data out, that's it.
The DDC lines (for EDID) are routed on a video card directly to the GPU (or through a level shifter). They don't go anywhere else.

sdz wrote on 2024-11-16, 22:51:

You can find a datasheet here: https://www.ti.com/product/TFP410 The TMDS encoders do not interract in any way with EDID. RGB da […]
Show full quote

You can find a datasheet here:
https://www.ti.com/product/TFP410
The TMDS encoders do not interract in any way with EDID. RGB data + sync in, DVI data out, that's it.
The DDC lines (for EDID) are routed on a video card directly to the GPU (or through a level shifter). They don't go anywhere else.

Wow, great information. Thank you!

I think I had been misunderstanding what the function of the TMDS encoder is. I had been thinking of it as taking the VGA signal and converting it to a digital format, but that doesn't make sense... why convert digital to analog and back to digital? Derp.

Now that I actually think about the existence of "data" lanes on the TMDS encoder, it makes a lot more sense to me. For a graphics card to have a DVI port, the chip\GPU actually has to be capable of outputting the display data (digital) to a TMDS encoder. The TNT2 apparently does this, but it seems to be quite picky or easily broken by unexpected EDID.

So, even though the VGA port is directly next to the little ribbon cable that jumps to the TMDS\DVI portion of the board, they have completely different signals. If there is no output on the DVI connector when there should be, the TNT2 itself is "confused" somehow and is not sending the required data to the TMDS, even though VGA would work in the same situation.

So, unless this can be fixed with a BIOS update, this is just a quirk of the TNT2 in general... or at least the TNT2 Ultra (which is a bit older than, for example, the TNT2 Pro).

I wonder if there are any TNT2 cards with DVI ports that will function on newer monitors? I see this Creative CT5828 has the SiI164: https://x.com/Foone/status/955968645955436544/photo/1

EDIT: Final question... is it even remotely possible that the TNT2 is simply sending data that this ancient SiI154 doesn't like, so it isn't working? Basically, the issue I had mentioned before, but with the bad\incompatible data coming from the TNT2 rather than the monitor. In this case, a different TMDS encoder may understand this data and function properly? Just throwing this out there.

Interesting... I had ELSA Erazor X GeForce 256 with VGA and DVI output. I couldn't get the DVI to work and just thought it was broken.

kixs wrote on 2024-11-17, 00:37:

Interesting... I had ELSA Erazor X GeForce 256 with VGA and DVI output. I couldn't get the DVI to work and just thought it was broken.

Do you know which monitor(s) you had it connected to?

EDIT: Just found your thread with the exact same problem.
ELSA Erazor X with DVI

I see that yours uses the SiI164CT64, which is the slightly newer one.

Maybe this is an ELSA specific issue? Perhaps flashing the vBIOS to something other than ELSA would be worth trying.
This article mentions that ELSA does some funky things with their cards (down at the bottom) that can impact 1600x1024 resolution specifically... maybe the issue in this thread is another side effect of whatever they did to their cards.

Ozzuneoj wrote on 2024-11-17, 00:20:

Wow, great information. Thank you! […]
Show full quote

sdz wrote on 2024-11-16, 22:51:

You can find a datasheet here: https://www.ti.com/product/TFP410 The TMDS encoders do not interract in any way with EDID. RGB da […]
Show full quote

You can find a datasheet here:
https://www.ti.com/product/TFP410
The TMDS encoders do not interract in any way with EDID. RGB data + sync in, DVI data out, that's it.
The DDC lines (for EDID) are routed on a video card directly to the GPU (or through a level shifter). They don't go anywhere else.

Wow, great information. Thank you!

I think I had been misunderstanding what the function of the TMDS encoder is. I had been thinking of it as taking the VGA signal and converting it to a digital format, but that doesn't make sense... why convert digital to analog and back to digital? Derp.

Now that I actually think about the existence of "data" lanes on the TMDS encoder, it makes a lot more sense to me. For a graphics card to have a DVI port, the chip\GPU actually has to be capable of outputting the display data (digital) to a TMDS encoder. The TNT2 apparently does this, but it seems to be quite picky or easily broken by unexpected EDID.

So, even though the VGA port is directly next to the little ribbon cable that jumps to the TMDS\DVI portion of the board, they have completely different signals. If there is no output on the DVI connector when there should be, the TNT2 itself is "confused" somehow and is not sending the required data to the TMDS, even though VGA would work in the same situation.

So, unless this can be fixed with a BIOS update, this is just a quirk of the TNT2 in general... or at least the TNT2 Ultra (which is a bit older than, for example, the TNT2 Pro).

I wonder if there are any TNT2 cards with DVI ports that will function on newer monitors? I see this Creative CT5828 has the SiI164: https://x.com/Foone/status/955968645955436544/photo/1

EDIT: Final question... is it even remotely possible that the TNT2 is simply sending data that this ancient SiI154 doesn't like, so it isn't working? Basically, the issue I had mentioned before, but with the bad\incompatible data coming from the TNT2 rather than the monitor. In this case, a different TMDS encoder may understand this data and function properly? Just throwing this out there.

You are correct, the DVI output has nothing to do with the VGA output. The GPU has a 12bit DDR or 24bit SDR data bus to the TMDS encoder, plus HSYNC, VSYNC, BLANK signals.

Ozzuneoj wrote on 2024-11-17, 00:20:

EDIT: Final question... is it even remotely possible that the TNT2 is simply sending data that this ancient SiI154 doesn't like, so it isn't working? Basically, the issue I had mentioned before, but with the bad\incompatible data coming from the TNT2 rather than the monitor. In this case, a different TMDS encoder may understand this data and function properly? Just throwing this out there.

It is possible, but unlikely. You can probe the data inputs of the TMDS encoder if you have an oscilloscope. I don't have any of these cards to test.

Well, after much (much!) messing around and having to scour the internet for long-forgotten files, it doesn't seem that flashing a standard Nvidia TNT2 Ultra 3.05 BIOS to this ELSA card will fix the DVI issues.

Thankfully, flashing the BIOS (using nvflash318 in DOS) did not brick the card... it still worked fine. So, I would say that if there are any driver related quirks with ELSA cards, they can pretty easily be flashed to a reference BIOS.

Anyway, this seems like a weird problem, but I guess it's just how these ELSA cards are. They seem to be quite picky about what they are connected to via DVI.

EDIT: Interestingly, one of the alternate BIOS files I found (there are four variations) actually behaves differently with DVI connected. The system boots normally, but the display is never initialized. With all of the others the system will actually hard lock when the DVI cable is connected. So, there does seem to be SOME difference between the various BIOS files and versions. It may be worth trying different ones then...

I am using the files from the TNT Ultra 3.05.00.10 zip from this page:
https://web.archive.org/web/20050211101639/ht … whitebunny.net/
The one that basically ignored the DVI connection was the 3.05.00.10.0 "SM" version.

There are also some older versions here I will be trying next:
http://www.x86-secret.com/articles/nvbios.htm
It looks like the variations have something to do with what other outputs the card has. So in the link from "whitebunny.net" the file ending in CH is probably Chrontel, BT is probably Brooktree. I'm not sure but PH and SM may be Philips and Silicon Motion.