Bitcoin Forum

This is what happened to my 3 weeks old gigabyte today, had to put out a fire  ??? It poped while i turned on the PC, not while mining :D

http://shrani.si/f/1R/Mr/2SzdSfqy/wp20130917001.jpg

What PSU ?
powered Riser ?

Thermaltake 850w, no powered riser on this one, there is only one more 7950 currenty in the mobo so power should be sufficent.

This is very common. i have had 2 do exactly the same and know people who have had 8+ go pop.
i was hoping they had fixed this issue with the new rev 3 cards :(

Does warranty apply to this? I sure hope so :D Did this happen only to gigabyte cards or 7950 in general?

Scorchio!  Those cards do run incredibly hot in that area - I stuck a load of RAM heatsinks to the metal bar, which kept mine a bit cooler.  I also had a 120mm fan blowing on the back of the cards, and they still ran about 85C around the VRMs when mining LTC.  Terrible design.

Thankfully mine all survived long enough for me to sell them for the same price I paid for them (bought used off eBay and sold again on eBay).

I had the costum bios installed 1,069V , it didnt run that hot at all about 75C

75 ºC was your GPU temperature, which isn't where the fire started. The failing component was a VRM (voltage regulator module), which do get quite a bit hotter than the GPU.

Sadly, there's no way to monitor the VRMs' temperature in rev. 2 cards. Not that rev. 1 cards are any better. While the rev.1 does let you monitor the temps, the VRMs aren't covered by the heatsink. That's just as deadly...

This.

I had a Gigabyte WF3 7970, and the GPU Core temps were always fine, but the VRM temps were outrageous. I never liked having VRM temps in the 115C range, and had to fight to keep them below 100C.

Those VRMs overheating is a known issue with those cards. It's a shame, cuz they mine REALLY well.

I have 18 x 7950 and so far only 1 x GB WF3 has blown up, fingers crossed!!!

http://i1292.photobucket.com/albums/b577/evo7ftw/VideoCards/7950_zpsea02dbf1.jpg (http://s1292.photobucket.com/user/evo7ftw/media/VideoCards/7950_zpsea02dbf1.jpg.html)

This is a common problem with Gigabyte cards. Personally I have 3x7950 dead with the same symptoms. Also 3x7970 with fried capacitors, but they are still working fine:

http://imgur.com/a/ovHug/

You need to RMA your card, here is instructions from Gigabyte:

Dear customer,

Please use the following link to register and request for RMA.
http://ggcs.gigabyte.com/service/Login_US.aspx

Thank you


Customer Service Center
GIGABYTE TECHNOLOGY
TEL:     1-626-854-9338 Op. 4
FAX:     1-626-854-9326
EMAIL: services@gigabyteusa.com
www.gigabyte.us

My dead cards have been replaced with the new ones. Easy process.

holy shit looks like a poor design. So even if i get a new replacement card it will be a ticking time bomb :D Should probably just sell it and get a sapphire/power color

Well, yesterday my another Gigabyte 7950 died after 5 moths of life.. With sparks and smoke.. These cards are made to die fast! Sh*t

Ouch that sucks, i feel you pain...personally i will never ever again buy a gigabyte card.

Yup, I just lost another one of these.  I've lost count of how many of these cards of mine have died this way.  I think it's 4 or 5 out of 8 that I bought brand new just 3 months ago.  
All are running the F43 bios at 1.09v.  GPU temps stay in the low 70's or less.  And it happens regardless of whether I'm using a powered riser or not.

Real shame they have this problem, as they are fantastic miners.

Yo, ive had 8 go. Here is a video of one i made earlier

http://www.youtube.com/watch?v=PG2ZRT1mRWE

hello. Best suggestion for a Gigabyte 7950 that has a fan going bad? I bought it in april, and have the box and all. I was thinking of buying a replacement fan and installing myself, but haven't found one online yet.

Same story: died after 3 months litecoin mining >:( bios F43

My advice is to not run these cards above 1000 mV

That sucks, imagine what would happen if its mining and no one is around.

Haha nice one that you got it on tape :D Looks like you jumped, so did i when it happened to mine :D

Lol, yeah, I jump every time.  Even when I know it's going to happen. 

It's not a problem.  The failure happens the exact same way every time.  It goes like this:

1. Rig is mining its own business, working along nicely
2. Rig completely shuts down for no apparent reason
3. Owner sees the rig is off and goes to check it out
4. Inspect the rig, everything looks fine, confused as to why it shut off
5. Thinking everything is OK, switches the rig back on
6. BANG! 

The pop doesn't happen until someone physically turns the machine back on after the actual failure.  So there's no risk of causing a fire while nobody is around.

Most of the pops really happen when the machine is getting powered on, but it's not always the case. I had one card under frequent monitoring because of the 2 already failed capacitors, when I noticed the sudden stop of fan noise followed by the strong burnt smell. Third pop... it was time to kill it, because I didn't plan to get one more barely alive card back from the shitty RMA process. Powering the computer on caused the fourth capacitor to pop so hard that a GLOWING chunk flew at least 10 cm before landing on a wooden furniture (ouch).
Don't underestimate gigabyte, next time we may hear about one of their products just catching fire. It's time for them to get rid of the only temperature sensor to shave off even some more cents per card.

This is the exact way it happened to me :)

So you expected this? I thought you was just recording randomly :D

Did you just tape by luck or you actually know it? !!!!

This is a damn fine advice. I'm running my cards at 950

Thermaltake is a terrible, terrible brand.

This isn't true. My computer restarted itself.

There's always going to be exceptions I suppose.  Every mobo/power supply combo might behave differently.  But that does appear to be the exception.

I even have my bios set to always turn back on after a power outage.  And that works, it turns itself back on if I unplug and re-plug the power supply.  But when this failure happens, it turns off and stays off until manually switched on.

+1
Buy a Seasonic.  Or a rebranded Seasonic.

Make sure to tell them it died while playing Cysis 3 and that you never overclock  ;D.

They never asked.    8)

I own the same one. What max temperature do you guys add in CGMiner? I mean the --temp-overheat command, which makes the fans go to 100% when they hit the chosen temperature.

As I have only 1 GPU, I really don't want to fire up.

--temp-overheat won't prevent this kind of failure. CGMiner monitors the GPU temperature, not the VRM temperature.

Allright, but doesn't there is a kind of relation between those two?

I'm still wondering which --temp-overheat you guys use, because i set mine on 82. Which is just a randomly chosen number I feel confortable with, but I haven't based it on any research or logic.

82 is fine for the gpu.  But my vrm's have been blowing up when the gpu is running in the mid 70's.  I've got them running in the 60's right now for a couple weeks, and have been issue free.

The changes I've made to prevent this issue:  Reduced intensity from 20 to 18.  Reduced powertune from 20 to -13.  Reduced overclock from 1075 to 1050.  
With these changes, each card is running 10 to 15 degrees cooler, and the whole system (4 cards total) is running at 50 watts less.  It's also hurt my hashrate by about 40Kh/s per card.  But, I'm now getting much fewer rejects than before.
I've also added a heatsink on the back of the cards along the metal support for the vrm's.  Not sure how much difference that makes, but those heatsinks are getting hot, so that means they're doing some good.  I just don't know how much good.

intensity 18 was chosen because 19 had no affect at all.  18 decreased hashrate just slightly.
powertune -13 was chosen because that is the highest setting that made a difference to hashrate.  -12 and they still ran at the same speed as +20.  -13 was the first setting that actually did something.
So by using those settings, I know that they will improve power usage and temperature for the better.
gpu clock 1050 was chosen somewhat randomly.

Allright. Thanks for your message.

What do you mean with powertune? What command do you use for that? I haven't done anything with that I think.

My intensity is always max 18.

I haven't overclocked the GPU myself and it runs now on: GPU = 1000 MHZ and memoryclock (if that's the correct translation) = 1250 Mhz. I haven't done anything with the voltage by myself. Where can I find stats about this?


My GPU reaches the 81 degrees maximum often. It only hits the 82 or 83 when I'm heating the room my PC is located in. Hope this will keep my GPU from burning.

Not really, the heat sink on the VRM on these boards is pointlessly small.  I think that airflow over the back of the card is much more important than worrying about the GPU temp.

When I ran mine, I had little heat sinks stuck to the metal plate on the back, and a fast 120mm fan blowing down the back of the two cards.  Brought the VRM temps down significantly.

Do you use default or modded BIOS, because default BIOS is with VGPU 1,25V, and modded one is with VGPU 1,09V?

I haven't done anything with my card, I put it into my PC without overclocking anything. So honestly, I don't know. I guess default?

Then it should be 1,25V. You can use modded BIOS with VGPU=1,09V. You will get lower temps and lower consumption.