Shes Dead Jim, no seriously this looks to be far worse than some dead Vram, likely the BGA balls under the GPU die have cracked, these are very simple cards so there isn't much that can go wrong with them. (usually its dead Vram which still displays an image but with corruption, fully corrupted image is likely the GPU die itself)

Also just because it was in a box out of the elements doesn't mean you can rule out dead Vram or a dead GPU die, solder ages and dries out becoming brittle. Solder from the time of this card is even more prone to such behavior and this card doesn't look like a more expensive model that would have used higher quality solder. Electronic components also age and the shock of being powered up after 20 years of sitting in a box is a great catalyst for such behaviour, just ask anyone who has powered up old motherboards with tantalum caps.

If you get a signal at the intended resolution (check the OSD of your monitor), the BIOS is most likely completely fine, and the primary issue for the completely broken picture is expected to be between the graphics chip and the video RAM chips. You might want to double-check the video RAM chips for bent pins that short to each other, but otherwise, I have to agree that the card is likely beyond economical repair at the moment. When people start paying $15k for TNT2 cards, this card might get worth repairing, though.

Reflash the BIOS anyway, won't hurt.

Thank you very much, guys, for the given directions.

Honestly, I'm wondering what may cause cracks in the solder joints under the BGA chip when this card was just sitting in plastic bag in a cardbox for years. The slightly newer one (GeForce 2 MX) has no such a problems. I'm 200% sure that this card was in working condition, before I've put it in "storе" it with the rest of the obsolete parts. I just don't keep defective hardware in cardboxes - I either repair it, or if it can't be repaired it goes for spare parts/in trash bin. But who knows... maybe Trashbytes is right.

@mkarcher, resolution of display is fine - got menu on the screen and the "mess" stays for background. No bent pins on the chips, I've inspected everything on the PCB with magnifying glass, just to be sure. No leaky caps, etc.

I may try hot air gun with plenty of flux near the GPU, but this will be the last tryout.

@Nexxen, I think to start with BIOS reflash, but I don't know where to find BIOS dump for the exact model (Inno3D RIVA TNT2 Pro/32M). I also don't know which utility will be the most suitable for reflash - NVflash, flashrom(?), uniflash(?)...

I know its not a "treasure" card, but will be nice to be in working condition again.

analog_programmer wrote on 2023-06-27, 17:21:

I think to start with BIOS reflash, but I don't know where to find BIOS dump for the exact model (Inno3D RIVA TNT2 Pro/32M). I also don't know which utility will be the most suitable for reflash - NVflash, flashrom(?), uniflash(?)...

You can find some BIOS dumps for the TNT2 Pro on this page.

No idea what to use for flashing, but if you have the original driver CD for your card, those usually contain both the original BIOS and flashing tool.

Joseph_Joestar wrote on 2023-06-27, 17:32:

You can find some BIOS dumps for the TNT2 Pro on this page.

No idea what to use for flashing, but if you have the original driver CD for your card, those usually contain both the original BIOS and flashing tool.

Thanks for the link, mate! I see there's a picture of the same PCB as on mine card, but from another manufacturer. I didn't know that the BIOSes for different manifacturers of these videocards are somehow unified.

Sadly I do not have original CD with drivers and utilities. I'll have to dig the web... 😀

You seem to have the same symptoms I get on a Diamond Viper V770 TNT2 card. The system boots and everything acts like it thinks it's working but the display is severely garbled, like what you showed. I was even able to boot a linux CD and see signs of a working mouse pointer. I wonder if this is a common failure mode on TNT2.
No idea about the repair though. I bought some RAM chips a couple years ago thinking I'd try replacing them someday but I haven't tried it.

It's disappointing to find out that it might be a GPU issue. My thinking has been that a GPU issue wouldn't leave the card in a still "logically" functioning state, so it must be RAM. Maybe I'm wrong about that.

I've had several Geforce2 MX cards from random lots and they always seem to work. They're very reliable, perhaps because they are low power. TNT2 is a hotter chip.

shamino wrote on 2023-06-28, 00:23:

You seem to have the same symptoms I get on a Diamond Viper V770 TNT2 card. The system boots and everything acts like it thinks […]
Show full quote

You seem to have the same symptoms I get on a Diamond Viper V770 TNT2 card. The system boots and everything acts like it thinks it's working but the display is severely garbled, like what you showed. I was even able to boot a linux CD and see signs of a working mouse pointer. I wonder if this is a common failure mode on TNT2.
No idea about the repair though. I bought some RAM chips a couple years ago thinking I'd try replacing them someday but I haven't tried it.

It's disappointing to find out that it might be a GPU issue. My thinking has been that a GPU issue wouldn't leave the card in a still "logically" functioning state, so it must be RAM. Maybe I'm wrong about that.

I've had several Geforce2 MX cards from random lots and they always seem to work. They're very reliable, perhaps because they are low power. TNT2 is a hotter chip.

Cracked BGA joints can leave it in a mostly working state, would only take one or two damaged balls to produce a garbled display, one option is to remove all the plastic parts and heat sink and hit the gpu die with a hot air gun for ~30 seconds from both sides. chances are that it may reflow it enough to start working correctly again. (This card thankfully doesn't have any SMD parts on the back making this an easy thing to try)

I would also check all the SMD components that are on the traces leading to the VGA port, one of them may have died preventing a clear signal reaching the VGA port.

Its not a permanent fix however and much like the baking fix for G80 cards it'll eventually break again but if you can get a proper picture from the card you can then diagnose if the Vram is also bad and if its worth trying to fix the card.

shamino wrote on 2023-06-28, 00:23:

You seem to have the same symptoms I get on a Diamond Viper V770 TNT2 card. The system boots and everything acts like it thinks […]
Show full quote

You seem to have the same symptoms I get on a Diamond Viper V770 TNT2 card. The system boots and everything acts like it thinks it's working but the display is severely garbled, like what you showed. I was even able to boot a linux CD and see signs of a working mouse pointer. I wonder if this is a common failure mode on TNT2.
No idea about the repair though. I bought some RAM chips a couple years ago thinking I'd try replacing them someday but I haven't tried it.

It's disappointing to find out that it might be a GPU issue. My thinking has been that a GPU issue wouldn't leave the card in a still "logically" functioning state, so it must be RAM. Maybe I'm wrong about that.

I've had several Geforce2 MX cards from random lots and they always seem to work. They're very reliable, perhaps because they are low power. TNT2 is a hotter chip.

Тhanks for sharing the information about exactly the same problem with this generation of videocards. I had no idea it could be a common problem. Maybe they'll become more expensive (as those in working condition will drastically drop in numbers with time) sooner than expected 😁 But I have some hope, that it's not physical damage to the BGA solder joints or physically degraded with the time VRAM.

Trashbytes wrote on 2023-06-28, 02:47:

Cracked BGA joints can leave it in a mostly working state, would only take one or two damaged balls to produce a garbled display, one option is to remove all the plastic parts and heat sink and hit the gpu die with a hot air gun for ~30 seconds from both sides. chances are that it may reflow it enough to start working correctly again. (This card thankfully doesn't have any SMD parts on the back making this an easy thing to try)

I would also check all the SMD components that are on the traces leading to the VGA port, one of them may have died preventing a clear signal reaching the VGA port.

Its not a permanent fix however and much like the baking fix for G80 cards it'll eventually break again but if you can get a proper picture from the card you can then diagnose if the Vram is also bad and if its worth trying to fix the card.

Everything you wrote has its reason. But today I intend to try with BIOS reflashing fix. And I'll explain why I keep some hope that there's a chance to avoid hot air gun treatment.

Sometimes hardware gives symptoms like its broken, but everything is software related (BIOS). Here's fresh example - not related to this NV TNT2, but another onboard video with dedicated VRAM:

In this very same box of antiques I found an old IBM Aptiva mobo with ATI Rage Pro Turbo AGP videochip and 4 megs of dedicated SGRAM. It is in fully working condition. On this board the original BIOS from IBM only supports the first generation PII Klamath CPUs (it lacks microcode for Celerons or newer PIIs) and works flawlessly with them (I still have one unlocked PII 266 that runs fine with 3* 100MHZ FSB). When I put in its slot some old Celeron Mendocino (I have Celly 333) the mobo POSTs (BIOS complains about mismatched CPU code (missing microcode) but everything is fine), loads OS, etc., except the fact that when the onboard ATI Rage is used in random time period after the system was booted the picture on screen becomes like patterned carpet (VRAM problem symptom) and the system freezes. But when this mobo is used with same unsupported from the IBM BIOS Celeron and some PCI videocard (no AGP slot is available) there's no such a problem. So I always thought for years (until now) that it has some defective onboard VRAM chip. These days I decided to solve the problem with unsupported newer (than Klamath core) Mendocino CPUs and eventually with newer Deschutes core PIIs... I managed to dig-up from the official Fujitsu(-Siemens) support site a newer BIOS image file for machine (thank you Fujitsu for your long-term support, but why the heck some things are so deep buried?!) with exactly the same mobo (Acer V66XA - Acer has the most terrible support, literally there is no support for like 10 years old machines and laptops) that is with support for these newer CPUs. I reflashed the IBM BIOS (it was not an easy process due to some checksum check restrictions in non-standard phoenix-like BIOS, but finally I've managed to get there) with the BIOS for Fujitsu-Siemens machine and now everything with Celly 333 Mendocino is OK during boot process. And the most interesting part - no more VRAM-like problems with integrated ATI Rage when used with Mendocino core Celerons!

So this gives me some hope that maybe this well preserved videocard has some "bit rot" problem with its BIOS. Recently I watched some videos on YT with degraded BIOS and reflash BIOS repairs (BitsUndBolts YT channel).

I have to find some suitable flash software (years ago I used NVflash in pure DOS to update BIOS on some newer than TNT2 videocards). Today I'll search for such a software. I'm not sure if Flashrom supports this card (on it site it is marked with ??? for untested, but maybe working) either not sure for Uniflash (It is too old).

My plan is to reflash BIOS with known dump from good card - Joseph_Joestar shared useful link here. Then if this does not give some positive result, I'll try to resolder VRAM chips and some other solder joints on the PCB. As a last resort - GPU hot air gun trick. At least trash bin is always an option 😀

For now I will not try to destroy this card with physical interventions 😀

Ive seen cards with a dead Vbios before and most of the time the card simply doesn't work at all .. no signal is the usual result, your card is being recognized by the motherboard and is trying to display an image so its a good bet the BIOS is just fine, rather than flashing it I would put it into a chip reader and dump the bios from it, this way you still have the original vBios and you can compare it to a similar one to see if there is any corruption.

By doing this first you still have the card as is, trying to flash the Bios with a non original bios may just result in a further bricked card.

Just something else to consider to avoid destructive fixes.

Trashbytes wrote on 2023-06-28, 05:39:

Ive seen cards with a dead Vbios before and most of the time the card simply doesn't work at all .. no signal is the usual result, your card is being recognized by the motherboard and is trying to display an image so its a good bet the BIOS is just fine, rather than flashing it I would put it into a chip reader and dump the bios from it, this way you still have the original vBios and you can compare it to a similar one to see if there is any corruption.

By doing this first you still have the card as is, trying to flash the Bios with a non original bios may just result in a further bricked card.

Just something else to consider to avoid destructive fixes.

I fully agree with you. I now have such a "brain dead" card (S3 Virge/GX PCI with SGRAM made by Jaton) with obviously degraded BIOS (system boots without any error beeps, but no video signal at all) from same "store" cardbox. Unfortunately there is no S3 tools to reflash its BIOS and the chip is soldered to the PCB with not so common for this era videocards formfactor (28-SOIC??). I'll keep it just as rare card from not very well known (here) manifacturer. Maybe one day I'll find adapter from 32/28-DIP to 28-SOIC/SOP and old S3 will be saved.

My plan for this defected TNT2 is exactly the same as your's suggestion. First I'll dump its current BIOS and flash/refresh with it (in BuB's last video he was able to resurrect his dead videocard this way). If there's no change, again flash, but with another good BIOS (from given link above - there's Inno3D dump file in zip archive). And if still there's no positive result - solder gun, hot air gun... trash bin 😀

I'm not worried that I can kill the card with simple BIOS flash (except if I kill the old EEPROM chip with write cycles). The EEPROM chip on this TNT2 is socketed, and I have working LAN card suitable for BIOS flash (up to 125kb BIOS images) for the same form factor old (parallel programmable) EEPROM chip - its very usable in combination with Falshrom, already used it to save a PI mobo with scr*wed up BIOS recently.

Right now I'm looking for the right software tool for BIOS reflash directly on TNT2 PCB (still I do not want to use LAN card with Flashrom as programmer - I'm rationally lazy person 😁 ).

Once again, thanks for your useful comments. I'll report after the BIOS (re)flash tryouts...

This is a handy resource for flashing early Nvidia cards, including the TNT2. It lists BIOSes and has links to the relevant tools - I've tried some of the TNT2 BIOSes and they worked fine:

http://www.x86-secret.com/articles/nvbios.htm#TnT2

Thank's for the link, badmojo! I think the older version of nvflash will be of use and I may try these BIOSes too.

Yesterday I was not able to try (re)flashing. Right now I have no working PCI videocard (my only PCI videocard - an old S3 Virge is surely with BIOS problem: boots normally, but shows only blank screen) and on my only mobo with integrated video and AGP slot, that I have in working condition, I can not make it boot to DOS with TNT2 inserted into its AGP slot and working onboard video (dedicated AGP vidocard deactivates integrated video and it has no BIOS settings to select only integrated video as active). So I have to work blindly with the flashing of the TNT2 card.

Today I'll make some autoexec.bat files with nvfalsh for "blind" autobackup and autoflash procedures.

Unfortunately a simple BIOS (re)flashing did not help to mine TNT2 Pro card. Now I'm left only with the option of physical intervention by soldering iron and hot air gun.

But at least in the process of trying to revive the TNT2, I've found that my only PCI videocard (Jaton S3 ViRGE/GX 4MB SGRAM) isn't dead because of BIOS degradation and also that rubbing the contact pads for PCI slot on a card with pencil eraser doesn't always helps, but rubbing with medical spirit (97% alcohol, don't drink it 😁 ) always helps 😀

Check all 8 pin resistors on data lines, sometimes they die due to thermal stress.

DundyTheCroc wrote on 2023-07-01, 09:16:

Check all 8 pin resistors on data lines, sometimes they die due to thermal stress.

DundyTheCroc, thanks for your suggestion! It is a very good direction. I'll check these resistor packages with multimeter for cracks and shorts. Yet I haven't started with the hardware testing of the "patient"...

Following Mr. Croc's advise I checked some SMD elements near the GPU. It is very tricky job using multimeter, magnifying glass and cheapo chinese USB "microscope". Unfortunately I have no schematic diagrams for this videocard.

Checked for shorted caps - none found so far. Also no shorted resistors so far... But I found some suspicious resistors. Here are some pictures with explainations

My investigation continues...

analog_programmer wrote on 2023-07-02, 06:34:

Following Mr. Croc's advise I checked some SMD elements near the GPU. It is very tricky job using multimeter, magnifying glass a […]
Show full quote

Following Mr. Croc's advise I checked some SMD elements near the GPU. It is very tricky job using multimeter, magnifying glass and cheapo chinese USB "microscope". Unfortunately I have no schematic diagrams for this videocard.

1.jpg
4.jpg
5.jpg

Checked for shorted caps - none found so far. Also no shorted resistors so far... But I found some suspicious resistors. Here are some pictures with explainations

2.jpg
3.jpg

My investigation continues...

Is the 150K still 150 when connected to the motherboard?
Only way to know is to desolder both. I guess that the 8.8 will read 10K, but the 150 is 150K, reading is correct.
https://kiloohm.info/smd3-resistor/154

Nexxen wrote on 2023-07-02, 10:13:

Is the 150K still 150 when connected to the motherboard?
Only way to know is to desolder both. I guess that the 8.8 will read 10K, but the 150 is 150K, reading is correct.
https://kiloohm.info/smd3-resistor/154

The measured values of 12.2k instead of 150k for R77 and 8.8k instead of 10k for R78 (this one is not so far from the marked value) are from the resistors as they are soldered to the PCB. I have to unsolder them to check their actual values for sure.

There is still one problem I couldn't catch with the multimeter: If the solder joints of the resistors to the PCB are not good. At the end maybe the hot air gun with plenty of flux will "play" over these resistors.