Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

[Joe24]

Moving this to it's own thread from the Schematics thread.

I've got an RTX 2070 here that doesn't artifact, no memory errors in MATS, boots fine in Windows Safe Mode, but freezes the instant you try to load a driver. BIOS manually reflashed, no change. Voltages are all fine, PCIe lines all look good. Suspect Vcore power supply issues, or I guess could be a bad GPU. My first 2070, so there may be potential causes that I'm not aware of. Not super familiar with the circuit.

If the vcore on iductors is 0.9 Volts , then mightbe a bad gpu ,or some bad mosfets of the core if you have oscilloscope try to check if the pmws are all simillar and equal going to gpu. when loading drivers the gpu boost and bad pmws signals can cause it crashing, also what gpuz is showing without drivers ?and does it crash also as secondary card using another grapgic as primary or builtin gpu if you got one ., some times mat can give you no errors because is running on 300 mhz memory speed , try increase that (which i donot know if is possible) or increase the tested size more than 20 megabyte.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Okay, so finally got some time to get back to this card (2 funerals, ugh).

GPUZ shows the card and some of the basic sensor inputs (temperature for sure, I forget what else). Doesn't show everything (fan speed) that would be there after driver installation.

Device Manager shows the card with error 43, which is pretty uninformative, just means there's some unspecified problem with the card (who knew?).

Back to the hardware. Measured the PWM input resistance of the 8x TI CSD95480RWJ DrMOSes, and the first 6 measure between 15.1k and 21.7k . . . but the last 2 are 172k and 223k, respectively. Verified that there are no other components on the PWM circuits between the uP9512 and the DrMOSes, so there's something screwy here with those last 2 phases.

No visible damage. Not sure whether the controller or the MOSFETS are faulty. Seems farfetched to have 2 chips fail at the same time, so I'm going to pop the uP9512 controller off the board (that's the simplest method; isolates all 3 components) and keep testing. I'm leaning towards a bad uP9512, but we'll see . . .

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Okay, so I pulled the Vcore uP9512P off the board (there's actually a pair of uP9512P; the one on the back of the board runs Vcore, and the one on the front of the board runs Vmem). Then I tested the PWM input resistance of the Vcore DrMOS chips.

PWM1 - 15.1 k
PWM2 - 92.3 k (*)
PWM3 - 21.3 k
PWM4 - 21.4 k
PWM5 - 15.7 k
PWM6 - 22.0 k
PWM7 - 247 k (*)
PWM8 - 235 k (*)

The 2 Vmem phases are also the same model of DrMOS, so I tested them as well:

memPWM1 - 15.0 k
memPWM2 - 91.7 k (*)

Who to believe? Normally I'd say that the outliers were defective, but here are 4 out of 10 chips with super high resistance?

[dl123]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I have the same issue on my zotac amp 2080 super. I can boot into windows with code 43, mats run just fine. I was able to boot it once and it was working fine until i restarted my pc.

[dl123]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Did you have any progress with the repair of this issue?

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Did you have any progress with the repair of this issue?

Well . . . 2 more funerals later . . .

This story is kinda embarrassing so far . . . as mentioned, I removed the Vcore controller so that I could test the resistance of the DrMOS gates. (NOTE: on these cards, Gigabyte used a pair of uP9512P chips for voltage controllers - one on the back of the board for Vcore, and one on the front of the board for Vmem. This differs from the Nvidia schematic.). When I was done testing, I put the Vcore controller back on the board.

And now the card was completely dead, no Vcore at all. Oh poop.

Thought I messed up the reflow, solder bridge or something. So I removed and put back the controller again. Being really careful this time. No joy. Very carefully inspected the surrounding components, and everything looks right. Near as I can tell, the heat from the reflow process killed something.

This could be bad, or it could be good. Either we just created another problem . . . or lucked into finding the original problem. Remember that I suspected a power supply issue. If there was a dodgy component that spazzed out whenever it would get warm, maybe we just pushed it over the edge with the controller reflow.

The other thing we know is that whatever the problem is, it's likely within 1-1.5 inches of the Vcore controller, although we don't know which side of the board. Nothing else got hot.

Keeping in mind that this section of the circuit is completely different than the Nvidia schematic. So I'm flying blind, and not very good at that.

Testing indicated that the Vcore controller is now not getting an Enable signal. Tried to trace the logic back to the various sources, and was unable to. Again, this card is quite different than the Nvidia schematic, and the card does not follow the boolean-and-diode array of the schematic, which takes the dozen-or-so Power Good signals from various monitoring systems and boils them down to a single OK signal which is then provided to the Vcore controller's EN input. And there are also plenty of unmarked chips just to keep life interesting.

So I gave up on trying to piece the logic together, and started working at the problem from the other end. I could see 2 chips that are used for monitoring input currents on various power rails (presumably 12V and 3.3V, I didn't check), both NCP45491. I could see them and what their purpose was, and knew that their OK signals would be required for Vcore to turn on, whatever the logic in-between was, so I decided to start with this known factor. Testing them (U605 and U608) revealed that although U605 has normal levels, U608 does not.

U605:
pin 28 (EN, low=enabled, measured 0.11V good)
pin 30 (BS_OK, high=monitored inputs are good, measured 2.79V good).

U608:
pin 28 (EN, measured 0.06V good)
pin 30 (BS_OK, voltage wanders around between 0.8-1.3V, bad; should be either ~0V low or 2.6+V high)

Both BS_OK pins, according to the schematic, are joined together and pulled high by the same resistor, but each chip has a 0-ohm resistor between its BS_OK and the shared pullup resistor. U605 is fine, so the pullup resistor is functional.

It's looking like I have a bad 0-ohm resistor on the U608 BS_OK line. Which is in keeping with the wandering voltage reading. Broken resistors are like that sometimes.

According to the schematic, R949 would be the culprit. Well . . . there's a resistor labelled R949 on the board, but it's not a 0-ohm resistor, nor does it seem to be connected to U605 or U608 on either end.

So I have to hunt manually and find the real resistor. Did some quick sweeping with a multimeter, but was unable to find it. Ran out of time, haven't got back to it yet. Cliffhanger!

If anybody's got a boardview, or a schematic for this specific card, hit me.

[dl123]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I see you've been doing a lot of hunting to find the problem, hope you find the broken resistor. as for it not working after you remove the chips the heat probably completely killed the chip or resistor that was already dying. I need to look into those areas on my rtx 2080 and see if it has a similar issue.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Soooo . . . short version is, I screwed up soldering, and now that I've fixed the screwup, the thing turns on again.

Now we're back to square one with the original problem where the memory seems fine and it'll run MATS all day, but the card crashes on driver load. This seems to extend to MODS, which runs long enough to generate a log file, but kills the screen and hangs the system.

A log file is attached. Most of it is Greek to me, but I see a lot of references to not being able to read various data from the BIOS. Also an MMU fault. Not sure how much of any of this to take seriously, but if the card doesn't know what voltage to set itself to, that could conceivably be a cause of the hanging issue.

Somebody asked before what the actual voltages are. I'm measuring GPU 0.780 V, and VRAM 1.320 V. And yes, I can see the card in Device Manager as either primary or secondary card. As a primary card, you'll only get to Windows in Safe Mode, so no driver. And as a secondary card, Error 43. And if no driver can load, then I don't have access to any frequency controls, etc. for testing. But this is not a software problem, obviously.

MATS doesn't have frequency control afaik, and MODS won't run.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Oscilloscope graphs of Vcore (NVVDD). Included a known working GTX1080 for comparison. This is at idle, of course, while the card is working fine.

So, changing gears here a little bit . . . assuming it was memory, like everybody insists it is . . . how would you go about figuring out which VRAM module is bad? I cannot generate any memory errors using MATS, can't run MODS, and can't load any drivers to run other tests.

[RJARRRPCGP]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Device Manager shows the card with error 43, which is pretty uninformative, just means there's some unspecified problem with the card (who knew?).

It could very well mean that the card Windows sees, doesn't match the driver .inf file. On some systems, it may mean poor contact with the slot.

Reminds me that I was very lucky for Windows to not have such error code, and the VBIOS flash utility failing, when I had a mea maxima culpa moment back in the winter with my Radeon RX 5600 XT. (I flashed a truly invalid VBIOS by mistake, because stupid me decided to reboot after flashing what looked like it couldn't possibly be more than a stub!) (To people that are new with AMD Radeon BIOS flashing, there's only 1 file and it's 512 KB, IIRC)
(That means if you see "128 KB" or whatever other size, stop right there!) (My GeForce GT 640 saved my bacon there, as if it were a little car with enough juice to jump start mine!)

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

It could very well mean that the card Windows sees, doesn't match the driver .inf file. On some systems, it may mean poor contact with the slot.

Slot is fine, and the problem is not limited to Windows. MODS itself crashes the card.

[taxxin]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Have you checked to see if the bios is corrupted as stated by RJARRRPCGP?

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Have you checked to see if the bios is corrupted as stated by RJARRRPCGP?

From the first line in the first post: "BIOS manually reflashed, no change." Yes, read back and verified.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I ditched the card. Somebody wanted it worse than me, and was willing to pay for it.

[dl123]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I ditched the card. Somebody wanted it worse than me, and was willing to pay for it.

I had a similar issue on my zotac 2080 super extreme amp where I would get code 43 and mats passed without any issues all of the voltages were fine yet i got code 43. Turned out two drmos on one of the powerstages were bad.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I had a similar issue on my zotac 2080 super extreme amp where I would get code 43 and mats passed without any issues all of the voltages were fine yet i got code 43. Turned out two drmos on one of the powerstages were bad.

Bad in what way? How did you determine this?

[Schorsch87]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

Bad in what way? How did you determine this?

You need a scope to check switching frequency and amplitude.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

You need a scope to check switching frequency and amplitude.

Yes.... Now what measurements led you to the conclusion that you had faulty DrMOS chips? In what way were they "bad"?

[Schorsch87]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

You have to compare the measurements between all DrMOS. If one is bad you will see it. You won't get a good looking rectangular voltage curve. Maybe you will have a higher ripple.

[Joe24]

Re: Gigabyte RTX 2070 error 43 but no artifacts, crashes on driver load

I did use an oscilloscope to check the output of each DrMOS (when I still had the card), and all phases looked even. Initially, that was my suspicion too, that I had a dead power phase. But everything seemed normal. I don't know how to explain the difference in gate resistance. All I know is, the outputs seemed correct.