7870 Hawk was most likely exploding

[Ivy]

...at stock (manufacturer OC) clock @1100/1200, no changes to the clocks made.

But why it was so hard to find out i will explain:

First, when i was finish with the build i was starting up the machine and everything worked proper so far.

There was only a few minor flaws such as:
1. Picture not centered. Not the first time it happens on a 7000 series, i dunno why, but i could fix that issue using a AV receiver between PC and TV, so the Receiver can fix it for me. Or i simply wait for the driver to kick in, and the driver will fix it too.

2. The fan was at full speed from its sound as long as the driver didnt start to work. But it worked properly when driver was enabled (short time after start).

However, for some reason the Windows Index only showed a 7.8 for GPU rating, even when driver enabled. So i found it weird because even a 4000 series could become such a rating and any GPU stronger than that will hit 7.9. Another thing which seemed weird: The GPU Z tool had a very hard time to readout the GPU, there was close to no stats when comparing to my other 7000.

In order to find out the truth i used Unigine Heaven and wanted to bench, so i may find out if just Index screwed or a real issue. However, in the middle of the Benchmark the screen was going black and the PC completly stopped working. All of a sudden, it was going that fast, it feels like something exploded in the middle of the bench.

Troubleshooting the failure:
Now i wasnt sure what happened but PC didnt start anymore, it kinda died. First thing i got in mind is that the PSU may have stopped working because the GPU for some reason was underperforming and maybe it became to less juice. As soon as the GPU was drawing even more juice the already weakened PSU may be blasted because of a manufacturer failure. At least it was my theory... i wasnt sure of course.

Luckily i own a PSU with a self check (Corsair 1200i) and i used the self check which failed on me. So i had in mind the PSU is defective.


Some other time, i already had in mind to get a backup PSU mainwhile, i found out the hard truth:
Because i was kinda unsure if that is truly the case. Corsair and digital PSUs? Its kinda the Ferrari of hardware, its very rare having them die. So with all my doubts, i checked out the Corsair forums and i suddenly noticed a important hint: In order to self check all the cables have to be plugged out except power cable of course. I had in mind the PSU will just ignore the cables when self checking. But finally it could be different, i just didnt think about when i was in such a "rage of loss". I was not in the mood to read the whole guide, so i just checked the first words from self check entry.

So i made another self check, this time without cables attached, and it worked properly, having a green light.

Leading to the question, but why didnt the PC make a single move?
The only thing possibly happening, in term PSU is fine, seems to be a busted up motherboard, and i was kinda frozen because of that shocking insight. It would be the hardest stuff to fix, thats certain, especially because there is a huge (usualy used on many full towers) and challenging cooler attached on a very small mATX board and case. That stuff is by far most challenging to build so far. I will have to remove the 2 backplates with 16 screws in order to acess the MB freely, remove cooler (difficult to acess) and clean TIM, its simply hardest work when MB dies. Its not a usual build, its a rather taxing one and most OEM builders are trying to avoid such stuff, to much work.

However, the truth is still different:
I decided not to trust that feeling, even when the risk is way to high. Prehaps its still the GPU, even when the MB isnt moving a single finger ehm... fan anymore?

I tried reseting the CMOS, prehaps some issue there. I just noticed, every time i used the reset button, the fan light was going on for a fraction of a second. So i made it several times in a row and tried to start up after. After a certain numbers of failed tryouts, it suddenly was becoming alive. Now the PC finally was running again but there was no picture on the screen. I made several restarts and still unable to see any pictures but i was certain that all cables are attached properly. I tried to put in the cables as hard as possible... but just nothing and checked channel of TV.

Could it possibly be a MB problem?!? I decided not to trust that feeling, even when the risk is way to high. Prehaps its still the GPU, even when the MB isnt moving a single finger ehm... fan anymore?

Well, i removed the GPU and was seating my prehistoric 3870 at that spot. A GPU way to old and crazy enough for it still to run properly. After testing out and connection all the stuff, finally i see a picture. The MB was in "initial setup" mode, and i had to enter some command first. I remembered that the same command was requested at the 7870 GPU, when that GPU was inside but only at the debug LED (known as A2 issue). So i was entering the command and finally the PC was booting up properly.

In order to be sure that the MB isnt damaged, i tried to make as much tests as possible. Made over ten restarts but not a single issue anymore. Now the PC is running for over 6 hours already, post was written using it too. Also i made a Prime 95 run which suceeded nicely. Max temp at 74 C after 5 min, with only a single fan attached to the cooler (because to less space in order to use 2 fans on the Phanteks). Even the 74C is far underrated because i still miss 2 active case fans which are not working at that moment (im in emergency testing mode, i dont care fans except the most critical one).

But why the MB completly failed to work and didnt make a single move anymore when the GPU died? I think my logical answer to this is that this board got no internal GPU, not any at all. So it was saved to the CMOS and a very crucial hardware part. On the other hand the GPU was dead, so the MB simply stopped working. The GPU is that critical that the MB apparently needed several resets in order to ignore that module. On the other hand, the GPU may have been turning itself off in such a harsh way, almost like exploding, that the CMOS got some error flaw or something like that, so it needed some time to restore. Thats my theory at least, im not a experienced PC support/tech.

Besides, when it comes to RAM the MB is behaving very nicely, it feels like the MB and the RAM is a perfect match running quadruple at 1600,9,9,9,24, 1T very fast. Such issues can be excluded, there was not a single hint. Bootuptime is astounding too, 18 sec including virus scanner.


Anyway, now im pretty certain that my 7870 Hawk must have died, its almost way to clear, its close to impossible having another failure when all the other stuff is working properly now.

So my 7870 truly died and what else might have happened? Yes pretty complicated story, but thats why troubleshooting is no easy work for many PC techs, i guess.

 

[azanimefan]

But why the MB completly failed to work and didnt make a single move anymore when the GPU died?

Because there was an electrical short somewhere in the GPU when it went poof. I've seen this exact same behavior when a GPU flakes out a number of times.

recently a friend spilled coffee on his hd6670, the system did the same thing. I pulled the gpu out, cleaned the gpu with rubbing alcohol and q-tips, plugged it back in and presto! saved his gpu. Of course I was able to confirm no damage to the system because his cpu had on chip graphics i was able to use for testing, just to be sure.

simply put, an electrical short on your mb/components will pretty much kill your system till it's removed or fixed.

since your system is working, we can assume no lasting damage was done when you popped your gpu.

 

[ivanlabrie]

Your gpu died...end of story. It seems the rest of the components are fine so no need to worry that much really.
Can you replace that gpu under warranty with Msi?
Good luck!

 

[Ivy]

Oh baby, what a mess. Didnt know that even a screwed up GPU can basically kill a entire system. Im not even sure if there is some sort of protection when it comes to such a condition inside MB. As far as i know the PSU got a reliable short circuit protection but what about the MB in general?!

If i can replace it? I think so why not. It should have 2 year warranty. I dont worry the GPU to much but i worry the possible damage it could cause to the rest of the system. A GPU is replaced easely enough, its one of the easyer stuff to tackle. It would be much worse when any other part got hurt. While im waiting for new one to arrive, i can still run the system using a oldtimer GPU, its fine for a HTPC, although i lack some nice features.

But how can that happen? I never spilled any water or what else, and there was no lose cable or something else causing such a short. It was simply the follow up of a damaged internal part from the GPU it seems. Oh dear, why is my luck always so terrible. The other hardware is very valuable, a six core CPU, Samsungs newest SSD and what else, i truly would feel sad if anything of those stuff would die. But i think its hard, because the PSU is the direct supplier of current. The other parts can only be damaged through the use of the low current "data bus" of the MB, it seems. In term the GPU is affecting the PSU, the short circuit protection may possibly kick in and prevent the PC from working. Well, the CPU is supplied by MB and PSU at once, so its surely possible to burn the CPU but it may not happen instantly. I never had any Intel CPU die, if so, it can only happen as a follow up damage when another part is dying, thats the almost only way to kill a Intel CPU.

Anyway, the GPU can be RMA, no problem. I just dont know how to prevent such terrible stuff. Its so much luck based stuff dealing with hardware.

 

[azanimefan]

I never had any Intel CPU die, if so, it can only happen as a follow up damage when another part is dying, thats the almost only way to kill a Intel CPU.

no... trust me... you can kill a cpu (any cpu, even intel) any number of ways. However, as far as electrical parts go, it's not like a mechanical part, like a car engine. Just because it wasn't running right for a period of time, doesn't necessarily mean any damage was done to the system. Undervolting a system due to a ground out (like you had) probably wouldn't break anything. And if it did, you'd know right now.

trust me, if you were going to break something else with that event, it would have happened already.

 

[Ivy]

Oh dear... i worry so much. Those Intel CPUs are dirty expensive. But i can only speak from my own experience. That experience is that i didnt have a single dead CPU so far, and i already made stress testing and even small overclocking on a CPU, so im not a totaly innocent being. I never had any bent CPU pin in my whole life and i was attaching a lot of CPUs already, so i know how to handle stuff sensitive.

no. Undervolting a system due to a ground out (like you had) probably wouldn't break anything.

Oh? I already had in mind that the case feels warm in some way, even if the system isnt running at all. Kinda like there is current going down the roof, the case is full Alu without coating it may have helped. I was powering the supply off after, because i did worry.


As far as RMA of GPU goes, yes i can RMA it, but MSI could still say "oh... its busted up, that person surely did overvolt or something, our GPU will never burn out of the box". And then they will simply tell my vendor that they do not hand out RMA, thats it. I dont think it may happen like that but i cant exclude it.

Sure i know that i didnt overclock or volt or anything like that, im not that stupid. Firstly a GPU have to work foolproof, and it never worked like that. The moment when the Windows Index was invalid, i already feelt something was wrong. On top of that there is no reason, i used that GPU as deluxe HTPC GPU, with way to overpowered specs. All i want is max endurance but i got a busted up short circuit instead.

Generally i only overclock my gamer GPU a little (not that one) but i never overvolt.


I truly hope it wont be a problem to RMA. It was surely not my failure. But generally a healthy and properly cooled GPU can take lot of abuse but the Hawk simply was not healthy. I already had GPUs running over 100 C and it didnt hurt them at all. It just comes down to luck and manufacturing quality. The Hawk had such a outstanding cooler and a backplate that it may not exceed 60 C not even under full load, so heat damage is absolutly impossible. Not even overclocking damage is any real on that card, it got a way to hardcore design which could take endless abuse. My card simply was defective from a fatal manufacturing failure.

But as usual, hardware is luck, RMA is luck.

 

[Ivy]

Besides the vendor was nice and i got my money back because they said, there is no RMA possible for the Hawk, because there is nothing on stock anymore.

Finally i got me a 650ti AMP instead and seems like it was a good choice and it works fine so far.

Although there seems to be a problem with brightness level of screen or card. Because on AMD cards i have not such a issue but the Nvidia card is to bright. Its connected over an AV receiver and i used a DVI to HDMI cable because i got to less HDMI cables left. May have to get HDMI cable and it may fix that issue, but prehaps there is other solutions. The wrong level of brightness will only affect desktop menu however (applications will fix brightness, so i see it going from gray to black... as soon as application will start), so it seems to be software related.

Win 7 was trying to force me a driver adapter download. But there is no reason because i used the official valid driver. Finally i had to disable the forced driver download manually because Windows didnt seem to stop it anymore and i always got the new update notification, even when official driver already installed. So its not fully free of mess but its never hassle free to jump manufacturers i guess.

 

[EarthDog]

650ti gets crushed by the 7870 performance wise...what is finer? Regardless, glad its working for you.

 

[Ivy]

I dont truly need it for gaming, really. Its for my HTPC, but it will act as a backup gamer PC in term my main Gamer PC is out of buisness for whatever reason, so thats why i made the choice to get a rather strong HTPC-GPU. So i was totaly fine with that choice. Highest GFLOP/W ratio, a good price/performance ratio and enough of performance for its intented use. I even got some bucks left for my next gamer card, so i actually could get a upgrade even cheaper for my gamer PC. A win situation.

Dont forget, the 650Ti AMP is nearly without audible noise at all, and its size is extremely small while maintaining a very solid performance. Inside a HTPC is a astounding GPU considering its specs. It will even allow me to put inside up to 6x 3.5 inch HDD, backed up by the power of a 6 core processor at the same time the case is still a easely movable mATX size. The main problem is the noise (at 1000 RPM) of the 2 front case fans, still to much noise but i will try to either reduce theyr speed or simply exchange with some less noisy stuff. GPU and CPU fan is as good as not audible at all. So i made great work considering those coolers, but i was unable to make better choice for case, may simply need to make some modding. Target is to have around 15 dbA (1 m range) which is quiet enough to even run the stuff inside a bedroom without audible noise at all (in a range of 2-3 m). Maintaining that db even at full load is what i look out for, so performance simply isnt everything and a 7870 yes, it will perform better but will create more noise during load. For that CPU, without using a expensive Phanteks cooler, it would be impossible but thanks to that cooler the impossible became possible.

Guess biggest challenge nowadays is "denoising of PCs", not "power up". But even more challenging when doing both at the same time and small size too?!!, lot of sweat!

At the time when i got my Hawk there was no 650 Ti AMP! But surely, the exploding Hawk gave me a new opportunity.

 

[ivanlabrie]

Your htpc rig is outrageously overpowered
You should make it a main gaming rig with dual 7970's

 

[Ivy]

Dual GPU? What for? I only play at 1080P because i use a Plasma TV. I can save up those bucks, i may barely feel the difference when even a single GPU is running at 50-60 FPS.

So a 7870 is basically all the juice i need at the current time. As a backup (gamer) GPU a 650 Ti is doing fine and may still allow good settings.


To much performance? Does a overclocker or enthusiast truly know such attributes? But finally, its mainly HTPC yes, but it will serve me as a backup for many different uses (Gamer, Workstation) too, in term there is a emergency. I may not get replacement hardware on time and it may then take over that duty for up to 2-3 weeks. Just so you know what i got in mind doing it. I enjoy security a lot and everything is backed up as high as possible, so my life is with less hassle because i already got enough of hassle. The data is stored inside a NAS (from every PC), so in term a PC is exploding, i can just acess the data from the destroyed PC using the NAS and install the stuff required, so i have as less delay as possible and low hassle.

 

[ivanlabrie]

Ah, of course...
You could have gotten a cheaper htpc that would work good as a fileserver or whatever, that's all. If you oc your 3570k you get great performance and better single thread than the x79 rig, but that hexa core with unlocked multi is begging to be tortured
Congrats on your stuffs

 

[Ivy]

Combined with the Phanteks, the 6 core definitely got great potential and may easely outpace the IB and can achieve nearly the same single thread even. Although the 3570K got "only" 4 threads and in any massively multi threaded application it will fall behind by a fair amount. So its surely mainly a gamer CPU, but even capable of taking over other dutys without problems.

However, in future i may not anymore use a 6 core for HTPC, i mainly did it in order to check out the possibility inside mATX cases. So it was kinda a prototype and i had great sucess doing so because the mATX comnbined with the Phanteks can handle it just fine.

In future build however, i may swap to a 4 core Xeon CPU, and add some massive amount of RAM (Prehaps 64 GB possible in 3-4 years, even inside mATX, currently 32 GB is limit). Although, a new gen SSD (close to 100k IOPS) is that dirty fast that a RAM disk will mean nothing in most cases and uses. Its still a fun option and i want to fully enable it in order to experience its potential. Its new playing-ground for me because i never used so far.

At that time the 4 core Xeon is even more powerful than nowadays (probably +50% in 3-4 years), and it definitely can replace a 6 core while even being able to maintain a solid backup duty for the other PCs.

 

[ivanlabrie]

Yeah...You should give benchmarking a try, you got plenty of toys already.