Overclock.net banner
1 - 20 of 54 Posts

·
Registered
Joined
·
72 Posts
Discussion Starter · #1 ·
Hi there,

I'm down to my forth dead 980 TI that I'm starting to diagnose...and getting burned out. Anyway, I have a 980 TI founders edition that has all the correct voltages and ohms (errr...maybe/mayb not). I was able to compare it to a working EVGA ACX2.0 that uses the same exacty FE layout. Everything lined up except for difference in ohms on the 5V rail. For the life of me, I can't figure out why there is a difference. The card has all the correct voltages and the ohms for VRAM and the GPU Core are spot on. The fan on the GPU spins and it doesn't have any of the usual shorts you'd expect. In fact, this card doesn't even have heat marks. It seems barely used.

Thoughts...
  • Dead VBIOS? I used my Multimeter referencing the spec sheet for the VBIOs and it seems ok. Voltages are what they should be and where they should be. Maybe the memory on the bios is toast, which isn't impossible. I'm going to try to flash it with a USB BIOS flasher and see if that brings it back to life. If it flashes without issues, then it's fine. If it doesn't flash, I'll replace it
  • Dead resistor or capacitor somewhere in the 5V rail. Where are those missing ohms?
  • Maybe missing Powergood (PGOOD). On the 980 TI PGOOD is on the 1.05V controller, which is a Ritek RT5771C, Synchronous Step-Down Converter. I've got a steady1.05V and not sure where the PGOOD signal goes from there. On Pascal PGOOD is important and used for initializing many things. Not sure what it's used for on the 980 TI.
I can't for the life of me figure this out
 

·
Registered
Joined
·
15 Posts
I'm interested to see where this goes.. Apologies, I replied with a similar issue I'm having with a Strix 980 it, then realised I hate it when someone adds questions about unrelated repairs to a thread! So I'm just here hoping you get some help with this that might give me some more info too ^^,
 

·
Registered
Joined
·
821 Posts
First, dunno if you know TiN or Kingping, but THIS is the resource for us modders for Nvidia gpu.

Log story short, how IC tells the rest of the circuit that they are is OK:
-EN pin: Enable pin is able to power up or shut down the IC, if there is a fault, the EN pin will be pulled up/down, disabling all the IC in the circuit.
-PGOOD pin: it is what the IC sends back to report it is running OK, if the IC pull up/down the PGOOD signal, it means that the IC is not in a good state, after being enabled by the EN pin.

I would check if the cards is not a dual bios one, and start by flashing a fresh bios or checking that the bios switching circuit is working correctly.
You can do the circuit resistance comparison troubleshooting, ONLY if you are sure that both cards use exactly the same circuit implementation.

The RT5771C is not a PGOOD controller, it is a standard buck converter.
Usually the gpu, the EC, or a simple transistor based circuit, take charge to pull all the PGOOD/EN together and sent this signal to the gpu core.
The gpu core will boot up, when all the VRM circuit pull up/down that PGOOD circuit, saying everything is ok, you can boot the bios.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #4 ·
The transis
I'm interested to see where this goes.. Apologies, I replied with a similar issue I'm having with a Strix 980 it, then realised I hate it when someone adds questions about unrelated repairs to a thread! So I'm just here hoping you get some help with this that might give me some more info too ^^,
Please tag along as we ought to share how to fix these things.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #5 ·
First, dunno if you know TiN or Kingping, but THIS is the resource for us modders for Nvidia gpu.

Log story short, how IC tells the rest of the circuit that they are is OK:
-EN pin: Enable pin is able to power up or shut down the IC, if there is a fault, the EN pin will be pulled up/down, disabling all the IC in the circuit.
-PGOOD pin: it is what the IC sends back to report it is running OK, if the IC pull up/down the PGOOD signal, it means that the IC is not in a good state, after being enabled by the EN pin.

I would check if the cards is not a dual bios one, and start by flashing a fresh bios or checking that the bios switching circuit is working correctly.
You can do the circuit resistance comparison troubleshooting, ONLY if you are sure that both cards use exactly the same circuit implementation.

The RT5771C is not a PGOOD controller, it is a standard buck converter.
Usually the gpu, the EC, or a simple transistor based circuit, take charge to pull all the PGOOD/EN together and sent this signal to the gpu core.
The gpu core will boot up, when all the VRM circuit pull up/down that PGOOD circuit, saying everything is ok, you can boot the bios.
Thank you so much for the helping out. There seems to be little information the 900 series as far as function, which makes this so difficult. My gut is that the PGOOD/EN are not being pulled together, likely due to some failed component. The previous owner said this card worked perfectly one day and then no display at bootup the next day. No artifacts, no lag, everything normal - then nothing. It sounds like, to me, that something failed that controls how the card initializes, maybe a failed cap or switch. Or maybe he bricked the BIOS trying to turn it into a mining card and didn't want the stigma of "broken mining card". Who knows but I'll keep working on it.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #6 ·
For grins and giggles I decided to install the gpu with my working 980TI in my test rig. The fact that it wasn't recognized as a 980TI but was recognized as device may actually indicate it's bricked bios. I've had some very broken GPUs and they still showed up in device manager as a GPU at least. I'll try to update the bios with a bios programmer as that's probably the only way to get it done. I ordered one from Amazon and it seems like nearly all of these bios programmers are clones of the real thing and also need to be modded for 3.3V.
 

Attachments

·
Registered
Joined
·
821 Posts
For grins and giggles I decided to install the gpu with my working 980TI in my test rig. The fact that it wasn't recognized as a 980TI but was recognized as device may actually indicate it's bricked bios. I've had some very broken GPUs and they still showed up in device manager as a GPU at least. I'll try to update the bios with a bios programmer as that's probably the only way to get it done. I ordered one from Amazon and it seems like nearly all of these bios programmers are clones of the real thing and also need to be modded for 3.3V.
Yep, could be a bios issue, usually you would get generic graphic adapter loaded with the Micro$haft basic vga drivers.

If you want a cheap programmer, i would warmly advise the cheap CH341A package, that include the 1.8v adapter to flash even the newer 32Meg 1.8v WSON chips.
You can use NeoProgrammer software, to read and flash almost all the common chip, the 32Meg 1.8v Wson chip included, instead of the usual Chinese software it ship with.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #8 ·
Fantastic. That's what I bought. Is it safe for 3.3V Bios? Many reviewers say it's risky without modification. I think that a couple of the pins still have 5V on them even when moving the jumper for 3.3V. I may do the mod just to be save. I'm still hoping that it's simply bricked bios and not some greater issue.
 

·
Registered
Joined
·
821 Posts
I have it with the jumper set to 3.3v and never killed a bios with it, until now.
Maybe the old version had this issue, since i also noticed people modding it, to get 3.3v.

Two weeks ago, i was playing with some WSON 32meg 1.8v chips, using the 1.8v adapter, checking which software would be able to read them.
I picked a random spare laptop motherboard, just to check if the programmer was working, removed the 1.8v adapter and probed a 3.3v chips.
The chip came up right away and i was able to read it without issues.

You would check the voltage on a sacrificial chip, the first time you use it, just to check that the jumper is working correctly and the voltage built-in regulator switch from 5v to 3.3v.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #10 ·
When you say "until now" did you mean you did kill a bios chip? I have a couple of motherboards that I used for parts that have working BIOS on them. I can test a read/write and see if I smoke it. Or do the mod if needed. I'm now wondering if my mobo bios are the the same p/n or slose enough to the gpu vbios? I'm going to take a look.

Also, I can't find anything online that talks about a bricked vbios and error code 28. Most bios bricks are a corruption issue. I think this one is fully dead, not shorted, but zero read/write. I haven't seen this on a GPU yet but on other devices.
 

·
Registered
Joined
·
821 Posts
When you say "until now" did you mean you did kill a bios chip? I have a couple of motherboards that I used for parts that have working BIOS on them. I can test a read/write and see if I smoke it. Or do the mod if needed. I'm now wondering if my mobo bios are the the same p/n or slose enough to the gpu vbios? I'm going to take a look.

Also, I can't find anything online that talks about a bricked vbios and error code 28. Most bios bricks are a corruption issue. I think this one is fully dead, not shorted, but zero read/write. I haven't seen this on a GPU yet but on other devices.
I meant that i had no issues with it until now, so i guess it is working as it should.
Still, i verified if i could read the chip on a dead spare motherboard and record the VCC pin.
Computer Personal computer Output device Space bar Peripheral

As you can see, NeoProgrammer detected the chip, i recorded and put on hold the value of the VCC pin8, on the multimeter.
It reads as it should with the jumper set to 3.3v.

About the error 28, it simply means that the OS didn't find a driver for this hardware.
If the card give an error 28, most likely the gpu booted up, the PCIe PHY is working and speaking with the motherboard.
So you were maybe spot on with the mining bios clue, which need modified mining drivers, that have nothing to do with the OG Nvidia ones.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #12 ·
Wow, amazing info as always. I'll try to pay it forward. I got the programmer today from Amazon...a credit to Prime and next day delivery. More good news is the programmer works and it's very easy to program an IC, even when on the pcb. Getting the jaws to latch is the hardest part until you learn how to do it reliably. So good news was the bios was not cooked. Bad news...after I flashed it with the correct BIOS...just to be sure there was no mining mods...no dice. In fact, same exact behavior as before. So it seems those missing ohms on the 5V rail are looking more like there is an electrical fault somewhere. Somewhere there is a bad resistor, transistor/switch, or capacitor. I think the missing ohms is a sign that something failed open and now an entire control phase is missing. No idea where to look...so I'm parking this one for now. I have other cards to fix that are probably easier to resolve.
 

·
Registered
Joined
·
821 Posts
Wow, amazing info as always. I'll try to pay it forward. I got the programmer today from Amazon...a credit to Prime and next day delivery. More good news is the programmer works and it's very easy to program an IC, even when on the pcb. Getting the jaws to latch is the hardest part until you learn how to do it reliably. So good news was the bios was not cooked. Bad news...after I flashed it with the correct BIOS...just to be sure there was no mining mods...no dice. In fact, same exact behavior as before. So it seems those missing ohms on the 5V rail are looking more like there is an electrical fault somewhere. Somewhere there is a bad resistor, transistor/switch, or capacitor. I think the missing ohms is a sign that something failed open and now an entire control phase is missing. No idea where to look...so I'm parking this one for now. I have other cards to fix that are probably easier to resolve.
Well, at this point, just grab another working 980Ti founder edition, and begin to probe the resistance on various components, on the 5V rail and on the card.
As i said, the best way to troubleshoot electronics, is by having a working piece of hardware to compare with.
In this way, you can compare the resistance and the voltages at a desiderated point of the circuit.

You would do the same with the power consumption monitoring, if you know how much power consume a device at boot.
You can build up a datasheet with the boot power consumption figures, at each stages of the booting process, until reaching the normal operation.
Deviation from these values are likely due to an issue, that you can pinpoint easily, when the boot procedure stop, at a determined power consumption value.

Did the bios get powered when you turn on the gpu?
 

·
Registered
Joined
·
15 Posts
Do you know the power up sequence for 980's, or does it vary from card to card? I've seen a few power up sequence charts for 10 series cards and some AMD cards, but nothing for 980's.. The Strix 980 ti OC I'm working on has gone from being shown in windows, to not being not detected at all now so I'm going backwards with the repair at the moment :cautious: The bios seems to show 3.3v on pin 1 for a second at power up, but then drops to 0v and stays there, previously it seemed to be read as per videos/information I've seen (rises to 3.3v drops for a short time, then returning to 3.3v).. I might switch out the bios for another one.. Knowing the resistor values around the bios would be really useful too!
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #15 ·
Well, at this point, just grab another working 980Ti founder edition, and begin to probe the resistance on various components, on the 5V rail and on the card.
As i said, the best way to troubleshoot electronics, is by having a working piece of hardware to compare with.
In this way, you can compare the resistance and the voltages at a desiderated point of the circuit.

You would do the same with the power consumption monitoring, if you know how much power consume a device at boot.
You can build up a datasheet with the boot power consumption figures, at each stages of the booting process, until reaching the normal operation.
Deviation from these values are likely due to an issue, that you can pinpoint easily, when the boot procedure stop, at a determined power consumption value.

Did the bios get powered when you turn on the gpu?
I get 3.3V to the BIOS when I power on the GPU but not sure if it's actually talking. I think there must be some communication that is not happening between the BIOS and VBIOS because I don't get any output from the display ports. I have an almost identical working one that I will continue probing and recording values. The only thing that concerns me is it uses Samsung GDDR5, which is <20 ohms, probed from memory phase to ground. The one that is not working uses SK Hynix GDDR5, which probes around 84 ohms.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #16 ·
Do you know the power up sequence for 980's, or does it vary from card to card? I've seen a few power up sequence charts for 10 series cards and some AMD cards, but nothing for 980's.. The Strix 980 ti OC I'm working on has gone from being shown in windows, to not being not detected at all now so I'm going backwards with the repair at the moment :cautious: The bios seems to show 3.3v on pin 1 for a second at power up, but then drops to 0v and stays there, previously it seemed to be read as per videos/information I've seen (rises to 3.3v drops for a short time, then returning to 3.3v).. I might switch out the bios for another one.. Knowing the resistor values around the bios would be really useful too!
Have you tried to see if you have 3.3V on the display ICs next to the BIOS? There are two of them and they also use 3.3V. If they're not getting 3.3V either, you probably have an electrical fault somewhere. The 3.3V comes from the MOBO as far as I'm aware so not as simple as replacing a failing buck converter.
 

·
Registered
Joined
·
72 Posts
Discussion Starter · #17 ·
Have you tried to see if you have 3.3V on the display ICs next to the BIOS? There are two of them and they also use 3.3V. If they're not getting 3.3V either, you probably have an electrical fault somewhere. The 3.3V comes from the MOBO as far as I'm aware so not as simple as replacing a failing buck converter.
Also, do you have Samsung or SK Hynix memory? If the latter, would you mind probing it from memory power phase to ground? I get 84 ohms on mine which actually may be high. I'd try to run MATS if I could get the GPU to boot to windows at least.
 

·
Registered
Joined
·
821 Posts
Do you know the power up sequence for 980's, or does it vary from card to card? I've seen a few power up sequence charts for 10 series cards and some AMD cards, but nothing for 980's.. The Strix 980 ti OC I'm working on has gone from being shown in windows, to not being not detected at all now so I'm going backwards with the repair at the moment :cautious: The bios seems to show 3.3v on pin 1 for a second at power up, but then drops to 0v and stays there, previously it seemed to be read as per videos/information I've seen (rises to 3.3v drops for a short time, then returning to 3.3v).. I might switch out the bios for another one.. Knowing the resistor values around the bios would be really useful too!
What did you do on the card, between when it was showing in windows and when it doesn't?
Nope unfortunately, i don't have the power up sequence diagram, but i suppose that checking if all the power rail provide a stable voltage, is the 1st thing to do.


I get 3.3V to the BIOS when I power on the GPU but not sure if it's actually talking. I think there must be some communication that is not happening between the BIOS and VBIOS because I don't get any output from the display ports. I have an almost identical working one that I will continue probing and recording values. The only thing that concerns me is it uses Samsung GDDR5, which is <20 ohms, probed from memory phase to ground. The one that is not working uses SK Hynix GDDR5, which probes around 84 ohms.
84 Ohms for the memory and mem controller don't shock me.
I checked the XDevs page for the 980Ti KP and no resistance charts were reported, unfortunately.
 

·
Registered
Joined
·
15 Posts
Have you tried to see if you have 3.3V on the display ICs next to the BIOS? There are two of them and they also use 3.3V. If they're not getting 3.3V either, you probably have an electrical fault somewhere. The 3.3V comes from the MOBO as far as I'm aware so not as simple as replacing a failing buck converter.
Yeah I expected to find something like that, but there isn't anything nearby that I would presume to be the display IC's? On this card there are two unpopulated (from factory) 8 pin packages the same size as the bios near the outputs & one up1641p that should be the memory controller.. I'll have another look at the weekend when I get a second.. If you can find them and point them out on the techpoweup pics I'd really appreciate it! ASUS GTX 980 Ti STRIX Gaming 6 GB Review This is the non-OC version, but it's basically the same as the OC I have component-wise..
 

·
Registered
Joined
·
15 Posts
What did you do on the card, between when it was showing in windows and when it doesn't?
Nope unfortunately, i don't have the power up sequence diagram, but i suppose that checking if all the power rail provide a stable voltage, is the 1st thing to do.
Good question.. I was trying to trace the platform reset pin, if that's what it's called (first pin left of the notch on the back of the card), but other than it going through a 0ohm resistor, I didn't find where it went, so guessed it must go straight to the GPU.. I've been trying a few different bios flashes (original ones from techpowerup) and I had just programmed one from a thread for 980 users that want custom bios, it was an original bios from the same 980 ti OC that someone uploaded looking for it to be updated to a custom bios.. I have modded my programmer for 3.3v btw.. That's why I was going back to the power up sequence, which to be honest I've never found a 1.8v? I assumed this card doesn't have this line, which is why I was now looking for a confirmed power sequence..
 
1 - 20 of 54 Posts
Top