47 Replies Latest reply on Apr 22, 2016 12:00 PM by defectivebydesign

    Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.

    latene

      Both GPU's were running BF4 when my computer shut down. The 290 produced an obvious burn smell from the power connectors so I wiped my drivers and used onboard video for a month. I retrieved my old Diamond 7870 and installed fresh drivers, of course, it ran flawlessly until suddenly my PC shut off and now will not recognize the GPU (no signal).

       

      I would most likely have been running the previous (bugged?) Crimson drivers when the r9 290 failed but I was running the latest version (non-whql-64bit-radeon-software-crimson-16.4.1-win10-win8.1-win7-apr4) when just last night the 7870 failed.

       

      Win 10 64bit

      ASUS P8B75-M/CSM

      Intel Core i5-3550 @ 3.3ghz

      Antec High Current Gamer M 750 Watt PSU

      4 X 4GB Kingston HyperX 1600mhz RAM

      Hitachi DeskStar 320GB 7200RPM HD)

        • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
          Hardwood

          Even the best PSUs can develop issues. I'd be much more suspect of the PSU.

          I've used most of the latest Radeon drivers, including 16.4.1., with my 290x on win10x64and there's no bugs/issues I've come across.

          • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
            daznash

            I would suspect the same. After I updated to latest crimson my pc refused to start, i had to do a fresh OS install to get them working, and i suspect it blew one of my 3 cards also.

            • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
              brucer

              I wouldnt..

               

              AMD released drivers that set default fan speed down to 30% and target temp up to max.  Which would probably burn up older video cards.

               

               

              The sad thing is you moderators know its a fact..

              • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                defectivebydesign

                16.4.1 is burning out videocards. DO NOT USE!

                I just had the same problem with my R9 290. Was playing Warhammer: Vermintide on 16.4.1, and BOOM! BSOD. Computer would not boot, bad VGA. Had to RMA the card.

                Edit: In retrospect, I may have been using 16.3.2. I have 16.4.1 downloaded though, so I'm not 100% sure which driver I was using when the card burned out. Regardless, this seems to be a trend with recent drivers. You don't have multiple cards from different users suddenly die out unless there is a driver bug causing the problem.

                 

                Thanks, AMD. At least I'll likely get upgraded to a newer model via the RMA process.

                  • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                    latene

                    I've reset the CMOS and attempted to run the 7870 using new cables on a new power slot. No luck. colsedav has emailed me a long list of trouble shooting advice and I've decided it's worth doing the work because the alternative is plugging in a new GPU without verifying my existing hardware is in proper working condition. Admittedly, I suspect the AMD GPU's and perhaps their drivers, that being, after having scrubbed the web to discover PC builders who've fail tested parts (especially with respect to r9 200 series GPU's) and produced very disappointing failure rates. PugetSystems shows Nvidia Maxwell parts at 2.3% failure while AMD r9 200's are at 18%for the year 2014.  their figures for my specific r9 290 maker, VisionTek, is at a terrible 25% failure rate. Obviously, people will reject or suspect these findings but I am rather confident that the first generation Hawaii I.E. r9 200 series GPU's (for whatever reason) were a bad batch. I'm still curious as to whether or not my Diamond 7870 will still function since it did not smoke, but, frankly, I'm ready to switch to Nvidia Geforce 970. For now I'll use onboard graphics for Civ 5 game play https://www.pugetsystems.com/labs/articles/Most-Reliable-Hardware-of-2015-749/#VideoCard

                  • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                    black_zion

                    The fan speed bug was only present in the very first Crimson release and was fixed days later, it is not present in any driver set released this year. I would agree with Hardwood that it's a PSU problem.

                    2 of 3 people found this helpful
                    • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                      colesdav

                      I do not think it is the Crimson drivers at all.

                       

                      HD 7970 and R9 280x running fine on pretty much every AMD Driver release since the past 2.5 years.

                       

                      In your case say something was wrong with the driver fan profile.

                      Even if your driver turned off the fans completely for example.

                      If your GPU Temps went up to over 95'C are you saying that your GPU self protection circuit failed to start because of the Driver?

                      Is this possible?

                      I have not seen a bug like that reported.

                       

                      One thing I would try is to re-flash your motherboard BIOS.

                      I had an instance where my PC crashed out and would not POST unless I removed my GPUs.

                      I re-flashed the motherboard BIOS and everything worked fine including the GPU's after reinstalling.

                       

                      colesdav.

                      • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                        colesdav

                        Also are you certain you went into the BIOS / Windows display settings and switched your Displays over from

                        iGPU to use the Pcie graphics?

                        colesdav

                          • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                            latene

                            I'm not familiar with the technology to know but I was assuming the drivers would have to play an active and correct role in any protective measures related to overheating. Perhaps that's not the case. However, there are quite a lot of stories going around about dead GPU's following the recent driver update and these stories seem to coincide with my troubles, that being, when I trace back my GPU/gaming usage I can see I would perhaps have been running the suspect drivers.

                             

                            But I will attempt a bios flash and look in to my windows display settings. (I'm not very hopeful about the 290, though, since it was a rather powerful smoke)

                              • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                colesdav

                                RE: I'm not familiar with the technology to know but I was assuming the drivers would have to play an active and correct role in any protective measures related to overheating. Perhaps that's not the case.

                                - OK. You might wan't to ask your Graphics Card tech support directly, But there should be an on board hardware only protection mechanism, that once a temperature is hit, switches your GPU off completely.

                                - What were your GPU Temps when you were running BF4?

                                - Did you hear your fans running or not?

                                - Have you tested the Fans on your GPU's separate to the Graphics cards themselves. You might have had a fan failure followed by GPU thermal shutdown.

                                 

                                RE: However, there are quite a lot of stories going around about dead GPU's following the recent driver update and these stories seem to coincide with my troubles,

                                 

                                - Don't know about that. Hardware does fail eventually. Could just be coincidence.

                                 

                                RE: But I will attempt a bios flash and look in to my windows display settings.

                                Good luck. Hope it works for you.

                                 

                                RE: (I'm not very hopeful about the 290, though, since it was a rather powerful smoke)

                                Was it emitting clouds of smoke? If it really was that bad you might want to think twice about even connecting it to your Pcie bus.

                                Maybe your GPU manufacturer could offer to test  it for you, even if it is out of RMA.

                                Just a thought.

                                 

                                colesdav.

                            • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                              colesdav

                              Final thought,

                               

                              That Motherboard comes with Lucid Virtu software, did you have that installed / running?

                              If so, try uninstall Lucid Virtu and reinstall Crimson.

                              bye

                              colesdav,

                                • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                  latene

                                  - What were your GPU Temps when you were running BF4?

                                  - Did you hear your fans running or not?

                                  - Have you tested the Fans on your GPU's separate to the Graphics cards themselves. You might have had a fan failure followed by GPU thermal shutdown.

                                  I can't hardly speak to these questions since I was wearing a headset while gaming and not monitoring fan speeds or temps. Although, I feel like my computer was running quiet.. I do get an impression fans were at low speed.

                                   

                                  I'm a bit hesitant about running a bios flash. I don't think there's any issue with my display settings fixed to my onboard GPU; the plug and play kind of functionality works the way it's supposed to.. i.e. when I plug my AMD GPU's in they are automatically detected - even though useless, now.

                                   

                                  No, there was not a cloud of smoke but it smelled bad. Bad enough that I pulled it out right away and very hesitatingly (after inspecting and seeing no visual signs of burning) attempted to boot with the GPU again.

                                   

                                  Nope, Lucid Vurtu is not installed.

                                   

                                  To be frank, when a PC is running flawlessly but you install new video drivers and your GPU smokes you then suspect the GPU/drivers. I'd consider trying a bios flash if this were an issue with not recognizing my hardware but I'm clearly dealing with GPU hardware failure. Obviously, I can't know for certain what's causing it but since VisionTek is not interested in helping me out and my Diamond 7870 is also out of warranty I feel like my best move is to switch to Nvidia.

                                    • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                      colesdav

                                      RE: I can't hardly speak to these questions since I was wearing a headset while gaming and not monitoring fan speeds or temps. Although, I feel like my computer was running quiet.. I do get an impression fans were at low speed.

                                       

                                      Yes I totally understand that. Lots of PC users wear headsets, especially when gaming with case and GPU fans running at high or max r.p.m.
                                      That's why it is so important for a Fan Control User Interface to be really careful to double check with the User if there has been any change to the Fan Control set up. Every potential change made should be confirmed. This has caught me out quite recently when using Fan Speed control on Crimson User Interface.

                                      I also think It would be good to add a visual warning message feature to Crimson User Interface if the GPU fans run below a certain threshold set by the User.

                                       

                                      RE: I'm a bit hesitant about running a bios flash. I don't think there's any issue with my display settings fixed to my onboard GPU; the plug and play kind of functionality works the way it's supposed to.. i.e. when I plug my AMD GPU's in they are automatically detected - even though useless, now.

                                       

                                      OK. It was just my advice. If you feel uncomfortable about doing it I understand. Totally fine. I am still apprehensive about running BIOS flash unless it is absolutely necessary. If your PC suddenly crashes it is possible the BIOS will get corrupted.

                                       

                                      I am not sure if your Motherboard supports USB BIOS Flashback?

                                      Mine (ASUS Z87) does. All you do is download your BIOS from the ASUS Support Downloads page, Rename the BIOS file. Load the BIOS file onto a USB Stick. Unmount the Stick. Power down your PC. Plug the stick into a special USB port on the back of the motherboard. Press a button. The Bios is updated in around 1 minute. It has worked for me flawlessly so far. I must have run about 5 BIOS updates and updated the BIOS after crash ~ 4 times.

                                       

                                      RE: To be frank, when a PC is running flawlessly but you install new video drivers and your GPU smokes you then suspect the GPU/drivers. I'd consider trying a bios flash if this were an issue with not recognizing my hardware but I'm clearly dealing with GPU hardware failure. Obviously, I can't know for certain what's causing it but since VisionTek is not interested in helping me out and my Diamond 7870 is also out of warranty I feel like my best move is to switch to Nvidia.

                                       

                                      It might also be the case that your cards had both been heavily used in the past, and nearing the end of their life anyway.

                                       

                                      You updated your drivers, and it is possible that they were actually running fine.

                                       

                                      Then you played an Intense BF4 session. There are lots of potential issues that can kill a GPU, including software exercising particular areas of the GPU heavily.

                                      Say one of the GPU's were suffering from an almost complete electromigration failure somewhere on the chip already and your intense BF4 session finished it off.

                                      Not a nice situation I agree. Certainly possible though. It is very difficult to put the blame on any one cause.

                                       

                                      Perhaps if one of the fans failed on one card whilst you were playing it caused the temps to rise.  That might be really easy to check.

                                      You might be able to remove remove the Fan housing from your GPU and there is likely a standard 3/4pin fan connector.

                                      Just plug that into a spare motherboard header and see if the fans spin when you boot up your PC. 

                                       

                                      If there were some failure already in progress on the GPU and the Drivers were running the fans at lower than expected speed, that would certainly cause GPU temps to be higher than they would have been, but again by how much. 10'C? O.K. might accelerate a pre-existing failure but not really certain it could be said to kill it.

                                       

                                      Anyhow good luck. I think all of the above issues are pretty standard for GPU's in general. I am sure NVIDIA cards also fail.  I have seen some data suggesting the failure rates from both manufacturers are similar.

                                       

                                      My AMD cards still run fine, (crossed fingers). I have, however, decided that running 3DMark Benchmarks with overclocked GPU's may not turn out to be be the cheapest hobby I have picked up recently.  

                                       

                                      Best of luck to you.

                                      colesdav

                                       

                                       

                                       

                                        • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                          defectivebydesign

                                          Neither latene or I were running the original Crimson driver that had the Fan Bug. Fan control should have been working perfectly fine, and even if it wasn't, the 290 will THROTTLE DOWN to stay within it's thermal limits. There is literally no way you can kill these cards via overheating over a short period of time. You would have to run the card 24-7 @ 90C for several days to kill it. I KNOW my card was not overheating, because I was monitoring the temps. Upper 70's to 80, but nowhere near 90c, which is the throttle limit.

                                           

                                          Many modern cards have quiet fan profiles that do not kick up until temps reach 80c. I can concede that happens, but it is a manufactured FEATURE of those cards, and it shouldn't cause a failure. In fact, it DIDN'T up until these latest drivers. There is NO NEED to mess with a BIOS, and I don't even know why that would come into question.

                                           

                                          Electromigration. Do you even know what causes that? Overvoltage. Nobody here was overvolting their card. HOWEVER, that doesn't mean the DRIVER wasn't doing it, which IMO could actually be what caused it. AMD has recently been having PROBLEMS with their power management, where people have previously used CLOCKBLOCKER to fix, so AMD has recently introduced a SIMILAR FEATURE to "FIX" this issue. Now i say "F-I-X", because they NEVER FIXED IT, AND ONLY INTRODUCED A WORKAROUND to the problem. My Hypothesis here is that AMD's power management is NOT working correctly, and they made it WORSE, instead of addressing the underlying issue. This could be an issue with Voltage Regulation, and if AMD's recent driver OVERVOLTED our cards, THAT could absolutely cause instant death, ESPECIALLY if some manufacturers were using low quality VRM's or something. Perhaps the drivers caused a voltage SPIKE, because of poor power management. That would definitely explain his card literally burning out, because standard overheating will NOT cause your POWER CONNECTOR to melt.

                                           

                                          I have not had any burning issues myself, but then again I am using a PSU that has all kinds of voltage protection built in, including dual rails instead of single. That doesn't mean my PSU will protect my video card from killing itself via Suicide, it just means that the PSU itself will not burn out the video card, and that there is no single point of failure that can take out the rest of the system like a single rail PSU. I did not experience any burning with my hardware, only a single BSOD while playing a game, and then the card was dead. There were no lightning storms, and even if there was I have a fairly decent APC surge protector, not to mention the safety features still built into my PSU. The only thing that killed my card, was the card itself, and I strongly suspect that the recent drivers were over-volting the GPU, albeit in a non-obvious way. I don't think it was even constant, but a random surge, because I occasionally do check my voltage levels. Just not on a regular basis, because WHY SHOULD I NEED TO. Guess I now have to add Voltage to the list of things to permanently monitor in Real Time for AMD hardware.

                                            • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                              colesdav

                                              Wow, calm down. I will try to respond without using block capitals. I believe that indicates shouting.

                                               

                                              RE: There is literally no way you can kill these cards via overheating over a short period of time. You would have to run the card 24-7 @ 90C for several days to kill it.

                                               

                                              I do not think that a brief or even long term run of the cards at temps of 90'C will in itself kill the cards. All it may do is worsen any pre-existing defect on chip.

                                              If I run my cards with fans manually set off and they hit 97'C the fans kick in at 100% and keep the temps there.

                                              I believe that if my fans failed at 97'C and the temps rose higher the card will automatically shut down at around 110'C region anyhow.

                                               

                                              RE: Many modern cards have quiet fan profiles that do not kick up until temps reach 80c. I can concede that happens, but it is a manufactured FEATURE of those cards, and it shouldn't cause a failure.

                                               

                                              Agreed. I didn't say anything else. I don't think that running the GPU's at these temps is really any concern.

                                               

                                              RE: There is NO NEED to mess with a BIOS, and I don't even know why that would come into question.

                                               

                                              It was an earlier suggestion made by me if the PC didn't POST with the GPU's in place after a crash then reflash the BIOS. It happened to me and it fixed the issue.

                                               

                                              RE: Electromigration. Do you even know what causes that? 

                                               

                                              High DC Current density at high temperature is a major reason for electromigration on an integrated circuit.

                                              Overvoltage can cause increased DC current.

                                              So can a clock waveform with > 50% mark space ratio at rated voltage. This can happen for various reasons.

                                              So can various other digital signal combinations resulting in greater than expected 'on' versus 'off' signals at rated voltage.

                                              Even if a designer uses the latest and greatest IC Design tools they cannot guarantee it will never occur in real life.

                                              Minor processing defects on the GPU could result in a thinner than expected wire on a metal track for example.

                                              Or, a contact or via could be created with some misalignment, which, over time, will fail due to electromigration.

                                              I only mentioned electromigration as one of many potential temperature related failure mechanisms.

                                               

                                              RE: AMD has recently been having PROBLEMS with their power management, where people have previously used CLOCKBLOCKER to fix, so AMD has recently introduced a SIMILAR FEATURE to "FIX" this issue. Now i say "F-I-X", because they NEVER FIXED IT, AND ONLY INTRODUCED A WORKAROUND to the problem. My Hypothesis here is that AMD's power management is NOT working correctly, and they made it WORSE, instead of addressing the underlying issue. This could be an issue with Voltage Regulation, and if AMD's recent driver OVERVOLTED our cards, THAT could absolutely cause instant death,

                                               

                                              I have seen some info about Clockblocker. I am not using it.

                                               

                                              I was not aware of this new similar feature from AMD  - where can I access it in AMD Crimson / Radeon Additional settings?

                                               

                                              Again not sure about seeing any issues or announcements from AMD to say that their Drivers are killing peoples cards with such issues.

                                              I am pretty sure if this was happening we would have had an official 'recall announcement'  or something to warn their users about it from AMD. 

                                               

                                              colesdav.

                                                • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                  defectivebydesign

                                                  colesdav wrote:

                                                  I believe that if my fans failed at 97'C and the temps rose higher the card will automatically shut down at around 110'C region anyhow.

                                                  Misinformation. The 290 has a maximum allowable temp of 90c, where it will throttle clockspeeds to address the thermals. Even if the fan completely fails, the card will downclock to clockspeed corresponding to the temperature, and that will render most games completely unplayable. The user would know there was a problem, whether or not they were monitoring temps. Mentioning anything higher than 90C is pure disinformation and speculation. Nor is it related in any way to the actual situation at hand. There are no reports here of overheating, as you are projecting that non-existent issue.

                                                  colesdav wrote:

                                                  .It was an earlier suggestion made by me if the PC didn't POST with the GPU's in place after a crash then reflash the BIOS. It happened to me and it fixed the issue.

                                                  I doubt it was flashing the BIOS that fixed your issue, but the fact that flashing the BIOS could reset the CMOS. Resetting the CMOS is fine and dandy, and will occasionally fix problems, but please do not mix the two. If you had an actual problem with your BIOS, your motherboard wouldn't have posted in the first place. It was resetting the CMOS that fixed it.

                                                  colesdav wrote:

                                                  .I am pretty sure if this was happening we would have had an official 'recall announcement'  or something to warn their users about it from AMD.

                                                  That NEVER happens outside of massive failure rates, and happens even rarer for Nvidia users. Bumpgate, anyone? There is no guarantee people are even using the latest drivers either, let alone are running hardware that is affected. This could be something extremely limited in scope, and as such doesn't mean a "recall" (wrong term) is necessary. What it does mean is that there is a bug that could potentially fry people's cards in random situations, and we have NO IDEA whether or not AMD is investigating the issue. Meaning that there is no point using replacement AMD hardware, when it could just get fried at any random moment. So far, it seems only older GCN hardware is affected with current drivers, and is caused by a sudden and random event. Possibly related to AMD's power management issues, the clockspeed "FIX", and/or Crimson. Again, this current issue has no relation to the original Fan Speed Bug, which was SOLVED.

                                                   

                                                  Electromigration could very well be related to the failure. But you are not asking what CAUSED the Electromigration. I am. You do not have simultaneous failures caused by random Electromigration, you have a BUG that overvolts your GPU in random situations that causes Electromigration. From what I've heard from latene, his card had a fried power connector. That is NOT classic Electromigration, but an extreme power surge. These cards are being FRIED by sudden power consumption, and not by a slow natural death.

                                                   

                                                  As for HOW our cards got FRIED, I know it wasn't my PSU, or a power surge. Nothing tripped my surge protector, and my PSU is currently powering a power hungry Nvidia GTX 470 with no issues. The fault lies either with AMD's hardware, which I dunno why 3 different cards from 2 different people would have the same problem, or it was a DRIVER ISSUE that caused our hardware to fail. I'm 99.9% positive that the driver caused it, as I've dealt with various hardware failures before. My experience tells me that this was driver related. AMD screwed something up with Power Management, and we know they've been having issues. The latest drivers are supposed to "FIX" some of these issues, but I don't think anything has actually been fixed, and whatever changed is what burned out our cards.

                                                    • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                      colesdav

                                                      Hey come on, I am not fighting with you. I am trying to help. What's all this ''Misinformation" stuff.

                                                      I will respond to your points below, respectfully.

                                                       

                                                      RE: Misinformation. The 290 has a maximum allowable temp of 90c, where it will throttle clockspeeds to address the thermals. Even if the fan completely fails, the card will downclock to clockspeed corresponding to the temperature, and that will render most games completely unplayable. The user would know there was a problem, whether or not they were monitoring temps. Mentioning anything higher than 90C is pure disinformation and speculation. Nor is it related in any way to the actual situation at hand. There are no reports here of overheating, as you are projecting that non-existent issue.

                                                       

                                                      O.K. So, A GPU has a number of Power States, that are frequently advertised and discussed. For AMD, I can find plenty of examples about 'PowerTune' technology such as WhitePapers etc which discuss Power States such as Idle state (GPU almost completely turned off) Intermediate States (an example could be your graphics card running at minimal GPU clock and GPU MEMORY clock speed to run basic desk top applications) at the top end is the Highest Power State, which will be the maximum GPU clock and GPU memory clock.

                                                       

                                                      Powertune 1.0 Used on HD7900 series and R9 280X, and I believe it is also used on the 7870 which is used by latene

                                                       

                                                      PowerTune 1.0 estimates GPU chip TDP  and  adjusts clock frequency down if TDP is exceeded. The GPU Temperature on my HD7970 will normally run

                                                      at max CPU and GPU clock speeds using a normal fan profile at around 40-50'C when I run a stress test on the|GPU.. With PowerTune 1.0.  I can adjust Power Control Limit setting (i.e. the 'TDP'), and I can also set a Manual Fan Control.

                                                       

                                                      If I turn off my fans using Manual fan control and exercise the GPU the temps will rise to 96'C. The fans remain off. Then when the temperature hits 97'C The fans go to 100% and the GPU Clock frequency drops.

                                                       

                                                      Powertune 2.0 is used on the R9 290 and 290x for example. It is somewhat different in it aims to keep a constant fixed target temperature, the optimum temperature for the card to operate at, I believe. PowerTune 2.0 also monitors the GPU board temperatures and voltages, not just the Die, to make a more accurate  estimate of the total GPU Card power. I believe you can adjust your target temperature and target fan speed and Power Limit on the R9 290. 

                                                       

                                                      So I agree with you about the normal behavior you see on your card. That is: "The 290 has a maximum allowable temp of 90c, where it will throttle clockspeeds to address the thermals."

                                                       

                                                      Regarding:  Even if the fan completely fails, the card will downclock to clockspeed corresponding to the temperature.

                                                      Yes I can believe that will happen, up to a certain temperature level.

                                                      Did you actually override your GPU fan control somehow  to find out what happens?

                                                       

                                                      Now for the final point.

                                                       

                                                      RE: Mentioning anything higher than 90C is pure disinformation and speculation.

                                                       

                                                      Oh no it isn't.

                                                      Disinformation? - You mean you think I am deliberately lying do you? Why would I do that?

                                                      Speculation? - The only thing I was speculating about was the exact temperature the final trip would kick in - "I believe that if my fans failed at 97'C and the temps rose higher the card will automatically shut down at around 110'C region anyhow."  - It is probably more like 110-120'C but I don't know exactly, and I have no intention of

                                                      trying to find out using my cards. 

                                                       

                                                      I believe modern  Nvidia and AMD GPUS (Similar to Intel and AMD CPU's) include an emergency thermal trip mechanism which uses temperature sensors on board the die / board to perform an emergency shutdown and a temperature higher than the CPU or GPU throttle temperature. It shuts the CPU or GPU down.

                                                       

                                                      This is done as a last resort safety mechanism implemented in hardware to avoid completely destroying the CPU or GPU die, in case the normal thermal control mechanism fails. It is done for Health and Safety reasons, to stop, for example, a CPU or GPU starting a fire.

                                                      You can look up this fact easily for CPU's. It is well known. Just look for 'thermtrip' on either an Intel or AMD CPU.

                                                       

                                                      For GPU's this feature doesn't seem to be generally discussed.

                                                      However I have seen a reverse engineering report on Nvidia cards which discussed this type of thermal trip.

                                                       

                                                      In addition I have read in the past that AMD ported the same tech from their CPU to GPU line. I cannot find the article in my bookmarks.

                                                       

                                                      However - I did find something that might back me up. Take a look at this: AMD preps hotfix for Radeon Crimson video card fan woes | ExtremeTech

                                                      You have possibly read it, however I will quote the relevant part:

                                                       

                                                      "Some users are reporting that the problem has killed cards, but it’s not clear why this would be happening. Both AMD and Nvidia cards contain thermal trip protections that yank down a GPU’s operating speed and voltage, including an emergency shutdown feature in extreme cases".

                                                       

                                                      That's enough for one post. I will come back to your next point later.

                                                      colesdav

                                                        • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                          brucer

                                                          colesdav wrote:

                                                           

                                                           

                                                          Powertune 2.0 is used on the R9 290 and 290x for example. It is somewhat different in it aims to keep a constant fixed target temperature, the optimum temperature for the card to operate at, I believe. PowerTune 2.0 also monitors the GPU board temperatures and voltages, not just the Die, to make a more accurate estimate of the total GPU Card power. I believe you can adjust your target temperature and target fan speed and Power Limit on the R9 290.

                                                           

                                                          The r9-290 does not have a target temperature control now, which I do think it used to have..  I also see no voltage monitoring with the r9-290 either.. as a matter of fact I just installed the latest 16.4.1 package on my r9-290 system and it downclocked my gpu memory to 625mhz, and noticed there is now no target temp control, go figure. I still dont think these driver engineers know what they are doing. maybe its me expecting too much..

                                                           

                                                          I also dont think these new fury&fury-x gpus perform as advertised, I think they have falsely bloated specs and cost entirely too much for what you get..

                                                           

                                                          I still get anomalies is Flash player.. I dont know if its Mozilla, Youtube, Flash or the Fury, but the Fury gpu is the only gpu I own that does it out of an r9-290 and a hd6850.. I think its something with the hbm memory or the platform in general. Its weird, because after watching a youtube video I get all kinds of video distortion within the flash player window and sometimes my borders go completely black and I have to minimize and then maximize the screen to fix it..... Its too the point I'm tired of messing with it.. I plan on waiting until pascal comes out, I'm going to wait for the dust to settle, then I'm going green whether it costs a premium or not, I do not care.. I've seen what AMD is about now and I dont want any part of it any more... I've been a dedicated amd purchaser for well over a decade and I've seen what they were and what they have become and I am not satisfied with their products at this point... I feel they have let the gaming community down for the all mighty $..

                                                            • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                              colesdav

                                                              O.K. thanks for the reply. I  read in past reviews and feedback on  R9 290 previously from others, I do not own an R9 290.  I thought Target Temp and Target fan speed were 'introduced' because of the new features and concepts in Powertune 2.0.  Maybe 'Target Temp' has been dropped on purpose., Maybe it is a bug. I don't know,  unfortunately I have older cards which only support Powertune 1.0.

                                                               

                                                              Maybe you should report lack of Target Temperature control it as an issue or maybe it was a deliberate change noted in the Release Notes for the drivers somewhere. I will have a quick check to see if I can find out anything.

                                                               

                                                              Sorry to hear about your bad experience with the Fury.

                                                              colesdav.

                                                              • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                defectivebydesign

                                                                The 290 still has a set temperature limit of 90c for throttling purposes. Temperature control is a software feature of Overdrive that may or may not have been left out of Crimson. Dunno, because I don't use Overdrive. The feature probably got moved around to fan control or something.

                                                                 

                                                                As for PowerTune, there is a slider called power limit, which is extremely vague, and has to do with regulating clockspeeds via TDP. It's poorly documented, and gives you no detailed information on how it works, but it basically adjusts clocks by your temps. Raising it will increase your thermal limits, but afaik, you still have a hard limit of 90c. The feature is basically useless, outside of cards with triple fan coolers, and high fan speed profiles.

                                                                 

                                                                None of the 290's features have been removed. Crimson just dumbed it down to the point of being useless and uninformative. You do get individual game profiles though, so it's a feature trade off. One step forward, two steps back. That kind of thing. Agree that the UI is horrible and needs work, however you still can use MSI afterburner to bypass Crimson's stupidity.

                                                              • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                defectivebydesign

                                                                colesdav wrote:

                                                                .Disinformation? - You mean you think I am deliberately lying do you? Why would I do that?

                                                                More like incessant Red Herring arguments.

                                                                Red herring - Wikipedia, the free encyclopedia

                                                                A red herring is something that misleads or distracts from a relevant or important issue.[1] It may be either a logical fallacy or a literary device that leads readers or audiences towards a false conclusion.

                                                                According to the Oxford English Dictionary, a red herring may be intentional, or unintentional; it does not necessarily mean a conscious intent to mislead.

                                                                 

                                                                EG: You're constantly bringing up temperatures, when temperatures are not particularly relevant to the discussion.

                                                                 

                                                                We already know, and you ADMIT that the 290 will THROTTLE @ 90c. Now, how playable is BF4 running your GPU @ 200Mhz? A: It isn't.

                                                                Therefore, temperatures were not an issue for the OP, and I KNOW mine were fine, because I was monitoring them with MSI Afterburner.

                                                                 

                                                                So can we drop the Red Herring temp arguments then? Thank you. It's not related to our card failures, and it's irritating as all hell to be constantly derailed into talking about irrelevant temp trivia.

                                                                  • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                    colesdav

                                                                    Look at the original post. "Fried" "Burning".

                                                                    Seems like something got hot somewhere for some reason.

                                                                     

                                                                    Even if you monitor your temps and can actually concentrate on that whist playing a game, temps could shoot up so fast

                                                                    you wouldn't even see it because the PC immediately BSOD.

                                                                     

                                                                    I can see when Red Herrings get thrown when someone is wrong, won't admit it, and keeps chucking insults around, and won't apologise for calling someone a liar on a public forum.

                                                                     

                                                                    colesdav.

                                                                      • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                        defectivebydesign

                                                                        colesdav wrote:

                                                                         

                                                                        Look at the original post. "Fried" "Burning".

                                                                        Seems like something got hot somewhere for some reason.

                                                                        Stop cherry picking words that you took out of context. The full quote is, "The 290 produced an obvious burn smell from the power connectors." I repeat, "from the power connectors". That's a power surge, and has nothing to do with the GPU temps. Power surges burn cables, Fan failure does NOT.

                                                                         

                                                                        colesdav wrote:

                                                                         

                                                                        Even if you monitor your temps and can actually concentrate on that whist playing a game, temps could shoot up so fast

                                                                        you wouldn't even see it because the PC immediately BSOD.

                                                                        Yeah, no. That's not how it works. You don't get fried power connectors from standard overheating. Something shorted out the GPU. Maybe you should do some research into electricity before making your ASSumptions. How NOT to Make an Electric Guitar (The Hazards of Electricity) - YouTube

                                                                         

                                                                        Not to mention, you're REALLY stretching your theory when the failures include THREE separate cards. Fan failure DOES NOT work like that. It's like getting struck by lightning three times in a row. If that happens, it's not random. You can't just sit there holding onto a lightning rod and say God caused it, which is essentially what you are doing with your Temp theory. There is NO PROOF that there was any fan failure on ONE card, let alone THREE.

                                                                         

                                                                        If you are going to assume ANYTHING, then the FIRST thing I would assume is the OP's powersupply shorted out the cards. Of course, then you run into the issues of why his PSU didn't affect the mainboard, and how my 290 died when I'm using a different setup. Either way, that kind of questioning has at least SOME basic relation to what actually happened, and is not some random faith based temp theory that has no root in anything.

                                                                        colesdav wrote:

                                                                        won't apologise for calling someone a liar on a public forum.

                                                                        If you don't want to be called a liar, stop lying. I'm not going to apologize for calling out your BS. The solution here is to stop pathologically lying, instead of blaming people who point out your logic is flawed. Your logic isn't any less flawed whether or not I say so, because BS is still BS. Ever hear of the Emperor's New Clothes? That's you. Drop the delusional leap of faith thinking, and develop some actual logic if you want a respectful conversation.

                                                                         

                                                                        PS. IN CASE YOU DIDN'T KNOW.. MSI AFTERBURNER CAN LOG YOUR STATS, AND THE HARDWARE MONITOR WINDOW IS A GRAPH. (Which I monitor on my second screen.)

                                                                        That outright invalidates your "sudden" overheating theory, which in and of itself is insane and not possible in reality. Drop it. Short amounts of overheating will NOT kill a card, throttling will kick in, and the game will stop working LONG BEFORE any damage happens. As such, there is NO PROOF any overheating happened here. All existing signs point to a power regulation failure of some kind.

                                                                      • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                        ray_m

                                                                        May I request all the participants on this thread to please calm down and try to maintain a respectful tone ?

                                                                         

                                                                        Colesdav is just a member trying to help,  he does not deserve the vitriol he is receiving.

                                                                    • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                      colesdav

                                                                      RE: I doubt it was flashing the BIOS that fixed your issue, but the fact that flashing the BIOS could reset the CMOS. Resetting the CMOS is fine and dandy, and will occasionally fix problems, but please do not mix the two. If you had an actual problem with your BIOS, your motherboard wouldn't have posted in the first place. It was resetting the CMOS that fixed it.

                                                                       

                                                                      My motherboard has a CMOS_RESET button. Believe me, I tried pressing it many times. I did not want to flash the BIOS. The Motherboard would not POST with the GPUS in place, after crashing. It booted up to a point where the VGA_LED lit, then attempted reboot, continuously. Pressing CMOS_RESET did nothing to help other than reboot. I removed the cards. My motherboard POST completed. I put the cards back in, reseating multiple times. Same thing, loop with VGA error. BIOS Flashback fixed the issue. POST fine immediately after running the Flashback, with no adjustment to the GPU's at all, Thats it.   

                                                                    • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                                      latene

                                                                      Thanks for your input so far. May I ask could you provide a short bullet list of the most obvious methods for trouble shooting? I'm thinking of resetting the CMOS to try and get my 7870 working again (I'd rather not even attempt anything with the smoked VisionTek 290)

                                                                       

                                                                      PS: The VisionTek r9's appear to be infamous for GPU failure so I'm perhaps still open to AMD but I will be very attentive to warrant policy and board maker credentials this time around.

                                                                      ^ Reproduced so colsedav may receive msg indication ^

                                                            • Re: Two GPU's fried 1. VisionTek r9 290 2. Diamond 7870 GHZ Edition .. both out of warranty, suspect Crimson driver bug, devastating.
                                                              hardcoregames™

                                                              There are no issues with my cards using any of the recent drivers. I think it could be bad luck that a card has failed.

                                                              I use 2 AMD cards, HD 5450 and HD 6970 and both work fine