Came home to the fair scent of magic smoke...

Status
Not open for further replies.
SilverSliver

SilverSliver

[H]F Junkie
Joined
Feb 23, 2004
Messages
12,526
Noticed my card had stopped mining. 780Ti, undervolted, ~75C for reference. Computer is off. That's weird, usually if I have instability while mining and I need to drop clocks, the system reboots.

Hit the button - * click *

Nothing. Check the mobo, light is on and I begin to notice the distinct smell of magic computer smoke. I light up the board to check anything obvious, power connectors, etc. No scorches. I pull out the video card and it reeks. Backside has no scorch marks but I did not pull the HSF. Computer boots without the card, will not boot with in any slot on the mobo.

Ugh.
 
Last edited:
Don't quote me on this but if the card is completely fried you'll probably be fine. Doubt they will take the GPU off the board and put it on another board to check bios version. Unless there is another way they can check the fried card ?
 
So you mod a card, it dies under heavy workload, and you want a new one from the manufacturer?

In on one.
 
schizrade said:
So you mod a card, it dies under heavy workload, and you want a new one from the manufacturer?

In on one.
Click to expand...

If by heavy workload you mean undervolted for safety and temperature, then yes. If I had been pushing a bunch of volts and an overclock through it, that is one thing. Appreciate the high horsing though, find another thread for it.
 
Last edited:
SilverSliver said:
If by heavy workload you mean undervolted for safety and temperature, then yes.
Click to expand...
Depending on how the voltage regulators are set up on that card, running below-spec voltage can actually be detrimental to component longevity.

For instance, if the card uses variable resistors to provide fine-grained control over voltage, running voltages between "ideal" steps would cause the resistors to run hotter than intended.
 
SilverSliver said:
If by heavy workload you mean undervolted for safety and temperature, then yes. If I had been pushing a bunch of volts and an overclock through it, that is one thing. Appreciate the high horsing though, find another thread for it.
Click to expand...

Running a card at 100% is still running your card at 100%. That's like saying "I drive my car at max speed all the time, but it's not redlining so I don't see a problem." The mileage is still adding up and you are still wearing it down.
 
lloose said:
Running a card at 100% is still running your card at 100%. That's like saying "I drive my car at max speed all the time, but it's not redlining so I don't see a problem." The mileage is still adding up and you are still wearing it down.
Click to expand...

well then where do you draw the line? running 100% for only an hour? i haven't read the warranty agreement on my gpu but i'd be miffed if they stipulate something along the lines of "warranty is void if card is fully utilized for more than X hours". i buy gpu's to use them.

edit: this is ignoring the other variables in the OP's case like bios flash etc.
 
Last edited:
SilverSliver said:
Slightly concerned that I had an aftermarket bios on the card that I am not unable to flash back to stock since there is a short on the board.
Click to expand...

Thats the price you pay... I've been there ....chalk it up to bad luck and buy another card
 
Porter_ said:
well then where do you draw the line? running 100% for only an hour? i haven't read the warranty agreement on my gpu but i'd be miffed if they stipulate something along the lines of "warranty is void if card is fully utilized for more than X hours". i buy gpu's to use them.

edit: this is ignoring the other variables in the OP's case like bios flash etc.
Click to expand...

Considering the GK110 was designed originally for intense parallel computing tasks -- ie, constant 100% usage - it was quite literally designed for this purpose.
 
Porter_ said:
well then where do you draw the line? running 100% for only an hour? i haven't read the warranty agreement on my gpu but i'd be miffed if they stipulate something along the lines of "warranty is void if card is fully utilized for more than X hours". i buy gpu's to use them.

edit: this is ignoring the other variables in the OP's case like bios flash etc.
Click to expand...

Ignoring the other variables, you are fine and are within the warranty. I don't even think flashing the bios should be something that breaks warranty. I think the problem falls where "regular use" is and where the line is drawn. Until bitcoin became mainstream, regular use was not running cards a 100% for very extended periods of times.
 
lloose said:
Ignoring the other variables, you are fine and are within the warranty. I don't even think flashing the bios should be something that breaks warranty. I think the problem falls where "regular use" is and where the line is drawn. Until bitcoin became mainstream, regular use was not running cards a 100% for very extended periods of times.
Click to expand...

EVGA warranty as an example. It mentions all of the above. Running 100% is not the issue. Running modded bios and power mods is. That is what fried the card.

Seriously, OP just buy a new one. We have all been there, don't defraud that company that makes the card. I have fried numerous expensive items over the years.

Product condition:
•Removal and or Defacing of Serial/Part number sticker(s) on ANY EVGA products WILL void ALL warranties.

•All products must be returned in the originally sold condition. Products received by EVGA for replacement that include 3rd-party attachments (CPU Heatsink Backplate, Memory chip heatsinks, etc) will be returned to the address on the package.

•There must be no physical damage to any component including the PCB, GPU/chipset, CPU Socket, or damage that is caused by: ◦inadequate or improper repairs carried out by any person or entity which is not authorized by the manufacturer to perform warranty services on its behalf;

◦negligence, accidents, modifications, defective installation or misuse and use of inappropriate spare parts;

◦fire, immersion in liquid, lightning, earthquake, inadequate ventilation, Acts of God, incorrect application of main voltage howsoever arising, or any other cause beyond the control of EVGA;

◦any other goods which can be used with the Product, but are not the Product itself;

◦consumable or wearing items requiring replacement as part of normal service, including (but not limited to) dry cells and rechargeable batteries.

•The product must be returned to EVGA in the original factory configuration and condition. All aftermarket modification must be reversed before sending in the product for replacement.
Click to expand...
 
SilverSliver said:
Considering the GK110 was designed originally for intense parallel computing tasks -- ie, constant 100% usage - it was quite literally designed for this purpose.
Click to expand...

The workstation cards probably have better components to deal with 100% load 24/7. Also I thought you ran the card over clocked ? You can always try to rma it. What did the nodded bios give you?

Try to RMA it. What's the worst that can happen they say no? I also doubt there is a short anywhere. Probably a failed component.
 
EVGA said to go ahead and send it in.

Digital Viper-X- said:
The workstation cards probably have better components to deal with 100% load 24/7. Also I thought you ran the card over clocked ? You can always try to rma it. What did the nodded bios give you?

Try to RMA it. What's the worst that can happen they say no? I also doubt there is a short anywhere. Probably a failed component.
Click to expand...

The workstation cards have the same coolers the Titan and 780Ti have. The ACX is a beefier cooler. I run overclocked during games, I undervolt/clock for mining.
 
LoL at the GPU police xD must have shares.

Good luck with your RMA, hope you get it.
 
Vrm's must have popped. Luckily for you evga will probably cover you.

Nvidia cards tend to lack vrm temp monitoring, so I would look into buying a laser temperature gun.
 
schizrade said:
EVGA warranty as an example. It mentions all of the above. Running 100% is not the issue. Running modded bios and power mods is. That is what fried the card.
Click to expand...

If EVGA didn't want you swapping out the bios they sure as hell wouldn't put DUAL BIOS cards on the market. 0_o

And on that note I've seen a post (I'll search for it) with a member asking an EVGA rep about bios/warranty information. The rep said swapping the bios is fine as long as the original bios is loaded onto the card when it is sent in for an RMA. He then went on to state that if the card is bricked while it has a bios that isn't stock... That EVGA has no way to check which bios the card is using and will be approved for an RMA (as long as it's not physical damage of course).
 
SilverSliver said:
EVGA said to go ahead and send it in.



The workstation cards have the same coolers the Titan and 780Ti have. The ACX is a beefier cooler. I run overclocked during games, I undervolt/clock for mining.
Click to expand...

By better components I don't mean the heatsink /fan I mean vrms etc... =] great that they accepted it. Probably a cheap fix for them
 
VRMs are cooked

cQUkwQe.jpg
 
I guess the core cooling for the ACX is great, but the VRM cooling is not too great. And by not too great, I mean the VRMs are not even actively cooled by the HS.
 
Just don't make a habit out of frying cards and RMA'ing them please. If you fry another one do the right thing and pay for its replacement.
 
Since underclocking came up may I ask, could running my 7870 using the graphics overdrive in CCC options at the lowest clock speeds and at power set to -20% hurt my card when watching streaming video like netflix or youtube? I ask because it does barely show some GPU activity and heats up a little bit, but at 350 on the core and 150 on the memory with the fan at 37-40% it stays very cool. I also ask because of what was mentioned about potentially harming the videocard by underclocking/undervolting? I also use those settings for 3 games. Braid, Magic The Gathering 2014, and Starbound. I'm sorry for going a bit off topic but I figure ask here instead of making a new thread. I've been curious for quite a while. I have never noticed any strange behavior while having the videocard underclocked. Thanks to anyone who can answer this for me.
 
SilverSliver said:
Noticed my card had stopped mining. 780Ti, undervolted, ~75C for reference. Computer is off. That's weird, usually if I have instability while mining and I need to drop clocks, the system reboots.

Hit the button - * click *

Nothing. Check the mobo, light is on and I begin to notice the distinct smell of magic computer smoke. I light up the board to check anything obvious, power connectors, etc. No scorches. I pull out the video card and it reeks. Backside has no scorch marks but I did not pull the HSF. Computer boots without the card, will not boot with in any slot on the mobo.

Ugh.

Slightly concerned that I had an aftermarket bios on the card that I am not unable to flash back to stock since there is a short on the board.
Click to expand...

Mining killed a 780ti lol. I thought people bought these to game on.

Last I knew RMA's didn't cover mining, unless of course you lie about what happened...

After re reading the thread I realized you are going to RMA it with those obviously fried VRM's. It'll be interesting to see if they cover it.

Especially after they see this thread with pics and the story lmfao.
 
Last edited:
SilverSliver said:
I guess the core cooling for the ACX is great, but the VRM cooling is not too great. And by not too great, I mean the VRMs are not even actively cooled by the HS.
Click to expand...

Thats pretty scary..it might be worth looking at the new one they send to make sure it wont happen again....I'm a little surprised they dont have some sort of cooling on them...kind of asking for trouble that way
 
So, if I'm reading this right; this card has no sensor to monitor VRM temps AND the VRMs are not actively cooled?!

Oh Wow, that's Insane!
 
SilverSliver said:
Noticed my card had stopped mining. 780Ti, undervolted, ~75C for reference. Computer is off. That's weird, usually if I have instability while mining and I need to drop clocks, the system reboots.

Hit the button - * click *

Nothing. Check the mobo, light is on and I begin to notice the distinct smell of magic computer smoke. I light up the board to check anything obvious, power connectors, etc. No scorches. I pull out the video card and it reeks. Backside has no scorch marks but I did not pull the HSF. Computer boots without the card, will not boot with in any slot on the mobo.

Ugh.

Slightly concerned that I had an aftermarket bios on the card that I am not unable to flash back to stock since there is a short on the board.
Click to expand...

If any part of the PCB is fried then the BIOS will not be accessible. That said why are you mining with a 780ti?
 
jmilcher said:
Mining killed a 780ti lol. I thought people bought these to game on.

Last I knew RMA's didn't cover mining, unless of course you lie about what happened...

After re reading the thread I realized you are going to RMA it with those obviously fried VRM's. It'll be interesting to see if they cover it.

Especially after they see this thread with pics and the story lmfao.
Click to expand...

I have nothing to hide. I explained everything to them. There are dozens of posts on the EVGA forum about modified bios, 'secret' EVGA bios that you get directly from EVGA, heck - they have an entire forum dedicated to Cryptocurrency and an active community there tinkering with making mining viable on Cuda. 100% load for an extended period of time coupled with ZERO VRM cooling obviously killed the card. I won't mine with the second one, as I will likely not keep it.

naticus said:
If any part of the PCB is fried then the BIOS will not be accessible. That said why are you mining with a 780ti?
Click to expand...

740 KH/s on a single card was why. Was decent hash rate for when I was not gaming on it. Mined $360 worth of Dogecoins in a week at present value.
 
Reminds me of the GTX 295's or GTX 570's all over again. I'd be very hesitant running nvidia parts under mining conditions until they improve their track record. That said, running the stock BIOS and screwing with voltage on cards that aren't designed for it are a big first troubleshooting step.
 
Mr. K6 said:
Reminds me of the GTX 295's or GTX 570's all over again. I'd be very hesitant running nvidia parts under mining conditions until they improve their track record. That said, running the stock BIOS and screwing with voltage on cards that aren't designed for it are a big first troubleshooting step.
Click to expand...

I wont do it again obviously. If I had been at the higher stock voltage, I imagine the card would have gone sooner. Going to close the thread as I was trying to be informative before others went down the same road.
 
Status
Not open for further replies.
Back
Top