Announcement

Collapse
No announcement yet.

Sapphire HD 7870 possible short

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Sapphire HD 7870 possible short

    Hi folks,

    I wanna start off by saying hi to all you guys as this is my first post, this is the only place I come to when in need of advise and knowledgeable answers. Great forum you all have here. I've been repairing my PC hardware for some time now (Minor cap replacement, Dc jacks). But that's kind of the extent of my knowledge so want to improve my skills and learn a bit more so what better way than to jump right it at the deep end and try fix a dead graphics card ay?

    The card in question is a Sapphire 7870 Ghz edition 2GB which I believe sticks to amd's reference pcb design.

    The problem is that the PC refused to boot at all when the card is inserted and pcie Molex connectors connected in the case of this particular card 2x 6pin. The Graphics card fans just 'Twitch'.

    I've ruled out motherboard and psu. A Tyan K8WE with dual opteron 285's & PC Power & Cooling turbo cool 850w (bought pre Ocz takeover, cost a fortune getting it from the states) love that psu , and I'm pretty sure more than enough power required for this card.

    From what I've read around the forums I believe the problem to be a short circuit and the psu's short circuit protection kicks in to prevent further damage.

    So I decided to lift off heat sink and take a look at the naked board expecting to find Burnt components maybe a blown capacitor, so I was surprised to see none of these, the board was in very good condition with no sign of damage at all.

    Due to a possible short I think we can cancel out caps? Correct me if I'm wrong?
    More likely VRM circuitry possible MOSFET at fault, or bad lead free Bga balls? What do you guys think? Thermal paste on the gpu core seemed sufficient and the size of heat sink and copper heat pipes make me think this card didn't overheat but I'm no expert. I've tried enabling chip select on the bios by shorting pin 1-8 I think it was but still nothing, so don't think its a bad flash.

    Now I know sapphire have had trouble with this particular model as a result of bad tantalum capacitors on the reverse of the card. I've checked cards serial number / date codes against known bad date range via sapphire website and I'm safe, not one of the 6% of duds.
    But from what I've read the symptoms for this are a black screen and not boot-up issues so I've ruled this out. Maybe I could RMA but I'm not the original owner and I've already voided warranty by dismantling. Plus where's the fun in that? I'm not out of pocket so it's not gunna hurt to learn and hopefully help others.

    I've tried to sort this myself but am a bit out of my depth and don't know where to start or in which order to test components. I've done the best I can to get a hi res pic of the card and I've sourced data sheets of the mosfets and a few controller ic's.

    I have a mastech 8211d pen style multimeter with continuity test buzzer (50ohms according to manual if this makes a difference) and an atlas esr70+ meter and a load of desoldering gear at my disposal.

    If someone could point me in the right direction or maybe talk me through what I should do next.

    KEY FOR THE IMAGE

    RED - 78M05 Voltage regulator
    BLUE - CHiL Semiconductor CHL8225 Multi-Phase Buck controller
    GREEN - RICHtek 8120 Single-Phase Synchronous Buck PWM Controller
    ORANGE - UNKNOWN CHIP AM113? LINEAR IC AMPLIFIER MAYBE?
    YELLOW - ON Semi 4904NG Single N-Channel Power Mosfet
    LIGHT BLUE - ON Semi 4963NG Single N-Channel Power Mosfet

    Im currently mapping out continuity between molex pins on both the 8 pin molex's, will post a pic shortly
    Attached Files

    #2
    Re: Sapphire HD 7870 possible short

    Yep, this is almost certainly a shorted MOSFET (most likely) or shorted ceramic capacitor (less likely).

    There are probably at least two VRMs/buck regulators on this board. The CHL8225 multi-phase buck controller most likely drives the MOSFETs for the GPU VRM, while the RICHtek 8120 probably drives those 4904NG and 4963NG MOSFETs for the RAM VRM.

    In a VRM/buck controller, you have two sets of MOSFETs - "upper" and "lower". The upper one connects between a PSU power rail (12V, in this case) and the GPU or RAM Vcc rail through a coil. The lower MOSFET connects between Ground and the GPU or RAM Vcc rail (again, though the same coil). The buck controller cycles these MOSFETs ON and OFF in an alternate order - that is, when one is ON, the other is OFF. This allows the coil/inductor to store energy when the upper MOSFET is ON and lower OFF, and then release that stored energy and dump it into the GPU/RAM Vcc when the lower MOSFET is ON and upper OFF. This allows the VRM to control the voltage very very precisely.

    Now, a thing to keep in mind is that the GPU and RAM Vcc rails will in almost all cases appear shorted when you measure their resistance to ground (it depends on the GPU, really - some will measure as low as a few Ohms, while others like a GeForce 7600 GTs I worked on recently, will show around 7 to 8 Ohms on GPU Vcc and 15 to 20 Ohms on RAM Vcc).

    It's wise to measure these two rails, as it could pinpoint where the short circuit is. On your board, GPU Vcc is likely connected to all of those "Magic R19" coils/inductors. So just measure resistance between one of the pins of those inductors and ground.

    For the RAM, the Vcc rail should be connected to those "1R0" inducotrs on the right side of the board where the 4904NG and 4963NG MOSFETs are. So measure resistance between the pin on those inductors and ground.

    Write down both resistances that you get above. Also, when measuring the resistance, make sure you use the LOWEST resistance scale (usually 200 Ohms for manual multimeters).

    After this, measure resistance between the 12V rail on the 6-pin PCI-E connectors and ground (I believe the upper middle pin on that connector is for 12V rail detection, so don't measure the resistance on this pin).
    - If you get a low-Ohm resistance reading very similar to either of the readings you got for GPU Vcc or RAM Vcc, then you likely have a shorted upper MOSFET in one of the VRMs (if the GPU and RAM Vcc rails have distinctly different resistances, then you can use this to know on which VRM the shorted MOSFET is).
    - If you do NOT get a low-ohm resistance between 12V rail on 6-pin PCI-E connector and ground, then you likely have a shorted lower MOSFET.

    In either case, you would start by removing the affected MOSFETs (upper or lower, that is) one by one and test them out of circuit for short between Drain and Source pins.

    Lastly, before I finish this wall of text, here's how you know which MOSFET is upper and which is lower...
    Upper MOSFET: for N-channel, Drain is connected to 12V rail and Source to GPU/RAM Vcc.
    Lower MOSFET: for N-channel, Drain is connected to GPU/RAM Vcc, and Source to ground.

    I hope this is helpful to you and let me know if something doesn't make sense. The process of checking is actually very easy and quick once you get the hang of it.

    Comment


      #3
      Re: Sapphire HD 7870 possible short

      Thank you sir momaka to the rescue. I very much appreciate your help, that's a very in depth reply still trying to wrap my head around it.

      So here's what I get:

      Upper IR0 to ground - OL
      Lower IRO to ground - 128 ohms

      R19's to ground - 6.8 ohms

      12v 6pin molex to ground measures 6.6 - 7 ohms

      So does this mean it's an upper MOSFET at fault?
      It's the whole upper / lower thing that I'm finding had to grasp. How would I find the right pair of mosfets. Sorry if you've already answered that I'll go back and try re read it all. I'm more of a visual man I have to see it to understand it if you get me.

      Many thanks for your help and patience.

      Comment


        #4
        Re: Sapphire HD 7870 possible short

        Originally posted by modman1337 View Post
        So does this mean it's an upper MOSFET at fault?
        Yes, you got it.

        Notice how your R19 resistance measurment is almost exactly the same as the one for the 12V rail on the PCI-E 6-pin connector? That's how we know it's most likely an upper MOSFET .

        So for suspects, you have MOSFETs Q601, Q621, Q641, Q631, and Q611 - those black square chips that look exactly like the CHL8225 buck controller. They are located right above and below those R19 inductors.

        Need to get the part numbers off of them so that we can determine what they are. Might be single or dual N-channel MOSFETs. Now, as far as determining which one is the faulty one exactly - that's not quite easy. So normally the method to determine the faulty one is to remove them one by one until the short circuit/low resistance between 12V and ground is gone.

        Originally posted by modman1337 View Post
        It's the whole upper / lower thing that I'm finding had to grasp. How would I find the right pair of mosfets. Sorry if you've already answered that I'll go back and try re read it all. I'm more of a visual man I have to see it to understand it if you get me.
        It's okay. This circuit isn't really easy to explain without a diagram. I am actually glad to see you understood as much as you did. If you look at the circuit diagram in this article, Q1 is the upper MOSFET (not because it's higher in the diagram but because it is connected to the +terminal of V_in) and Q2 is the lower MOSFET (it's connected to the -terminal of V_in, which is normally ground). V_in in our case is the voltage between the 12V rail and ground. The two wires going to the right are the outputs - the lower one being ground and the upper being the output voltage (this can be the Vcc rail going to the GPU, RAM, etc.).

        The reason there are 5 MOSFETs on your card (and also 5 inductors) is because there are 5 of these buck circuits, all running in parallel to each other, supplying the GPU with the needed power.
        Last edited by momaka; 02-08-2014, 05:38 PM.

        Comment


          #5
          Re: Sapphire HD 7870 possible short

          Nice thanks, I think I understand the circuit better now.

          Right so I got my camera out again to try find out what these mosfets were and who manufactures them.

          Heres what I got:-

          FDC18AB
          FDMF
          6705B

          Which I think translates to a Fairchild FDMF6705B 6x6 PQFN XS DRMOS FAMILY MOSFET, do you agree momaka?
          Attached Files
          Last edited by modman1337; 02-09-2014, 06:58 AM.

          Comment


            #6
            Re: Sapphire HD 7870 possible short

            Originally posted by modman1337 View Post
            Heres what I got:-

            FDC18AB
            FDMF
            6705B

            Which I think translates to a Fairchild FDMF6705B 6x6 PQFN XS DRMOS FAMILY MOSFET, do you agree momaka?
            Yes, that looks right to me.

            Man, these things are complex! - more than just two MOSFETs inside. Anyways, that's not that important. However, the faulty one will need to be replaced with the exact same part (i.e. you can't stick just any good ol' MOSFET in there - something I often like to do ).

            You found the datasheet too, so that means you probably found this page as well:
            http://www.fairchildsemi.com/pf/FD/FDMF6705B.html
            $1.6/piece. A bit pricey but not bad. If you are a student or engineer, however, the good news is you may request some free samples .

            So yeah, not much else to say. I guess it's time to start removing them one by one and see when the short circuit on the 12V disappears. Do you have any BGA/hot air tools available at your disposal, though? Because, if you don't, this may be a little bit more complicated.

            In addition to that, it may also be wise to preheat the whole board a little before trying to remove anything. If you have an electric stove, that works great - just place the video card on something a few centimeters above the heater and put it on low heat. If you have a gas stove, this won't work, though (as gas tends to concentrate all of the heat in the middle).

            Also, as far as which MOSFET to start with - just look for the one that looks the most discolored or the one that has tiny cracks or holes on it (if any!). If they all look fine... then just pick one and go with it.
            Last edited by momaka; 02-10-2014, 08:39 PM.

            Comment


              #7
              Re: Sapphire HD 7870 possible short

              Okay so Fairchild denied my request for a free sample not sure why? So I went ahead and ordered them from Mouser with some stupidly over priced delivery charge £12, so I order a few to make it worth my while. Why is the delivery from these chip companies so high? Even from my aliexpress account the Chinese where quoting something crazy to send something weighing next to nothing and costing pennies. I've ordered hundreds of capacitors and delivery was really cheap.
              If someone could shed some light on that I'd appreciate it.

              So back to the topic at hand, I was wondering what the chances were of more than one of these mosfets failing? And wether or not they can cause the failure of any other component in the vrm.

              how much heat from hot air desoldering can the surrounding components take? Especially those r19 inductors?
              Does this type of package normally connect to thermal vias?
              I'll probably be using a skillet to preheat the board. Is it wise to cover components with kapton tape to protect them?
              I'm also going to practice on an old junk board or a practice kit.
              Any advice/recommendations on which desoldering station to go for, probably a multi function one with a nice high wattage iron? I was always going to upgrade to one it's not just for this project. I wanna get into Bga repairs eventually, but one step at a time.

              If I manage to pull this off somehow I owe you big time momaka, do you accept bitcoin donations?
              Last edited by modman1337; 02-11-2014, 10:50 AM.

              Comment


                #8
                Re: Sapphire HD 7870 possible short

                Sorry I haven't responded for a while. Didn't have much homework at school but was procrastinating it, so you know how it goes .

                Originally posted by modman1337 View Post
                So I went ahead and ordered them from Mouser with some stupidly over priced delivery charge £12, so I order a few to make it worth my while. Why is the delivery from these chip companies so high?
                Wow, that's quite a bit indeed. I sure hope they send you the part in a gold-plated envelope or something. I order from Mouser regularlry, but here in the USA, shipping with them is not expensive. Same with Digikey. Did you try Farnell, RS Components, or some of the other more popular part stores in the UK?
                On a tangent, I bought two of my required university text books online. One of them came from India via DHL in less than 3 days and the total shipping was under $7 (and under $4 if I wanted "regular" 5-day shipping, lol). Keep in mind, this textbook is an engineering book, so it's got some weight to it.
                /tangent

                Originally posted by modman1337 View Post
                So back to the topic at hand, I was wondering what the chances were of more than one of these mosfets failing?
                Very small. Usually when one of the MOSFETs shorts, it causes an over-current condition on the 12V rail, which quickly triggers the PSU short circuit or over-current protections.

                Originally posted by modman1337 View Post
                And wether or not they can cause the failure of any other component in the vrm.
                If they were just plain old MOSFETs without drivers, they can sometimes take out the buck controller. But since these have integrated drivers, I doubt has this happened.

                Originally posted by modman1337 View Post
                how much heat from hot air desoldering can the surrounding components take? Especially those r19 inductors?
                The most sensitive components in that area will be the caps. I would suggest you shield them with a piece of aluminum foil on the side (but don't cover them completely, so if they get hot, they will be able to radiate some of the heat. The inductors should be fine with the heat, I think. At least I haven't damaged any so far.

                Originally posted by modman1337 View Post
                Does this type of package normally connect to thermal vias?
                I don't know for sure.
                All I can tell you is those MOSFETs use the large copper planes on the board as heatsinks, so make sure they get soldered to the board.

                Originally posted by modman1337 View Post
                I'll probably be using a skillet to preheat the board. Is it wise to cover components with kapton tape to protect them?
                Skillet will work. If you have an electric stove in your kitchen, that will also work. Ours has 1.5 KW burners (or thereabouts) for the small elements. Set on medium, they can bring the board temperature up to 140-150C easily. I highly recommend you get a thermocouple, though. Type-K are cheap on eBay, particularly the TM-902C, like this:
                http://www.ebay.com/itm/TM-902C-Digi...item53f193d769
                I have one and my friend uses quite a few of them in his repair shop. They are quite well-made, actually. The only problem is it may take a while for them to get to you if you order from someone in China on eBay.

                Originally posted by modman1337 View Post
                Any advice/recommendations on which desoldering station to go for, probably a multi function one with a nice high wattage iron? I was always going to upgrade to one it's not just for this project. I wanna get into Bga repairs eventually, but one step at a time.
                If you're looking at the 2-in-1 stations (that is, hot air and soldering station), I wouldn't spend too much on one because they are all the same more or less. I have a KADA 852D+. If you have something on which to pre-heat the boards, the hot air is adequate enough for reflowing/reballing small BGA chips (such as Northbridge and Southbridge in newer laptops) as well as RAM chips. The hot air is also adequate for putting solder balls on bigger chips after they have been removed. But for larger chips like the Xbox 360 GPU or (please DON'T try this) a PS3 GPU/CPU, it's not powerful enough. Similar results when I tested a YiHua/Saike 898D -type station.

                As for the iron that comes with these soldering stations - it's not very powerful, especially if it uses those clone Hakko tips (I think all do), but not bad either. It all depends on the quality of the heating element. The one that came with my KADA is very weak. The 898D -type stations I tested were okay - I was able to remove MOSFETs from an Xbox 360 with 2 of these irons just by globing up the pins with solder and heating the MOSFET from both sides.

                That said, if you do want to get into more serious BGA work, you should look into a rework machine that has it's own bottom and top heaters. Dark IR are fairly cheap these days. Can get a fairly decent station for around $500.

                Originally posted by modman1337 View Post
                If I manage to pull this off somehow I owe you big time momaka, do you accept bitcoin donations?
                Nah, don't worry about. Getting this one fixed is enough to make me happy .
                Last edited by momaka; 02-14-2014, 07:54 PM.

                Comment


                  #9
                  Re: Sapphire HD 7870 possible short

                  Hi,

                  Interesting read for someone like me who thought he was in the same boat. Unluckily it's not my fets or anything powerside as this is a quick farnell order. (free shipping on all orders, but the minimum is £20) I'm an EE student and hobbiest and have reflowed some stuff. awaiting my new reballing stencils for another project and fear I may need them on this card I picked up.

                  Don't know the history on this card but have scoped and measured all the power rails (CORE/DDR/5Vreg/3.3/both PCI-E 6-pin) and worked out where each goes and does what (apart from the 5vreg, I suspect it's used as a ref for the CHiL 8225G) was a bit worried when i spotted 5.5 ohms on the core rail but I get voltage when the buck driver wakes up.

                  Long story short : Before I reflow/reball I think TP101 between the core and PCI is open circuit and TP102 is lower ohms than the rest. This is bad due to it being Transmitter Lane 0 HSOp(0) part of the differential pair. maybe it would boot is it was lane 4 and above but could I ask modman or momaka or anyone reading to measure the resistance between TP101 and ground and post results? Unintentional hijack sorry; It's just this is the most technical post on these cards on the whole internets according to google. and I cant read another "just RMA it" without popping a cap(acitor) in anger

                  Cheers,
                  eekcage

                  Comment


                    #10
                    Re: Sapphire HD 7870 possible short

                    Hi eekcage,


                    don't worry about the hijack, the more we can find out about these cards the better, like yourself I could barely find any technical / repair info on the net that's why I came here and asked the questions myself. Ill check TP101 for you in a bit and give you my reading.

                    Would you mind sharing your knowledge on the lanes / differential pair? Am I right in thinking TP101 stands for 'test point'? If we can map out all 'normal' test point readings on a working card it may well provide a source of reference for others.

                    Comment


                      #11
                      Re: Sapphire HD 7870 possible short

                      okay so I went ahead and tested TP101 -104 for you eekcage, heres what I get.

                      TP101 - 3.56
                      TP102 - 4.58
                      TP103 - 4.10
                      TP104 - 3.74

                      If you require anymore readings just ask, hope this helps.

                      Comment


                        #12
                        Re: Sapphire HD 7870 possible short

                        Hi OP and all other members .
                        I have been reading this thread for a while and its really interesting , video cards are really easy to fix once you know how they work , but fixing them must be a pain with all those SMD/BGA ICs .

                        So modman have you managed to find the bad mosfets ???
                        Last edited by SM-Piyes90; 04-09-2014, 08:12 AM.

                        Comment


                          #13
                          Re: Sapphire HD 7870 possible short

                          Hello,

                          since it's a similar porblem with the same card, i don't think i should start an new thread^^

                          To my question: There are 3 resistores gone from the board (i marked the positions at the Picture - i borrowed it from the first post^^ ) - so i need to know the specifications of these parts.

                          Can anyone help here to get these information?

                          thx so far^^


                          (sry for my bad english... im from germany and not well with other languages^^)
                          Attached Files

                          Comment


                            #14
                            Re: Sapphire HD 7870 possible short

                            I know this is more than a month late, but these are ceramic capacitors, not resistors. So if they appear open-circuit on your multimeter when they are removed from the board, that is normal.
                            Now, if they were broken off and you can't find them, just put ceramic capacitors that of similar size there. It's not ideal, but in most cases it will work. And in many cases, the video card will work without those ceramic capacitors.

                            Comment


                              #15
                              Re: Sapphire HD 7870 possible short

                              Okay... when ARE we going to solve the OP's question!?!?!?
                              Don't buy those $10 PSU "specials". They fail, and they have taken whole computers with them.

                              My computer doubles as a space heater.

                              Permanently Retired Systems:
                              RIP Advantech UNO-3072LA (2008-2021) - Decommissioned and taken out of service permanently due to lack of software support for it. Not very likely to ever be recommissioned again.
                              Asus Q550LF (Old main laptop, 2014-2022) - Decommissioned and stripped due to a myriad of problems, the main battery bloating being the final nail in the coffin.


                              Kooky and Kool Systems
                              - 1996 Power Macintosh 7200/120 + PC Compatibility Card - Under Restoration
                              - 1993 Gateway 2000 80486DX/50 - Fully Operational/WIP
                              - 2004 Athlon 64 Retro Gaming System - Indefinitely Parked
                              - Main Workstation - Fully operational!

                              sigpic

                              Comment


                                #16
                                Re: Sapphire HD 7870 possible short

                                @ TechGeek: When you come up with a better solution .

                                Lol, seriously man, why even post that?

                                Comment


                                  #17
                                  Re: Sapphire HD 7870 possible short

                                  interestingly user Trodas in MDL recapped older videocard with slightly better ratings and gets better OC from it.

                                  Comment


                                    #18
                                    Re: Sapphire HD 7870 possible short

                                    I was in a similiar situation but with a radeon 5770. Short-circuited on 12v 6pin conector. It was a bad mosfet ,but i replaced it with a used one with similar specs and it worked kind of , its artifacts in 2d/3d but the card also had 3 blown caps -changed with crapy Yageo, so now i don't now the cause of artifacts ,caps or mosfet .So that's why i'm here !

                                    Comment


                                      #19
                                      Re: Sapphire HD 7870 possible short

                                      Artifacting is probably bad RAM\GPU or bad solder joints on either, possibly even damage was caused when the FET blew.

                                      OR your new FET is not good enough and now the power supply to the chip is too low or something...
                                      "Tantalum for the brave, Solid Aluminium for the wise, Wet Electrolytic for the adventurous"
                                      -David VanHorn

                                      Comment


                                        #20
                                        Re: Sapphire HD 7870 possible short

                                        Hey, my first time here but i already see there are plenty of guys here that know what they're talking about. Anyhoo, I'm reviving this thread because i seem to have a similar problem. I have an R9 280X gpu that fell victim to what i think was a short circuit on the external PCIe power connectors. Here are the pictures of the card.
                                        http://imgur.com/a/t94Da

                                        I tried the card in several computers and none of them boot up (one of the PCs had a loudspeaker and beeped a GPU error, so it's certain that the card is faulty). I started investigating what could be the cause, and made a few following assumptions, after reading the thread:
                                        1. The three ribbed heatsinks on the left side of the card should be the inductors for GPU Vcc
                                        2. The 8 sapphire logo ribbed heatsinks on the right side of the card should be the inductors for RAM Vcc
                                        3. There are three FETs on the right side of the card for each RAM inductor, the bigger FETs are IRF6725, while the smaller ones are IRF6721


                                        I grabbed a multimeter and started measuring and investigating, according to momaka's instructions in his first post in the thread. Here are the measurements:
                                        1. Resistance between GPU Vcc rail (measured from the three inductors on the left side) and GND is 1.2ohm
                                        2. Resistance between RAM Vcc (sapphire inductors) and GND 0.6ohm
                                        3. Resistance between IRF6721 "smaller fets" DRAIN and 12V PCIe rail is 0.3 ohm
                                        4. Resistance between IRF6725 "bigger fets" DRAIN and GND is 0.6ohm
                                        5. Resistance between PCIe 12V and GND is in the order of few kilo ohms


                                        From this we can conclude that the smaller 6721 fets are the upper ones, and 6725 are the lower ones. Now, the main most important measurement are i think the ones in bold. Since FETs usually fail in such a manner that they short the drain and source pins, i concluded from the short circuit between 6725 drain and GND that this is indeed a blown lower mosfet , which is confirmed by the big resistance between 12v and GND on the external connectors.

                                        Anyways, i'm not an expert on such complex circuit and am basically just seeking some kind of confirmations that my detective work is correct. Therefore, i summon master momaka to the thread!

                                        Furthermore, if my diagnosis is correct, I was wondering if there is any easier way of finding out which fet is faulty apart from desoldering every one of them until the short is gone. I was thinking of maybe rather removing the 8 sapphire chokes so as to remove the connections between fets and the remaining parts of the board, and then sending some current through the fets to actually blow the faulty one to the point where its identifiable by naked eye (heat, smoke..?). But i must admit, it seems a very risky strategy, since i dont know what voltage to use, how the controllers for the FETs (which are on the backside of the card ) work and if there are any other connections to the board that bypass the chokes etc..

                                        Comment

                                        Working...
                                        X