Picked up this “defective” ASUS ROG STRIX RTX 2070 SUPER for 47€ a few days ago and finally managed to get it working again.
The original eBay description basically said: “RGB LEDs turn on, but the fans don’t spin and there is no display output. Card already opened before. Sold as defective.”
And that was exactly the condition it arrived in. RGB worked, but absolutely nothing else. No fan spin, no display signal and the motherboard always stopped at the VGA LED.
The first thing I found was a missing DrMOS, so I replaced that and honestly thought the card would boot afterwards. It didn’t. Still completely dead.
At that point I started tracing the startup sequence manually because a lot of rails were missing. I basically started at the BIOS and traced 1V8_AON backwards through the whole enable and monitoring chain trying to figure out why the card never entered the proper startup sequence.
One annoying part was that I couldn’t find a proper boardview for my exact revision. The only one I found was for a different revision of the card, but luckily most sections were still very similar or only slightly changed, so it still helped a lot.
Finding datasheets for some of the chips was also pretty painful. A lot of components only had tiny SMD markings and almost no useful search results. Eventually I managed to identify every chip I needed though, which made understanding the pinout and switching logic way easier.
Before finding the actual fault, I also found another missing component with a ripped pad, which I believe was probably R1196 based on comparing the PCB to the boardview. I carefully scraped a bit of the trace open and got pretty lucky because the spacing between the remaining pad and the exposed trace was almost perfect for directly soldering in another 10k resistor without needing a bodge wire or solder bridge.
While tracing the startup logic I eventually ended up at the 5V enable path. At first I thought PS_5V_EN_Q1 was the problem since it goes through a resistor to GPIO1_GC6_FB_EN, so I followed that whole path for quite a while until I eventually couldn’t get any further there.
After going back to the 5V_EN section and checking the surrounding components again, I noticed that R1222 wasn’t getting 12V even though it was connected to the exact same trace as R462, which did have 12V present. Replaced it with a 10k donor resistor from another board and now the card works perfectly again.
Before finding that though, I injected 1.2V into the 5V_EN line with a lab bench power supply just to see what would happen and the card instantly came to life. Got display output, Linux detected the card correctly, USB-C worked, audio showed up and OpenGL was running fine.
Ran multiple stress tests afterwards and everything seems stable. No PCIe lane errors, no crashes, normal temps and normal power draw. I also tested the card in actual games afterwards and it runs perfectly there as well without artifacts, crashes or instability.
This was actually my first proper board level GPU repair and also my first time using an SMD rework station and infrared preheater, so I’m honestly pretty happy with how it turned out.
Would be interesting to hear what you guys think about the repair and especially the soldering work.