EEVblog Electronics Community Forum
Electronics => Repair => Topic started by: Traceless on June 20, 2022, 07:29:32 pm
-
Hey everyone,
I'm currently trying to revive a dead Zotac Palit GeForce GTX1060 with a short on the PCIe 12V lines, (the short does not show up on the PCIe PSU connector). The card uses SIC632 integrated power stages (https://www.vishay.com/docs/62992/sic632a.pdf) and an NCP81174 Buck controller (https://www.onsemi.com/pdf/datasheet/ncp81174-d.pdf). During diagnosis I found that the short showed on one of the SIC632 PWM inputs and pins CS4 and G4 of the PWM controller (NCP81174). Also all of the SIC632 outputs (V_SWH) are shorted to 12V. I removed all SIC632s and the NCP81174 but still see the short. I'm pretty confident meanwhile that the problem is somehow related to memory I suspect either a dead core/memory controller and/or one or more defective memory modules. What worries me in particular is that I see ~21.5 Ohms between all four test points on the GPU (marked with green dots on the picture below) and PCIe 12V. However since there is absolutely no visible damage/discoloration/bubbling on the card or the GPU, I'd like to determine for sure if the core, one or more of the memory modules or both are at fault. Of course I could take off the memory chips one after another till the short is gone or no memory is left which hint at the GPU being the culprit. Unfortunately reballing and resoldering the memory is a major hassle. I thought about trying voltage injection but I'm worried about damaging the core (if it is not dead already). Does anyone have ideas/suggestions on how to proceed here safely?
-
Today I found some time to look at the graphics card again. I used my bench PSU to inject 0.7V into the 12V rail hoping to identify a faulty component. However the current drawn peaks at ~40 mA. There are three components on the board that still receive power. One is an RT8128C (https://pdf1.alldatasheet.com/datasheet-pdf/view/956368/RICHTEK/RT8128C.html), it receives the full 700 mV on VCC I also see ~270mV on BOOT. The other other two components I couldn't identify yet. They are marked "MA C" "and 08=3RM" (see pictures below).
Does anyone know what the two mystery components are and/or has some tips on how to test the components (preferably in circuit)? Obviously EN on the RT8128C is low (~60mV) and I can't supply higher voltage, let alone the full 12V without risking damage.
-
may struggle to have schematic put here being discussed nicely together?
-
may struggle to have schematic put here being discussed nicely together?
Hi abdulbadii, I agree schematics would make things a lot easier. Unfortunately I don't have a schematic for that card (it is a Zotac Palit Geforce GTX 1060 - P/N: NE51060015J9-1061D)
-
I finally found the culprit: SC532 has a small fissure which caused it to go short-circuit. That mean thing broke on the underside (the bottom right corner on the picture below)- the top still looks pristine. Now I need to find a proper value replacement for SC532 and also SC87 which I accidentally broke during voltage injection.
P.S.: SU5, which is connected to SC532 seems to be a Richtek RT9403EGV8 (https://www.richtek.com/assets/product_file/RT9403D=RT9403E=RT9403F=RT9403G/DS9403DEFG-00.pdf?spm=a2g0s.imconversation.0.0.65783e5fni3M0T&file=DS9403DEFG-00.pdf)
-
http://www.smdmark.com/en-US/search/code?id=08%3D (http://www.smdmark.com/en-US/search/code?id=08%3D)
RT9068, Ricktek, 60V, 30uA IQ, Low Dropout Voltage Linear Regulator, marking 08=:
https://www.richtek.com/assets/product_file/RT9068/DS9068-07.pdf (https://www.richtek.com/assets/product_file/RT9068/DS9068-07.pdf)
http://www.smdmark.com/en-US/search/code?id=ma (http://www.smdmark.com/en-US/search/code?id=ma)
MBT3904DW1, dual 2N3904, SOT363, marking MA:
https://www.onsemi.com/pdf/datasheet/mbt3904dw1t1-d.pdf (https://www.onsemi.com/pdf/datasheet/mbt3904dw1t1-d.pdf)
-
RT9068, Ricktek, 60V, 30uA IQ, Low Dropout Voltage Linear Regulator, marking 08=:
https://www.richtek.com/assets/product_file/RT9068/DS9068-07.pdf (https://www.richtek.com/assets/product_file/RT9068/DS9068-07.pdf)
Thanks fzabkar, I hadn't been able to identify the chip marked "MA". Interesting is also that you found another Richtek chip with topmark 08=, the RT9403 (https://www.richtek.com/assets/product_file/RT9403D=RT9403E=RT9403F=RT9403G/DS9403DEFG-00.pdf?spm=a2g0s.imconversation.0.0.65783e5fni3M0T&file=DS9403DEFG-00.pdf) also has mark 08= so I guess SU5 could be either an RT9068 or an RT9403 then. Since the board is a graphics card the RT9403 would be sensible from an application perspective - to be sure I'll check continuity between pins 6 and 7 if there is continuity SU5 could be an RT9068, otherwise it is probably a RT9403. I'll also double-check the package format SU5 looks like SOT-23-8 which points towards RT9403 as well, RT9068 is available as SOP-8 or MSOP-8 according to the datasheet and thus should be less rectangular.
-
@fzabkar The resistance between Pin 6 and 7 on SU5 is ~ 120 kOhm based on that I assume it is an RT9403, also for now I assume it is not faulty because the shorts around it and on PCIe 12V disappeared after I removed SC532.
I'm now trying to reassemble the graphics card, resoldered SU1 (NCP81174 Buck Controller (https://www.onsemi.com/pdf/datasheet/ncp81174-d.pdf)) and now see another problem namely a short (.18 Ohm) between pin 23 (CS4N) and pin 24 (CS4) and a short between GS4 and ground (.18 Ohm). Between the other CS[1,2,3]N/CS[1,2,3] pairs is see resistances of around 4.7k Ohm, GS1-3 read all MOhm resistances to ground. So something seems to be wrong with phase 4.
Additionally the resistance between Pin 9 (VSP) and pin 10 (VSN) is pretty low (~7.3 Ohm). Resistance of VSP to ground is ~3.8 Ohm, resistance of VSN to ground is ~Ohm 3.6. I'm not sure what the actual resistance between VSP and VSN should be but the schematic in the datasheet (page 4) (https://www.onsemi.com/pdf/datasheet/ncp81174-d.pdf) shows a couple of resistors and an OpAmp between those pins so 7.3 Ohm seems dodgy unless VSN and VSP are connected to the GPU core.
For now I'm trying to find the problem in phase 4. Is there any chance to determine in circuit if NCP81174 is faulty internally or if the problem is external to the buck controller? Note: I did not yet resolder the SIC632 power stages and the mosfets (4503NH, SM4377) that supply the GPU and presumably the memory - so those can not be at fault.
-
There is a capacitor connected between CS4 and CS4N. Perhaps it is shorted?
-
I'm not sure if there is a cap directly connected to CS4/CS4N. I tried to highlight where the traces seem to run on the picture below. There are also three vias (which I marked with dots and a V). There are two columns of three caps next to the NCP81174. It seems that CS1-3 and CSN1-3 are connected to the right (green) column and then run off towards the GPU. CS4/CS4N however run past the two unpopulated capacitor pads, also towards the GPU.
I wonder: On the graphics card only the power stages U10, U12 and U14 were populated with SIC632s. U11 and U13 were unpopulated (you can see this in the pictures attached to my original post, next to the big inductors). Is it possible that phases 1-3 correspond to U10, 12, 14 and Phase 4 is grounded because it is not in use? Even if that was the case there is still something dodgy. Since you mentioned the caps I also measured the first (red) row of caps and they all show up short (~1 Ohm) between pads. So either they are connected to the GPU core (hence the low resistance) or something is wrong.