RTX 4090 failure to POST and power limit = 0

[Relatyvity]

Specs first:

cpu: AMD R9 5950x

vga: INNO3D RTX 4090

ram: G.Skill Trident Z (4x8GB)

Old board: Asus b350 TUF Gaming

New board Asus x570-E Gaming

psu: EVGA Supernova 1000GT

 

Day 0

Issue started a week ago. Used the PC as normal, went to sleep, woke up, turned it on, no signal. After installing the beep/code speaker, I was getting 1 long - 3 or 4 short beeps. Opened a thread on Tom's hardware, and with a few (somewhat) helpful people we came to the conclusion, that the PCI-e slot was probably dead. Borrowed a 1050 from a relatives PC, crammed it into the lower slot, posted, booted, worked. I went online to buy a second hand x570 board. Ended up with the one above.

 

Day 4

New board arrives, I put everything together. Fails to post with a d6 error code (I was trying with the 1050 as the 4090 is a pain to move around). Tried in all 3 slots, didn't work. In the top it kept trying to post, in the bottom 2 VGA led stayed on and didn't post. Flashed the newest BIOS from a thumb-drive. Cleared CMOS  a few times and at one point I managed to get it to post and even reach the BIOS. After switching the cables about I even managed to get an image (I think it worked from HDMI but not from DP). Flashed the BIOS again rebooted, no signal again. Gave up went to sleep.

 

Day 5

Next day, after I came home from work started experimenting with a riser cable I just bought. 1050 was unwilling to work, so I started trying with the 4090. In the top slot, same d6 and VGA light on, no post. In the middle one it would give a beep code but does POST and boots (I can hear the windows welcome sound) but no image. Third slot... doesn't fit in there. So I figured I'll try with the riser as well just to be thorough. Lo and behold it worked. In both the bottom and the middle slots, it POSTs successfully, boots up and there's even an image. However. If I fire up afterburner: GPU clock: 855MHz, VRAM clock: 10501MHz. Power limit and temp limit are both at 0, when I try to change them when I hit apply they go back to 0. GPU utilization also reads 0. The cooler is also off, so really glad I invested in a few decent front fans, cause otherwise it would start to bake itself.

 

So what is this?

Hardware failure: could be but having 2 boards with very similar issues feels unlikely. So maybe the GPU then. But the 4090 is only 3 weeks old, but ok, maybe it's a pretty extreme piece of kit in most ways, so maybe. But the 1050? I went and put it back into the relatives system to check, and it booted right up...

Software issue then: After I managed to get into windows, I've reset the system just to be sure. Then reinstalled it, just to be even more sure. I installed all the drivers from ASUS's page, the the Geforce driver. I also tried downgrading the chipset driver, downgrading the GPU driver, downgrading the BIOS, removed the driver with DDU then reinstalled an older one and a combination of these, none of which worked.

 

Additionally: I dug up an old "emergency" GPU, an R7 250x. Guess what. Booted right up no matter the slot. So for now my running theory is that there is some kind of a clash between the board/chipset and the nVidia firmware.

The only thing I can think of and did not try was flashing a new VBIOS but I believe that voids the Warranty which in the case of the card in fact being faulty would be a very severe financial hit for me.

 

Of course I would much appreciate any kind of help but I'd also like to document this, as similar "GPU power-limit stuck at 0 and clock-speeds at minimum" threads have been opened on other forums but had no resolution.

I will go over to another relative's tonight, who has a somewhat similar system to test the 4090 in their system, and hopefully their 3070Ti in mine. I'll update the thread about any findings.

 

[lopj245]

It only voids warranty if its a non-stock bios ( not from the manufacturer ). I've seen this in the past and I'm trying to remember what was causing it. I do write-ups on everything that comes into the shop so I'll dig that out and see what my findings were at the time.

[Relatyvity]

@lopj245 Very kind of you, I would certainly appreciate it

 

So I promised an update, here it is:

 

After a trip to my dad, we checked the 4090. Unfortunately we were only able to "fit" the card in, with the riser cable, and it exhibited the same exact behavior, that you can see on the screenshot above. With his 3060 (which I thought was a 3070ti) the sliders were at the correct position and adjustable.

BUT. When I hooked up the 3060 to my riser cable, it booted (though it took relatively long) then dropped the signal. It kept turning the fans on and off, but that's it. When we tried to reboot, it did the same thing. No signal just periodically spinning fans.

So, what's the conclusion: Even though the riser cable was brand new, it seems that it's faulty. I will try and test (not during the weekend I have an OS test coming up, gotta brush up on my C knowledge) if the riser fails with the R7 250x as well (though that's a very different card so it may or may not fail in the same way). My guess is that the clock-speed issue may also stem from the riser cable.

Now, you might be thinking, great but that doesn't exactly explain why won't the card pass POST when it's plugged into the mobo. Well, remember the D6 code I mentioned? It means: "No Console Output Devices are found" which seems to be GPU or VGA related but not specifically VGA failure (I think that should be B2). But let's just say the board cannot detect a graphics card. It didn't even occur to me, but check out the pictures I've taken. Does it seem seated correctly? I didn't even check as I tried to push it in as much as it was willing to go, but it does seem lopsided. So much so, that towards the rear of the card, you can see the top of the contact pins peeking out slightly. This is btw with the 4090 screwed into the case which is flipped on the side.

 

[Relatyvity]

Update again,

 

The problem is solved.

The solution: I've made the mistake of reading the manual. The cable mod, riser+bracket that I bought did not mention, that the cable they sold with the kit was in fact a Gen3 cable. Asus, does not mention in their mother board manual, that setting the PCI-e behaviour to Auto does not in fact detect what type of device is plugged into it and instead default to Gen4. Meaning the only thing I had to do was go to the BIOS -> select Advanced -> Onboard device configuration (or something like that) -> Find the PCI-e slot I plugged the card in -> set it to Gen3, and voi-fckin-la, it works perfectly.

 

I'm gonna leave this here, hopefully some other helpless soul in a similar situation might stumble on it