Bitcoin Forum

So this happened just now, I saw a flame near the graphics card for about a second or two (running 2 x 280x and 1 7950) and the rig shutdown. I have a 1250 W PSU. I tried smelling the wired for burnt cables but could not find any smell. However the burnt smell was on one of the 280x cards and subsided soon thereafter. This had happened once earlier too, but I thought that I was seeing things and a flame I did not expect. Cards were running near about 80 C temperature. And it would run fine before but I had a burnt cable problem on the PSU and am using different port now for the graphocs card supply

What do I test how do I go about this ? Should I continue to mine expecting that if I stop and restart it every 12 hours or so I will "get away" with this or should I completely stop ? Kindly advice.

I am not too technical a person but can look up things online.

Need urgent help.

UPDATE : Tried with the remaining (280x and 7950) cards after removing card from which burnt smell came for a bit. And seems to be running fine for now. So I am assuming it wasn't anything with mobo or psu perhaps ? But still no idea what to do with the other card (from which burnt smell came) and how to test it ?

STOP! Shutdown the PC NOW

I did stop it then and removed the cards. Checked which one was smelling of carbon and plugged it back the other 2 which werent. They seem to be mining fine for now. What do you think is the problem ? Any idea ?

I lost a 6950 exactly like that due to poor ventilation of the VRM area

use one card at a time.

First of all, lower your target temperature from 80ºC to 70ºC. Set also the fans to at least 60 - 70%

On the miner config
"gpu-fan" : "60-90",

"temp-cutoff" : "78",
"temp-overheat" : "74",
"temp-target" : "70",


Load GPU-z and select the "Sensors" tab. Scroll down to see the VRM temperature. Then, load you miner software with intensity = 8 and start mining. Raise intensity and check those temperatures!

Are you undervolting the cards? If not, do so with VBE7

Actually I have all 3 cards running now. I was earlier overclocking the 7950 and that seemed to make an impact on the other 280xs hashrate as well. Weird.

I have not undervolted any of the cards. Now the cards are running at about 80% +- 5 fanspeed. And this time I have not overclocked the 7950.

I have GPU-z loaded now and had it earlier too. For one of the 280x it shows VDCC from 1.056 V to 1.044 V.

About the VRM I can not find that in GPU-z under sensors. All I see is voltages and fan speed, etc

Hashrates now are :

280x - 710 khps earlier was 750 khps (when 7950 was overclocked to E - 1020 M - 1420)
280x - 690 khps earlier was 750 khps (when 7950 was overclocked to E - 1020 M - 1420)
7950 - 560 khps earlier was 650 khps (when 7950 was overclocked to E - 1020 M - 1420)

So they are not running at full speed. About the temperature : They have always run @75 + ... right now at about 78-79 ish and I always have a fan blowing air on them.

Since my cards are not undervolted .. what do you mean by VBE7 ?

Oh ok got the VBE7 - program used to undervolt... I have sapphire trix on my system .. will that do ? and how much should I  set VDDC to for undevolting 7950 and 280x ?

Is there any way to know a GPU is faulty/burnt besides smelling it ?

What brand of power supply are you using?

Silverstone

Often the software based do not work. You will have to follow these instructions: http://www.techpowerup.com/forums/threads/vbe7-vbios-editor-for-radeon-hd-7000-series-cards.189089/

Start with VDDC at 1050. You can go lower depending on the quality of the core and voltage regulators

So, you insist on running 3 non-undervolted cards side by side without knowing the temperature of the voltage regulators??? I asked you to STOP and follow my advice. The intent of that rushed post was clear. Yet you ignored it... Good luck, I guess

It happened before too and then worked fine for another day. So I thought it would be ok to run it without overclocking. I value your advice but its just that I have made too much of an investment in these cards and I want to run them to get the return as much as possible.

I want to know a way to get to the root of the problem. How do you suggest I find out the VRM , GPUz does not show it ? And does not cgminer show the temperature correct ?

Please correct me if I am wrong on anything. Dont mean no disrecpect bro.

Also when you say run 3 non-undervolted cards side by side - are you implying something to do with PSU ? because its a 1250 Watt PSU. Or is there some other rationale ?

Won't that void my warranty ? Because if something happens I want to be able to claim the card back atleat .. they are only about a month or two old.

Also I have never flashed BIOS before :-(, if something goes wrong I would not know much to undo it.

Oh, I see. Sorry about that...

Ok, skip the bios flashing but it's just that you're pushing the cards to the limits.

Do you have good ventilation? Do you have PCI-e risers so that the cards are spaced? Are the fans running at 70%?

Use at least a large fan (those household ones) blowing at the back of the cards and get PCI-e risers

Yes I have a large fan running on them .. so far its been running smoothly, I think the problem arises either when I overclock the 7950 or if its been running for like say more than 12 hours.

I do not have riser cables. I have had the cards running with the same space between them for about a month now. And these are lower temperatures of 78-80 . There have been days when the cards have touched 90 C due to accumulation of dust. So as of now I don't think they are being maxed out. Also without overclocking 7950 I am getting lesser hashrates right now compared to when overclocking (see above for hashrates).

Also the fans are running at a fairly high % :
7950 - 67% fan
280x - 70%
280x other - 80%

So I thinnk the fan is doing its job.

Do you think getting riser cables might solve the job ? Also I am worried about the burnt carbon smell from one of the 280x I mentioned earlier. How do I diagnose if its ok or not (because apparently its been mining fine now)?

One weird thing I noticed was when overclocking my 7950 , it used to affect both the 280xs ... dont know if that counts or is normal .. but I have failed to find an explanation for that ... seems weird to me how overcloking one can afect hashrate of other cards.

Up to 90ºC on the core and unknown at the VRMs? No risers, no undervolt, no space between cards, accumulating dust?

Understand: you must do _something_ to fix that problem. Not saying that it's ok now, and hope. You're degrading those cards fairly quickly.

Being "maxed out" or not is barely relevant. That "maxed out" is for casual use of the devices as intended (e.g. a few hours a day with 3d gaming loads), not 24/7 mining with the conditions above.

How do I calculate/finout the VRMs ?

I believe 750 for 280x is normal without needing to be undevolted . is it not ?

for 7950 yes I did overclock it to get 650 but that too would be taken care of by the psu which has enough power isnt it ?

The cards have some space between them and with the fan running constantly they are about 80 C for the past few days . now without overclocking it is at about 77-78 currently. The dust part was normal house dust that had accumulated (living tropically).

About these cards 24/7 .. people do it for months together without any problems isn't it ? So mine is just a month old .

And how is one cards overclocking affecting others ? I fail to understand

Find out the temps on your VRMs. 

How do I do that ? I have asked the same above. Kindly help. GPU-z is not showing any info on vrm

Perhaps MSI Afterburner or ASUS GpuTweak will show.

Could you please post a screenshot while mining, of the miner and Gpu-z?

Cant find VRM on MSI after burner either .... here is the image below while mining. Its been about 24 hours and the rig had a break of 1 hour in between and has been running fine without any overclocking.

Are you running the newest ATI drivers?  It should say R9 200 series in GPUZ, then maybe VRM temps will show up on the bottom row.

You can also use HWinfo

Yes the  drivers are the latest Catalyst drivers that came with the cards. The GPU-z is for the 7950 .. below it on the bottom there are two dropdowns for 280x's .. and all three don't show any VRM field.

Did you find the source of the fire? I recently had one on my 7950 XFX DoubleD catch fire on startup (power on) and can't really figure out what happened, my temps were 80-81c which doesn't seem too bad.

Fires are caused by dust balls building up in the PSU, CPU and GPU heatsink and fans. Buy a can of compressed air and clean them out once a month. Put a smoke alarm near your equipment and never place equipment where fire can spread, near curtains etc.

Hi,

No I could not find the source of the fire. Can't seem to figure out how to go about doing it :(. Although I am not overclocking now and once every 8-12 hours or so I am shutting it doewn fro about 15 mins or so and draining out power by taking out the plug. It seems to be working fine now but I am getting about 15% less hashrate :(. But ill take that anyday compared to not running one of my cards.

I am not sure if it was because of the dustballs near the GPU (because I saw the flame from the GPU section only), reason being : it happened twice in 2 days, so I am not sure it was because of the dustballs.

Although your theory seems very plausible and it may very well be because of the dust balls, reason being: the fire has not recurred since I have not overclocked and the rig is running at close to 77 C, i e , 3-5 C less than it was when overclocked. Maybe that might be the critical temperature for dustballs to ignite, who knows lol.

Anyways taking the advice of cleaning with compressed air seems good eitherways and wont hurt, it has been on my mind for the past 5-10 days, but now you saying it has pushed me to go ahead and do it, so thanks buddy for shaking off my laziness :)

What brand of motherboard and vid cards are you using? Asrock motherboards are KNOWN to cause fire when used with multiple vid cards for mining. Lesser brand vid cards have poor heat ventilation designs.

You mentioned problem with your silverstone psu: replace the psu with a quality one (corsair) ASAP. Additionally, if your motherboard is ASROCK, replace it with a quality one (Asus, Gigabyte, MSI, etc...) ASAP.

Your problem is HARDWARE, NOT SOFTWARE: no amount of fiddling in software will fix it. FIX your hardware and then tweak your software.

Also one thing I have noticed is that the VDDC on GPU-z fluctuates + 0.006 V and very rarely by + 0.012 V with avg voltage being around  :

1.087 V  for 7950
1.019 V for 280x no 1
1.050 for 280x no 2 (gives more hashes see in pic)

not sure if his is relevant.

Mobo : MSI Z87-GD65 ... i think its a good one .. doubt whether the problem is due to that.

About silverstone PSU : isnt that a well known brand ? Anyways have to take it for repair/replacement because of 2 burnt ports (but that I think was more due to the connection being loose). But i am thinking of running it for now as long as I can .

Trust me it's more than a theory. I've worked in server rooms most of my life. Seen many fires. They're all caused by idiots overloading cables or poor maintenance (dust build up). I've seen power supplies shooting flames like flame throwers several feet out the fan exhaust!  :D

Hmm im not doubting your honesty or intention. I appreciate all your help buddy, thanks for that. But however if it was of dustballs then in that case 2 questions arise :
1) Why did it not happen again without overclocking ? possible reason: it may be that that the dust bulls got burnt in the first 2 instances or flames perhaps.
2) Why did it not happen again without cleaning it ? possible reason : maybe it wasn't because of the dust balls in the first place perhaps.

So thats why I am calling it a "theory" for now, until further evidence either way. But sure I got to give it to you - its the most "plausible theory". And to test that I will wait another 24 hours and then clean it with compressed air perhaps, cause I have a feeling it wont give flames. Lets see again thats just my "intuition/theory".

The fire I had recently happened from off-state to when I turned it on (although it shouldn't have been off, it seemed to have crashed or powered down and not due to a power failure). Just 5 weeks old, definitely no dust and no pets/hair, with open rig and external fan. I'm almost certain its not a dust issue as it was really new. The rig I have was running at 82c and was 2 XFX 7950's Double D's, from what I've read they aren't a good miners card, but they should not be catching fire. I've stopped overclocking and brought the temp down to 79c and will be RMA'ing the one that caught on fire as that thing is not going near my PC ever again!

Even mine is about 5 weeks old or so and cards are sapphire brand, so I am also starting to doubt whether its because of dust balls.

About you RMAing it : don't you need to pinpoint some fault in the card as a reason to do it ? I mean is your card actually burnt .. can you see it ? cause I can't see it on mine (I have just seen superficially though and could not see any carbon marks).

I want to know how I can pinpoint the fault in the card(s) so that atleast when I take it to the shop to return/replace I should be able to tell them what the fault is.

So still no definitive answers yet.

280x will blow out caps sometimes, it will probably still mine fine after (its kinda scary/exciting though)

+1 for scary/exciting ..thats for sure :)

How do I check if the capacitor is blown though ? any idea/simple way ? I'm not that much of a techie.

I would love to know the fault as well but there's no way I'm trying to plug this back in again, its going back for sure as a full out fire started in the card. There's visible damage on the actual card of where the flames burst out from. Really scary situation and I'm grateful it happened when I was home.

Well it just happened to shutdown and restart again ... don't know if there were flames or not since I wasn't looking towards it this time .. I will leave it running through the night and if I find it has shutdown and restarted again then I will stop all mining I guess or try it with 2 cards at a time.

Now its getting more scary than exciting :(

I'm going through the same thing, I now have 2 fire extinguishers and 2 smoke detectors next to my rigs. Doesn't bring comfort when I'm not home though. I've been fortunate in that the first time I just had a spark and second time I had a small fire but in both cases it was when turning on the rig. So if I come home and find the rig off and it wasn't a power failure, my heart just races at trying to turn it on and hope it doesn't blow up. Not what I had hoped for when I got into mining. It does seem that the mobo or PSU does power down the rig when there's an electrical fault however, I just have no idea how to fix it other than turning it on and hoping for the best.

Shutdown this time after 12  hours and did not restart .. now will try without a 280x and see if it shutsdown .. wish there was a better way to diagnose .. this is getting too risky now .. since I do not know even which card it is.

If you too happen to find a solution let me know .. still desperately looking for a solution .. dont want my Mobo to get fried.

The sparks / fires are usually caused by ripple capacitors for the VRAM exploding on startup :] I've had 2 Gigabyte 7950's do this so far, however, they both still work fine.

It's always the same... the rig shuts down without warning, then when you turn it back on... BOOM...

The good news is I never had an issue (so far) with rigs exploding while running.

I've had it happen on both powered and unpowered risers... usually the issue is just poor PCB layout in that not enough cooling air gets to this part of the board.

I've also destroyed 1 Mobo when this occurred and the other Mobo lost a PCI-E slot so be careful...

Ok one more spark and now I can see that a 280x has a capacitor blown .. I can see the carbon !!

Now the problem is that the other 2 cards (one 7950 and one 280x are plugged in ) .. both are showing in GPU-z :
GPU Core Clock : 300 MHz
GPU Memory Clock : 150 MHz

This is clearly not what it was earlier. However the sapphiretrixx shows proper clocks for memory and core clock readings that had come with the cards.

Are the cards dead ? Should I still try to mine with the cards ? Why the discrepancy between GPU-z and sapphiretrix ? Or is the PCI-e slot gone ?

Now I'm really scared about the cards , don't have the courage to even try mining with the remaining 2 cards .. what should I do ? All cards and mobo are under warranty.


Please Please Please help !!

Ok so since I am almost going crazy as to what is happening .. I tried to see if the 7950 card works with mining .. turns out that the "sensors" tab in GPU-z was showing a lesser reading when not mining for GPU core and GPU memory clocks ... when I started mining with the 7950 it showed proper reading of 925 MHz and 1250 Mhz which is default and now after stopping mining it still shows the same.

Is this normal ?

TLDR : Is GPU-z "sensors" value for GPU core and GPU memory same as that specified for the card when the computer starts compared to when its mining ? cause mine is way different

Did you ever get to the bottom of this @heretolearn ? I have the same problem but I'm running 3x Sapphire 280x. I believe the advice about VBE7 is the best - cards with older BIOS installed don't seem to have the same problems, as they run around 10C cooler than the same hardware with newer BIOS installed. Basically, any undervolt needs to come from the BIOS and not an application like trixx, or the cards throw a fit.

The other possibility is that the motherboard isn't delivering the right amount of power to the cards all of the time, MSI is not an expensive brand, they are a relatively young company in terms of their competition. I have been having the same issue as you, and my Asrock board is definitely a possible factor.

Does this ring true with anyone else lurking here?

I am. I also had an spark coming out from one GPU. The riser burned. Righ now I'm planning to change all the powered risers, since the molex once in a while disconnects by itself.
I tried to direct connect he molex but the rig continues to shutdown. I believe that is the molex that isn't passing any power.

About the GPU despite the spark nothing changes. Everything seems to be working ok.
Running 4x Sapphire 280x @ 1090mhz engine, 1.175v bios