What are some of the dangers of voltage tweaking GPUs?

[gigabytecoin]

Every tutorial I read that mentions voltage sounds like you are giving your GPU card a death sentence and to only be done by "technically experienced individuals"... blah blah blah... you will break your card. I assume this is all legal fodder. :/

So I have increased the voltage on the highest voltage range setting approximately +15% and this has allowed me to reach up to 1025 mhz on a card that typically tops out at 950 mhz with stock voltages.

The temperature is at approximately 75 degrees Celsius when I leave the stock fan settings running - or 70 Celsius when I have the fans set to 95%.

Am I doing any hard to my video card? Do I have anything to worry about? Will this action decrease it's live by multiple times?

What is the danger of setting my voltage too high? Is it simply that the card will get too hot eventually and "melt down" so to speak? If so then I should be safe at 75 degrees Celsius correct?

Can I increase the voltage from +15% to +25% so long as my temperatures stay below 85 degrees Celsius?

[hardly]

Heat does more damage than voltage. Assuming you don't run your card up to 100c+ temps you should be fine.

[gigabytecoin]

Heat does more damage than voltage. Assuming you don't run your card up to 100c+ temps you should be fine.

Yes I don't see how a card could detect a certain voltage and then freak out/burn out over it?

Higher current will definitely cause more heat, and that is the only thing in my mind that can harm the video card. Am I right?

[hardly]

It's not about the card detecting the voltage and killing itself, it's just about the core not being able to handle that much voltage. You decrease the resistance and increase current flow into anything that wasn't designed to handle that much current then you have trouble.

The card's range of voltage is well within the safe limits of what the card can handle on air cooling, though, so you should be perfectly fine.

[gemm]

Take al look at this: http://en.wikipedia.org/wiki/Electromigration

[gigabytecoin]

It's not about the card detecting the voltage and killing itself, it's just about the core not being able to handle that much voltage. You decrease the resistance and increase current flow into anything that wasn't designed to handle that much current then you have trouble.

The card's range of voltage is well within the safe limits of what the card can handle on air cooling, though, so you should be perfectly fine.

Not quite getting that last sentence?

I increased the stock voltages by 12% and am able to increase the core clock about as much and keep it stable. Is my card going to last 12% less time in general? Will it die within days?

What is a maximum recommended voltage increase for any ATI card in general? Does anybody know?

[gigabytecoin]

Take al look at this: http://en.wikipedia.org/wiki/Electromigration

And what exactly are we looking at there?

[hardly]

It's not about the card detecting the voltage and killing itself, it's just about the core not being able to handle that much voltage. You decrease the resistance and increase current flow into anything that wasn't designed to handle that much current then you have trouble.

The card's range of voltage is well within the safe limits of what the card can handle on air cooling, though, so you should be perfectly fine.

Not quite getting that last sentence?

I increased the stock voltages by 12% and am able to increase the core clock about as much and keep it stable.

What is a maximum recommended voltage increase for any ATI card in general? Does anybody know?

It's not a general number across all ATI cards, for the 5830/5850/5870 max safe voltage on air, would be around 1.3v assuming you could keep it cool.

And what I meant was that the card has a voltage range, that is defined by the bios that tells the card not to let the voltage go any higher. That programmed range was tested by the GPU manufacturer and is known to be safe voltages to run 24/7.

[Grinder]

Heat does more damage than voltage. Assuming you don't run your card up to 100c+ temps you should be fine.

Technically you're right. The problem is that what you're measuring is the heat of the silicon block, when it's really the heat in the microscopic wires that matters. If you're over volting they may be red burning hot even if you are able to cool the measured temperature of the chip to 50c.

[hardly]

This is true but, generally, if you can keep the chip, and the VRMs cool, the card shouldn't run into any trouble from overvolting.

[gigabytecoin]

Heat does more damage than voltage. Assuming you don't run your card up to 100c+ temps you should be fine.

Technically you're right. The problem is that what you're measuring is the heat of the silicon block, when it's really the heat in the microscopic wires that matters. If you're over volting they may be red burning hot even if you are able to cool the measured temperature of the chip to 50c.

Makes sense enough.

Could I benefit from a product like this perhaps? (Linked to an infared thermometer on eBay.)

My cards are "floating" in the air above their respective motherboards. I do not have them in a typical computer case. So I can easily check the temperature of the PCB.

What temperature should the back of the PCB in general be limited to?

[gigabytecoin]

It's not about the card detecting the voltage and killing itself, it's just about the core not being able to handle that much voltage. You decrease the resistance and increase current flow into anything that wasn't designed to handle that much current then you have trouble.

The card's range of voltage is well within the safe limits of what the card can handle on air cooling, though, so you should be perfectly fine.

Not quite getting that last sentence?

I increased the stock voltages by 12% and am able to increase the core clock about as much and keep it stable.

What is a maximum recommended voltage increase for any ATI card in general? Does anybody know?

It's not a general number across all ATI cards, for the 5830/5850/5870 max safe voltage on air, would be around 1.3v assuming you could keep it cool.

And what I meant was that the card has a voltage range, that is defined by the bios that tells the card not to let the voltage go any higher. That programmed range was tested by the GPU manufacturer and is known to be safe voltages to run 24/7.

Do you mean the voltage range that I can see in RBE (0.95 for the "low level", 1.063 for the "mid level" and 1.163 for the "high level")...? Or is there an actual lock on the card that will not allow it to increase over X voltage (say 1.4v for example...) no matter what I change my bios to?

Right now I have changed my BIOS to 1.4v (a 20% increase on the "high level") and can achieve ~1060mhz gpu core on my card for at least 20 minutes. The card rests at about 78 Celsius at that speed.

[bcpokey]

This is true but, generally, if you can keep the chip, and the VRMs cool, the card shouldn't run into any trouble from overvolting.

EDIT: Quoted wrong guy, ah well.

Source for allowable range of voltages?

Anyway, every chip is different, some have higher tolerance and are limited by heat, some have low tolerance for voltage even when heat is not a factor. I believe it has to do with power leakage, though I can't say for sure as I'm not a CompEngi.

[Jack of Diamonds]

Take a heatgun and point it at the VRM/power intake.

I'll give you a hint: It's about 110 celsius when you operate at stock frequencies which is well past boiling point.

It jumps up to 130-140c even with a 15% overvolt.

Yes it can be sustained for a short time, but it's simply dumb for 24/7 operation.
It makes no sense for bitcoin mining. It might make sense for short term gaming gains.

Core temp of the chip itself tells nothing of stability like pointed above (but it has a much lower thermal threshold than VRM; it will downthrottle at about 100c.

[gigabytecoin]

Take a heatgun and point it at the VRM/power intake.

I'll give you a hint: It's about 110 celsius when you operate at stock frequencies which is well past boiling point.

It jumps up to 130-140c even with a 15% overvolt.

Yes it can be sustained for a short time, but it's simply dumb for 24/7 operation.
It makes no sense for bitcoin mining. It might make sense for short term gaming gains.

Core temp of the chip itself tells nothing of stability like pointed above (but it has a much lower thermal threshold than VRM; it will downthrottle at about 100c.

What card are you using that has 110 celcius at stock freqs? And is that when bitcoin mining is going on at 100% load you mean?

[nebiki]

he's talking about the vrms which usually get as hot as the gpu, except they are not cooled as efficiently (thus they get even warmer). 110C is not unusual on stock cooling. vrm overheating is the most common reason for graphics cards to melt. especially when overclocking/not providing an airflow with fresh, cool air or using a non-reference design which does not efficiently cool the vrms. my gpu only solution on a gtx460 results in very high vrm temperatures (i didn't think the 460s would produce that much heat and a fullcover block was relatively expensive).

[actually, i'm really tired and don't know if what i said makes sense]

in other words, be careful when your gpu temperature goes beyond 90°C, as your vrms will be much hotter at that point.


another point is, especially on the newer cpus, voltage peaks which occur right after your cpu goes from idle to load. that was the reason a lot of sandy bridges died. i don't know if gpus suffer from that, too. so high voltages may damage your hardware, too.

[Jack of Diamonds]

Take a heatgun and point it at the VRM/power intake.

I'll give you a hint: It's about 110 celsius when you operate at stock frequencies which is well past boiling point.

It jumps up to 130-140c even with a 15% overvolt.

Yes it can be sustained for a short time, but it's simply dumb for 24/7 operation.
It makes no sense for bitcoin mining. It might make sense for short term gaming gains.

Core temp of the chip itself tells nothing of stability like pointed above (but it has a much lower thermal threshold than VRM; it will downthrottle at about 100c.

What card are you using that has 110 celcius at stock freqs? And is that when bitcoin mining is going on at 100% load you mean?

No GPU core can go to 110 celsius. It would downthrottle immediately much earlier.

VRM is the voltage regulation module on a GPU silicon. The temperature of the component is about 100-110 celsius even if you run the core (which shows in afterburner) at 60-70c.

The more voltage you add to the card the hotter the VRM gets.
When it gets past 145 celsius (for comparison; much past the boiling point) at 20-30% overvoltage, you are accelerating it's death. They are not specified to handle those temperatures in the long term. For game benchmarking it might be OK to run the VRM at those temperatures for half an hour.

For 24/7 bitcoin mining you are just asking for the card to die within months or even weeks.

Here is an example of a voltage bump from 0.912 to 1.025v in conjunction with a 170mhz overclock & what happens.

http://www.youtube.com/watch?v=sRo-1VFMcbc

[huayra.agera]

I can add to this, I tested my cards with aftermarket coolers, and forgot to stick the heatsinks for the VRMs, this is with a 5850. Using GPU-Z, it reached 145C for VRMs although GPU is like 48-50C (DeepCool V6000). Run well for a while eventually, my system would freeze or the mining stops and everything goes back to 2D mode stock clocks. So I think the cards themselves throttle down when reaching the danger zone. What I did, I bought my cards VRM R5 to cool the VRMs.  Temps down to 60C but with voltage of 1.25 @ 90C.

[klaaster]

.... with a 5850. Using GPU-Z, it reached 145C for VRMs although GPU is like 48-50C.

Where do you see the temperature of the VRMs in GPU-Z?
I have a Sapphire 5830 Xtreme 1G, is it maybe GPU-Temp. #2??

[MaGNeT]

I lowered the voltage on al my HD5870's with 0,025-0,050 Volt and overclocked it with 50-100Mhz and they all run cool, stable and energy efficiënt.

Electricity is 0.32$ over here so that's the main reason I like to keep the voltage down...