Mobo hangs on GPU check

[laserman367]

 I recently ordered a new motherboard (AB350 Gaming 3) with a new CPU (Ryzen 5 1600) and RAM (LPX Vengeance). The motherboard has a little disco before boot where it tests the CPU, RAM, GPU (labelled VGA) and then Boot. For each test, a separate light goes on and if there's no problem the light goes out again. However, with my current GPU (ASUS R9 270x DirectCU II ) the light doesn't go out. With an older GPU (Zotac GTX470) the system boots fine and everything functions. Seems like a clear cut case of a dead GPU, however my R9 270x still works just fine on my older motherboard+CPU+ram combo (P7P55-M with i5 760). When the PSU gets power, the R9 270x shows green lights above where the power is connected and when I try to boot, the fans spin. Not sure if that helps, but I thought I might as well mention it

 

As for things I've tried,

I tried plugging it into a different PCI-E slot on the motherboard, however a friend said that was a bad idea since only the top one was PCI-E 3.0.

I tried using a brand new PSU (so I've tested it with two, both of which should be plenty to power the new system at 550W since pcpartpicker only rates it at about 350)

I updated my new motherboard's BIOS to the latest version.

I tried changing the PCI-E settings from Auto to Gen3, in case it didn't recognize the R9 270x as gen 3.

At friends recommendation, I tried turning off XMP but it was already disabled by default.

[laserman367]

I think this video from JayzTwoCents shows the exact problem I'm having


Sadly he seemed to have no idea of what caused the issue and I can't find an update about it. For him it's probably not such a big issue since he can just run another card on that system, but I can't magically switch my GPU for an identical one

[DaveLembke]

Between a known good video card and a known ( run of tolerance ) video card that works fine in one system vs another, both cards same make/model & Bios Version. I'd be looking at them side by side and comparing the rest of the chips on the video card that are standard components. Its quite possible that the one video card has 2 different standard non-custom universal chips used as for example a common Quad Comparator in which one card is using a Texas Instruments chip and another is using STMicroelectronics. These chips are manufactured to stay within a specific tolerance but by which you can drive these chips fast and one chip may handle it way better than another brand due to the slightest difference in design. Its quite possible that his situation is a ( run of tolerance ). A run of tolerance is when 2 parts combined place the equation of their successful functionality outside of permitted functionality. In this case its likely a timing run of tolerance. the PCI Express BUS speed is not a constant between those 2 motherboards the AMD and the Intel. So the slightest tolerance offset which is still within acceptable timing by nVidia and the AMD motherboard also offset in the opposite direction for example would create a timing synchronization issue. Computer internals have to work very fast but also be timed very precisely. There are virtual traffic lights that keep from collisions from happening, but if the light turns green and the bit isnt able to pull away from the light in time then you have problems, and thats where this guys issue is.

If you tested your system with a different video card and its fine, and tested the video card that has issues in another system and its fine there, and the power supply isnt the problem of a combined power draw that is too much, then its very possible that its a run of tolerance issue. Unfortunately at the consumer level a run of tolerance means that you have to isolate it down to the known culprit as to what added to the build makes it no longer work. You have it narrowed down to video card from the sounds of it.

As the guy in the video stated you send your video card back RMA under warranty and you could get the same card back. They test it on an ATE and it passes and its shipped back. If not wanting to try to find a different video card, then the motherboard itself is the other part of the equation however the CPU itself also controls the PCI Express BUS.

Something like this to successfully test what exactly the problem is would require a test bench with a PCI Express riser with test points to take signal readings to look between a known good and the troubled system and look to see where the timing issue is at. I believe its a timing issue that is choking up this guys test build.