| Electronics > Projects, Designs, and Technical Stuff |
| GTX1080 Memory VRM Going into OVP |
| (1/1) |
| Snoopy04:
Hi, as the Title suggests I've got a Problem with a GTX1080 from Nvidia. I've bought it as defective for a few (relative) Bugs and I'm now trying to repair it. I think the main issue was an Overcurrent sittuation on the PCI-E 12V Rail because the Fuse (10A) was Blown. Surprisingly none of the VRMs, which rely on this Voltage, or the MOSFETS of the Main VRM for the Memory was in Short. I've checked both MOSFETS with a Multimeter in Diode Mode an everything seems to been just fine. After changing out the Fuse of the Rail, the Card seems to work Normal except for the fact that there is no Image. The fans are Controlled and The Card is recogniced by my Computer. (With help of another dedicated GPU). After further Investigation I found that there was no memory Voltage Present. So my first thought was that some of the Control lines for the Controller IC (RT8816 )are not in proper state. Turns out everysingle Line Voltage is where it should bee. So obviously the Controller IC must have died in the Overcurrent. Therefore the Memory is not Powered and I'm not getting a picture. I've searched nearly everywhere but I couldn't find an exact replacement for it. So after further Research I've comealong a NCP81278 which seems to bee a complement of the RT8816. But after trying out and Comparing the two datasheets even more I think now that there are indeed some differences. After Powering the Card with the NCP81278 the Memory VRM tries to Provide Power but shortly is switches off due to Over Voltage on the Output. (High Side FETs are disabled and LOW side Enabled) It's doing some weired stuff. REFIN Voltage is around 1.35V so if I've understand the working principle correctly the Output Voltage of the Memory VRM should be 1.35V. Instead it exceeds 2V before it finally shut's off. My first thought was that there is something wrong with the LOW side of the DUAL MOSFET but after Goofing around with my Oscilloscope I don't think so. See my Attachments. In the Picture I've marked the Gates of the LOW side MOSFET, the HIGH Side MOSFET and the Output Voltage after the Coil. I've moved the Wavefomes a little bit so you can read the Lables. Has anybody an Idea what's going on |O Someone who is more into VRMs or Buck Converters. |
| dmendesf:
It could be a problem in the feedback or compensation network. One idea would be to remove the VRM or cut it's input somehow and power the memory bank from an external power supply to see if this is the only problem remaining. |
| wraper:
--- Quote from: Snoopy04 on December 19, 2018, 11:32:35 pm ---I've searched nearly everywhere but I couldn't find an exact replacement for it. --- End quote --- Aliexpress is full of RT8816. It is available on ebay too. |
| wraper:
There is a difference of pin 9 (TON/FS) operation. Direct replacement as minimum will cause incorrect switching frequency. |
| Navigation |
| Message Index |