Announcement

Collapse
No announcement yet.

AMD R9 270x - Gigabyte Windforce

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

    AMD R9 270x - Gigabyte Windforce

    Hi all,

    I have purchased a "spares or repair" AMD R9 270x on the bay, just as a fun project.

    The board worked in my test system. I ran some tests and all was well. When I tried it in my main desktop, it refused to display a picture.

    Back on the bench and it was booting up 50% of the time.

    I found a short on a line - on the right hand side of the board, not the main phases on the left. Injecting 1.5V I identified a shorted capacitor at the back of the board. I temporarily replaced it with an electrolytic (I measured the one close to the shorted one to guess the capacitance) and the short was gone.

    When I tried the board again, it was still booting up 50% of times - then it started behaving well and so far I haven't been able to replicate the issue.

    I notice that when I power up, the fans "try to spin" twice before they eventually spin normally. Also, when it fails to boot, the fans spin at full power - and the GPU gets very warm.

    My question is: how is it possible for a video card to work with a short on it? And is that fan behaviour normal for that type of card?

    Thank you!
    Attached Files

    #2
    Re: AMD R9 270x - Gigabyte Windforce

    I've made a video of the fan behaviour. You can see they try to spin twice before actually spinning.

    So far the card works - but I am still puzzled on how could it work with a short on it.

    https://www.youtube.com/watch?v=W0GAYJb_JzM

    Comment


      #3
      Re: AMD R9 270x - Gigabyte Windforce

      I have this card, haven't had this problem though two of the fans on mine stopped spinning, until I took them apart and WD-40'd the bearings. At first they still didn't work but then they started up. Sorry this is no help

      Comment


        #4
        Re: AMD R9 270x - Gigabyte Windforce

        Thank you - any input helps!

        So are you saying that your card does not show the same fan behaviour? It's not a mechanical issue, I see there is no voltage going to the fans until they start spinning.

        Regarding the card not working in my main system, I see many posts online mentioning some BIOS incompatibility so it may just be my system not liking that model.

        As it seems to be working on another test kit I have, I'd say the card is working fine. Still, I fail to understand how a SHORTED CARD could have worked in the first place - even thought 50% of the time.

        Comment


          #5
          Re: AMD R9 270x - Gigabyte Windforce

          Originally posted by tony359 View Post
          Thank you - any input helps!


          As it seems to be working on another test kit I have, I'd say the card is working fine. Still, I fail to understand how a SHORTED CARD could have worked in the first place - even thought 50% of the time.
          Well a graphics card will have several voltage rails and various places where you could have a short. But regardless of that I am also somewhat puzzled how a short could cause an intermittent problem.

          I am assuming the short was not on the Vcore supply to the GPU otherwise you couldn't have found it the way you did.
          Follow me on YouTube
          ------------------
          Learn Electronics Repair
          https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

          Comment


            #6
            Re: AMD R9 270x - Gigabyte Windforce

            Originally posted by tony359 View Post
            As it seems to be working on another test kit I have, I'd say the card is working fine. Still, I fail to understand how a SHORTED CARD could have worked in the first place - even thought 50% of the time.
            I suspect that cap you removed may not be shorted. It appears to be a filter for the VRAM V_dd rail, so you might have been measuring the low resistance of the VRAM. After removing the "shorted" ceramic capacitor, did you check it to see if it's still shorted? If no, then check its capacitance? If it shows any capacitance (should be somewhere in the 1-20 uF range, probably), then that wasn't the issue.

            Also, the electrolytic cap you soldered with those long leads will be ineffective at high frequency filtering, due to extremely large XL from the long lead - so much that the circuit probably won't even "see" the capacitor... so you might as well not bother installing it. You need either a similar SMD ceramic, or it just won't matter otherwise.

            Lastly, I've been seeing a "rash" of dead R9 video cards for the last year on eBay - more dead ones that working ones, actually. In particular, it seems that HD7800, HD7900, R9 200 and R9 300 series are just not well-made cards. I haven't been able to pinpoint if it's the GPU or RAM dieing in these, but I suspect it may be the GPU more likely. I have an R9 290 and a R9 280 - both appearing to work OK until the drivers are loaded in Windows. After that, it's either artifacts and crash or insta-crash.

            *EDIT*
            I've seen a similar "fans spin up-and-down" multiple times on faulty ATI/AMD Radeon cards. On the older ones (HD4k, HD5k, and HD6k) that I'm more familiar with, the fans don't really twitch like yours, since they are designed in such a way that if VGA BIOS does not boot, then fan(s) run at 100%. In any case, some of my dodgy HD4850's and HD4870 love to multi-spin the fan on boot before they show video (up to 5x, sometimes)... and these were faulty GPUs that I have reflowed... so probably still dodgy. On supposedly good HD4850/70, I don't see that behavior. So yours could well be a video card doing multiple boot attempts before finally working.

            Give it a stress-test with FurMark or some other heavy 3D benchmark and see if the card croaks. If it does, it's probably just another of the 10's of failed R9 cards I've seen so far this year.
            Last edited by momaka; 02-16-2021, 02:50 AM.

            Comment


              #7
              Re: AMD R9 270x - Gigabyte Windforce

              Thanks for your input

              dicky96, momaka

              I found the short by injecting 1.5V at 300mA on that rail and I saw the capacitor glow on my thermal camera. I removed it and tested and yes, it is definitely shorted. I guessed the capacitance by removing the adjacent one and measuring it - 25uF.

              Once the capacitor was removed, I did not measure the short anymore.

              And yes, I am aware that very low resistance is normal on a GPU, I do have 7 Ohms on the VCore rails indeed. But on that rail I had 0Ohms.

              Cool to know about the long leads - I just didn't want to leave it empty. A proper replacement is on its way.

              Too bad about the idea that the attempt to spin may be a sign of early failure. I have run Unigine for several hours, everything seems fine, the temperatures are fine.

              I've also run furmark - not for very long to be honest - and all seems to be ok. This is why I am a bit puzzled.

              I wanted to test the card in my main PC - so it would be on for most of the day - but, again, it won't boot up.

              Comment


                #8
                Re: AMD R9 270x - Gigabyte Windforce

                Originally posted by tony359 View Post
                I found the short by injecting 1.5V at 300mA on that rail and I saw the capacitor glow on my thermal camera. I removed it and tested and yes, it is definitely shorted.
                Ah oh, just checking. I wasn't trying to demean your repair skills in any way. It's simply that I often see people get tripped by this (typically because they use continuity setting on their multimeter instead of properly measuring resistance), so that's why I had to check.
                Your findings, of course, confirm I was wrong with my assumption, and indeed the cap was bad (shorted).

                Originally posted by tony359 View Post
                Too bad about the idea that the attempt to spin may be a sign of early failure. I have run Unigine for several hours, everything seems fine, the temperatures are fine.
                Well, it could also be just something about the way the fan profile is set up in BIOS on boot. Like I mentioned, I'm not as familiar with the R9 and RX cards as much as I am with the older HD series. So the fan "twitching" could be normal.

                Originally posted by tony359 View Post
                I've also run furmark - not for very long to be honest - and all seems to be ok. This is why I am a bit puzzled.
                Hmmm, so it's not failing under load.

                In that case, maybe check if the BIOS is the original stock one. Perhaps the seller / previous user flashed it with something else incorrectly or played with the various power states.

                If you have an oscilloscope, it would also be interesting to measure the ripple output on the V_core and V_ddq (RAM supply) to see if there is perhaps excess ripple somewhere. Also, maybe check GPU V_tt(s) - there may be several of these on AMD cards. Typically they are generated from the V_ddq and 3.3V rails with linear regulators (at least on old HD series cards.) Probably OK, though, as I haven't seen one of these regulators go bad to date on the older cards.

                Comment


                  #9
                  Re: AMD R9 270x - Gigabyte Windforce

                  Momaka

                  Don't worry, I didn't take your comment in a bad way - in fact, thanks for your input.

                  Interesting about the BIOS.

                  But I have a new element.

                  I've tried the card on a different MB, an i7 system so much younger. The startup behaviour is completely different. The fans spin immediately at start up and stay at full speed until the system has booted up. And, so far, I got video out of it every time. No signs of that "glitch" I video'd before.

                  Maybe I should indeed check that the BIOS on that board is correct but this is quite interesting. Any chance that the other board did not have enough PCI power for it? Still, the board would not display a video out of my main i7 Desktop system. Interesting!

                  I do have an oscilloscope and I can check the ripple indeed. The fun part is going to be to find where to check! I think I can check out of the chokes?

                  Comment


                    #10
                    Re: AMD R9 270x - Gigabyte Windforce

                    little update - no, the board does not start when "hot", I need to power cycle the system to have it working again - only a few seconds turned off will do. Even a restart will trigger a blank screen. But the fans behave ok this time.

                    I have managed to re-flash the same version of the BIOS on it, there is nothing better available. But that didn't change anything.

                    I will keep trying - and I'll check the ripple as suggested.

                    Thx!

                    Comment


                      #11
                      Re: AMD R9 270x - Gigabyte Windforce

                      Here is how the fans behave with a different motherboard.

                      Nothing weird. Fans spin at full-ish speed until the driver kicks in and they slow down.

                      Usually when I restart the system from within Windows, the fans stop to then start again showing the same behaviour, but I have no picture.

                      In that case, I need to power cycle the system.

                      https://youtu.be/aoosm7ccQJo

                      Comment


                        #12
                        Re: AMD R9 270x - Gigabyte Windforce

                        Ok, I tried the R9 on a different motherboard, an Asus P5t with a Core2 Quad on it. It wouldn't work at all. Fans glitch as mentioned before and no video ever.

                        The same R9 on a similar Asus (P6T) and an i7 920 works most of the time.

                        Ripple on GPU seems to be ok, the spikes I see are there anyways even if I probe the ground chassis so I suppose it's RF from the supply maybe?

                        What I notice is that when the GPU starts, the GPU Vcore starts at 1.5V and then drops to 1.17V when the video is output. When the GPU does not start, the voltage stays at 1.5V all the time.

                        As the PSU is the same in my tests, I wonder if that card is somehow upsetting the PCIex power supply rail? Maybe something is wrong there and the card exceeds the max power the PCI slot can supply?

                        I really don't understand that kind of behaviour.
                        Attached Files

                        Comment


                          #13
                          Re: AMD R9 270x - Gigabyte Windforce

                          Originally posted by tony359 View Post
                          Ripple on GPU seems to be ok, the spikes I see are there anyways even if I probe the ground chassis so I suppose it's RF from the supply maybe?
                          Are they there when the PSU is off and/or unplugged too? If no, then something in the system is causing them (PSU itself or other component.) They don't look bad, though.

                          Originally posted by tony359 View Post
                          What I notice is that when the GPU starts, the GPU Vcore starts at 1.5V and then drops to 1.17V when the video is output. When the GPU does not start, the voltage stays at 1.5V all the time.
                          That's typical behavior for many GPUs - they always start with a high core voltage until their BIOS loads all the power states, and then the GPU determines what state to go in, based on what state the system is in too.

                          So the voltage being stuck high means either BIOS is not getting read or GPU chip is problematic and unable or too unstable to load the BIOS.

                          What's interesting is that this seems to be consistently (so far, at least) happening on your older P5T / Core 2 board but not on the newer P6T / i7 board.

                          Originally posted by tony359 View Post
                          As the PSU is the same in my tests, I wonder if that card is somehow upsetting the PCIex power supply rail?
                          Well, try a different PSU if you have one, just to see if you can get the same behavior to repeat on both boards.

                          Originally posted by tony359 View Post
                          Maybe something is wrong there and the card exceeds the max power the PCI slot can supply?
                          Unlikely.
                          Mother consumer boards don't meter the PCI-E slot power, and even not all workstation boards do. If you do a resistance test, you will see that the 12V and 3.3V power on the PCI-E slot on the motherboard are typically directly connected to the PSU and there are no shunt resistors anywhere to meter the current.

                          Originally posted by tony359 View Post
                          The same R9 on a similar Asus (P6T) and an i7 920 works most of the time.
                          That's the worst part - an intermittent issue.
                          This suggests even more the GPU chip may be going bad, along with the fact that the card won't start from a reboot when "hot".

                          On the other hand, it may also be worth cleaning the PCI-E gold fingers on the card that go into the PCI-E slot, just to rule that out. Pencil eraser wiping followed by cleaning with IPA is a good way to remove any residue or possible corrosion.

                          Comment


                            #14
                            Re: AMD R9 270x - Gigabyte Windforce

                            thanks again for your input.

                            I'll check the spikes.
                            The R9 also refuses to start on my i7 second gen mobo which I use as main desktop system.
                            I have tried a different PSU and cleaned the connector several times unfortunately.

                            I want to invest a bit more time on that board as I thought I found an area which once warmed up would cause the issue to appear quite consistently. But I could not identify a single component.

                            Comment

                            Working...
                            X