Bitcoin Forum

Has any of this happned to you? I went to my rig this morning, and it was off... Strange... So I hit the power button...

ZZZzzzzttttt!!! Sparks fly! I get scared and pull the plug quickly...

http://s02.imageupper.com/1_t/5/I1374247193276655_1.jpg (http://imageupper.com/i/?S0200010050011I1374247193276655) http://s02.imageupper.com/1_t/5/I1374247193276655_2.jpg (http://imageupper.com/i/?S0200010050021I1374247193276655) http://s02.imageupper.com/1_t/5/I1374247193276655_3.jpg (http://imageupper.com/i/?S0200010050031I1374247193276655) http://s02.imageupper.com/1_t/5/I1374247193276655_4.jpg (http://imageupper.com/i/?S0200010050041I1374247193276655) http://s02.imageupper.com/1_t/5/I1374247193276655_5.jpg (http://imageupper.com/i/?S0200010050051I1374247193276655)

SHIT! Glad my house didn't catch fire... Also glad I had a in store warranty... Wonder why this happened? Does gigabyte suck? Or did I do something wrong? Its been working fine for about 2 months....

Only time I have seen something like this is when someone is mining LTC.  Seems to be pushing those cards to do more memory based things versus GPU based things really is havoc on those cards.

Yea, it was mining ltc... Damn it man... So is it just this GPU or litecoin mining in general?

LTC

damn it man... so how do u mine ltc without mining ltc? my mobo, cpu, and ram was free (old pc) then i bought 3 gpus... i need 4 more months to pay it off. lol

This shit happen with my 3 Gigabyte 7950 cards (they are dead, RMA in progress) and with my 2 7970 cards (alive, still in service for ltc). Have a look to this post: https://forum.litecoin.net/index.php/topic,4886.0.html

Holy shit!
What clocks did you run em?
Seems like low quality parts...

I run my HD7850's mem@ 1500 (6000 effective), stock is 1200, and havent had any problem, I have Asus ones....

Stock clocks, not overheated. My pics is here: http://imgur.com/a/ovHug/

Those Gigabyte coolers keep the core nice and cool, but the VRM temps can reach ridiculous numbers. My 7970 was hitting 110+C before I realized it. Now I undervolt, and I never see >80C. And that was BTC mining, so I'm sure LTC pulls more A thru them, causing more heat.

Stock coolers have the VRMs making contact with the main heatsink directly, before the core. The fan blows over the heatsink, reaching the VRMs, and then the core. Keeps those VRM temps down, while increasing the core temps a little bit. On these Gigabyte ones, the fans blow the air through the heatsink, and then onto the tiny little VRM heatsink. That VRM heatsink is way too small, and it's only receiving hot air which has been used to cool the core. Bad design.

clocks werent too high. like 1100/1350... Temp varies from 60-78c.... damn it man...

I'm going to go out on a limb and say you do NOT have powered risers. An expensive mistake.

No...Never really knew about the risers till too late, and now wtf is a powered riser? please explain? :-)

If you are not using the pci-e x16 slots on your motherboard, and are using the pci-e x1 slots and running your cards off risers, there are two different types, powered, and non-powered.  This is how you  are able to see people say they are running 5 cards off one board, they are using risers. 

The biggest difference is that powered risers give more power to the bus of the cards, the powered risers have an additional molex connector attached.  This is important in ltc mining as the power to overclock your vram on your cards for ltc mining comes from the bus.  While btc power comes from the pci-e power connectors from the psu since thats what is powering the gpu overclock, thus you dont need powered risers to run them.

http://www.amazon.com/Micro-SATA-Cables-Express-Extension/dp/B00DDCT5IK
http://www.amazon.com/Micro-SATA-Cables-Express-Extension/dp/B00DDD3012

Guys, this very likely has nothing to do with powered risers, nor LTC mining (although LTC mining may make it more likely.)

I've had the exact same thing happen with a Visiontek 7970 that had been only BTC mining on a x16 PCIe slot for about a year.  Machine powered off.  When I found it and turned it back on, There was a spark from inside.  (I havent opened the reference case and cannot see any burn damage like in these pictures, but I'm sure it exists inside there somewhere.)

I attribute it to the VRMs running far too hot in the 7970 reference design.  I tried to keep mine in the 90s, but it often went into the 100s with a fairly good overclock.  Mind you, it did run 24/7 since about mid 2012.

Anyway, am about to test Visiontek warranty service.. will see how it goes, but I, like you, do attribute this to VRM temps.  7950s don't have this problem.. their VRMs are often cooler than the GPU itself.  Reference 7970s often really do have issues with VRM temps..

I tried saying this up above, but they didn't listen. :/

You can add extra heat sinks to the metal plate on the back of the card that holds the VRM heat sink on.  I've two of those 7950 Gigabytes, and I added some of the cheap eBay stick-on RAM chip heat sinks.  Seemed to help a bit. 

I've also a 140mm fan running on the side panel of the case, blowing straight over the cards. 

I ended up doing something similar almost a year ago (https://bitcointalk.org/index.php?topic=102167.msg1138807#msg1138807).

Well I personally have never mined LTC, and have always down-clocked my Gigabyte cards memory so I guess I never have had an issue.  But you are right, from a design stand point, they are very flawed.  I was merely giving him some info that may benefit him in the future.

great. thanks! you wouldn't believe how i have my risers set up... smh.. im just glad it works as is. lol.

I'll just add my 2 cents:

My Gigabyte 7970 is currently in RMA with status "repairing".  I'd been mining both bitcoin and litecoin.  I'd just started mining Litecoin and was tuning cgminer when I realized that my HW error rate was really high.

I tested the card with memtestCL and found that my card failed the 'random blocks' test in a major way (over 672,000 incorrect bits).  Took a screenshot and sent it in with my RMA request.  Gigabyte service has been great so far.

The card did overheat a few times while I was using it and after examining it when it failed there was some discoloration on the VRM heatsink rear strap suggesting extreme heat in that area.

I'll be much more careful when I get the card back.  I've already picked up risers and after reading this thread I've ordered some cheap heatsinks off ebay.

Thanks guys!

To be fair, miners are running the cards way above what they were ever designed for.  99% of people buying those cards maybe play a game an few hours a day, which isn't as hard on the VRMs as mining LTC. 

Mining LTC on a GPU is like driving your car up a very steep hill at maximum revs constantly for 24 hours.  It'll do it for a certain amount of time, then the engine will pop.

When I bought my 7970 GHz edition almost a year ago, it was $450! It was one of the best video cards you could buy at the time. I've also been careful to give it good airflow, watch the temps, and clean it regularly.

In you're car analogy, this isn't just your mom's Toyota Camry running up hill 24/7. This is a Nascar Toyota Camry. See the difference? It should be able to handle the load.

You're crazy psychotic running your cards at 90°C

Use cgwatcher..........................

to fix this problem: stop mining in the summer and go out and fuck some chicks instead

autumn/winter are for mining

I have 4 rigs and started out with 16 cards total. I lost 5 due to this exact issue and needed to RMA. The cards do have design issues as they cannot take running 24/7. All of the cards have failed in the exact same way. I am using the F43 bios which allows lower voltage for the Gigabyte cards. Also, all cards were running stock. Literally not setting any options except intensity and thread concurrency. A few months of straight mining and they kick the bucket...

wrong again use robnitro's hacked bios.... w cgwatcher.... w powered risers....

otherwise = PHAIL

Well yes this happens alot with gigabyte stuff
I have had 4 different gigabyte nvidia based videocards blow up in similar ways
They never took responsibility on all these items.
Also had a very expenssive motherboard which was failing alot
I did an rma and got the board back with the message that it was fixed but the same problems kept appearing.
They simply messaged i was wrong and never anwered any follow up mails.
When it finally completely died we found what really was wrong by accident.
Some friend wanted for his same type motherboard the cool blocks from the chipset when we took it off we found the main chip was almost completely covered in melted metal.
Hence that was the weird problems i had all the time.
The last issue i have had was a amd card which suddenly failed and guess what gigabyte also never wanted to replaced it, well not when it
was under warranty they dare to ask him to buy a new instead with 30 euro off ........

So i am staying away from gigabyte products, and keep super long distances since all this fiasco's happen.
Worst is this company never cares at all.
In my opinion they may go bancrupt tommorow

Do you have the link to this BIOS? I am using powered risers and the cards themselves never go over 80C as reported by cgminer. What features does cgwatcher provide over cgminer for temperature protection? I know that cgminer already has overheat protection. The cards have all failed while running within temperature limits. It's always the chips on the back of the card that fail. The main processors and everything under the heat sinks are fine.

got 10x 7950 Gigabyte.

3 of them with exploded VRM. STOCK CLOCKS!! even slightly underclock/volted!!

Bios F43 with 1,09v Core (i dont know memvoltage)

LTC Mining

Searching for BIOS now.... so easy to flash, don't worry. As for CGWatcher... it will automatically change the intensity to your liking for when it overheats. You specify the values. You can set timed profiles. It automatically restarts for all kinds of issues. Dead cards/No accepted shares in X time, Speed falls below XXXX MH/s, etc. It's a must have especially if you have more than one card or rig.

I WAS GETTIN' SUM BIOS, GETTIN' GETTIN' SUM BIOS! Use the FY1 BIOS if you have Rev. 2 cards. If you go to flash in DOS and you get a locked error, google it. You have to command line unlock them suckers.
http://www.overclock.net/t/1313813/new-bios-update-for-gigabyte-hd7950/720
http://www.overclock.net/attachments/13654