GTX680 dying? gpu usage/temps peak fan activates happens every 5-10minutes

[notanmmmonewbie]

Hey everybody,

I built this PC 10/2012. I'm good at following instructions so i can build and assemble a PC and shop for the right parts. But i'm no good at diagnosing and i only have a vague understanding of how the internals of a GPU work. So i dont know if this is because the graphics card is dying, because summer is coming on, or if it could be driver related. I mention drivers because about 2 months ago i did a roll back for some FO4 modding work.

http://www.newegg.com/Product/Produc...82E16811133191
http://www.newegg.com/Product/Produc...82E16814127695

Thats my graphics card and case. I have three 210mm intake fans-top, front and side and a standard exhaust fan in the rear. The MSI card has two fans of its own as well. When i play games on high graphics like FO4, B&S and BDO (even Source Filmmaker) the GPU fans will kick on every 5-10 minutes and run for about 30-60 seconds. I installed MSI afterburner when the this issue first started happening about a month and a half ago and the GPU usage goes up as well. Now the PCB of the card does have some dirt buildup and its not something that canned air will get rid of. The GPU fans have no dust after cleaning and the fins underneath the fans do have some dust but i cant get to it.

So could it be a driver issue or is the age finally starting to show? or do i need to tweak the card settings out of game?

[Miyagie]

What are the system/CPU/GPU temps ?

[notanmmmonewbie]

http://i.imgur.com/2RtIyie.jpg

That was the first time the fan went off, The second time it went off the indicators on the right showed 53* for the first one. My room is running a bit hot right now too. Do you think it could be the CPU too? i dont know if it makes a difference but i can use Maya 2016 for a few hours and the fans wont come on at all. I dont know what to make of that though cause i dont know how SFM or Maya use CPU vs GPU.

- - - Updated - - -

Looking at some CPU related maintenance stuff and FYI/FYAI i havent changed the thermal paste since i assembled the PC. And it was also the stock paste that comes on the CPU out of the factory box. I bought my own tube of thermal paste but i didnt even pop it open. If i can find that tube would it still be good to use or would it have hardened even if i didnt break open the metalic seal on the nozzle?

- - - Updated - - -

OK i found the tube of thermal compound. I bought it along with the rest of the PC parts back in 09/2012 but i never used it. Its the Dynex brand and looks like its in a capped off syringe still in the original packaging.

[Deleted]

Why do people think thermal compound gets bad as it gets older?

Depending on how much there is, you may need to get rid of the buildup of dust. This may mean disassembling your GPU, or taking off the CPU cooler fan (not the heat sink itself).

Be careful with using canned air near any fans. You may ruin their bearings by spinning them up beyond their limits.

If your PC is kept clean, then nothing will change in regards to temperatures as the years go by. CPUs and GPUs do not procude more heat as they age, and if they go bad, which is really, really rare, they will usually just flat out die on you. Boom. Dead.

Edit: If your actual GPU usage suddenly went up, then it is probably not a hardware related problem at all. Something causes your GPU to do more work, be it drivers or programs running in the background.

Did you install MSI afterburner before the problem occurred? That is an overclocking utility, and overclocking your GPU will by definition increase the power dissipated by and load shown on your GPU.

[Evildeffy]

Why do people think thermal compound gets bad as it gets older?

Depending on how much there is, you may need to get rid of the buildup of dust. This may mean disassembling your GPU, or taking off the CPU cooler fan (not the heat sink itself).

Be careful with using canned air near any fans. You may ruin their bearings by spinning them up beyond their limits.

If your PC is kept clean, then nothing will change in regards to temperatures as the years go by. CPUs and GPUs do not procude more heat as they age, and if they go bad, which is really, really rare, they will usually just flat out die on you. Boom. Dead.

Edit: If your actual GPU usage suddenly went up, then it is probably not a hardware related problem at all. Something causes your GPU to do more work, be it drivers or programs running in the background.

Did you install MSI afterburner before the problem occurred? That is an overclocking utility, and overclocking your GPU will by definition increase the power dissipated by and load shown on your GPU.

Several corrections:

It isn't that thermal paste gets worse as it ages directly it's because (depending on type) it either dries up or evaporates over time.
Thus requiring a fresh coat every so often (most of the time 4 - 5 years on normal usage).

You push temperatures more and this accelerates of course.

Regarding temperatures on the CPU they actually do rise with time as the silicon degrades thus leading to more voltage droop and resistance over time.
So as a CPU gets older it DOES increase in temperatures and does lose it's overclocking capability.

Your statement would have it that silicon does not degrade at all which means from the day it's built to the day it dies it remains identical.
This simply is not true as silicon in the sense of chips have an actual natural degradation cycle.

Just correcting the info given.

[Deleted]

Well, that is certainly news to me. Admittedly I have only been working with electronics for 35 years or so, and you always learn new things.

So you are saying we need to take all power electronics apart every few years to replace the thermal compound? Electronics like those found in power distribution, cars (including electrics), medical equipment and more, *and* additionally microchips only last a few years?

Interesting. Seems I have a lot of high paying work ahead of me. This will make the Y2K problem look like a joke by comparison.

I am aware many TIMs change consistency once heated, but just because the TIM feels crusty when you pull the HS doesn't mean it has failed. The grey stuff Intel uses partially falls into this category, as it intentionally flows when heated, allowing it to fill every nook and crevice.

[Evildeffy]

Well, that is certainly news to me. Admittedly I have only been working with electronics for 35 years or so, and you always learn new things.

So you are saying we need to take all power electronics apart every few years to replace the thermal compound? Electronics like those found in power distribution, cars (including electrics), medical equipment and more, *and* additionally microchips only last a few years?

Interesting. Seems I have a lot of high paying work ahead of me. This will make the Y2K problem look like a joke by comparison.

I am aware many TIMs change consistency once heated, but just because the TIM feels crusty when you pull the HS doesn't mean it has failed. The grey stuff Intel uses partially falls into this category, as it intentionally flows when heated, allowing it to fill every nook and crevice.

Most devices do not use TIM and a heatsink is plenty without the need for contact.
A CPU and GPU is a different matter as it continuously changes it's "phases" depending upon load.
Thus changing the properties of the TIM as it goes along (like metal fatigue at some point causes a snap).

And yes every form of electronics that uses silicon will degrade over time, this is a fact.
Barring the fact that silicon also retracts and expands and thus also cracks over time. (nVidia laptop chips from 2006 - 2010 anyone?)

I've also worked in the electric industry (Noratel) and have designed and produced in more than your just named example fields.

But this discussion doesn't help the OP.

[Deleted]

Most devices do not use TIM and a heatsink is plenty without the need for contact.

You have got to be kidding me.

A CPU and GPU is a different matter as it continuously changes it's "phases" depending upon load.

You mean completely unlike the power semiconductor devices in an electric car, or the output devices on many radio transmitters, just to mention two examples?

Thus changing the properties of the TIM as it goes along (like metal fatigue at some point causes a snap).

Never heard of this being an issue. Google only mentions PC builder concerns, and thermal expansion cycles should do nothing, considering how soft the material is. Which mechanism(s) do you propose is responsible for the degradation of the thermal conductivity? The goo is basically a slurry of very fine metal oxide particles, like BeO, suspended in a mixture of synthetic silicone oils.

And yes every form of electronics that uses silicon will degrade over time, this is a fact.

Oh yes, no argument. It just takes a lot longer than a few years, if the devices in question are run within specs. We would see absolutely insane infant mortality rates on just about every piece of electronics, if the average MTBF on a particular device was only measured in a few years.

Barring the fact that silicon also retracts and expands and thus also cracks over time. (nVidia laptop chips from 2006 - 2010 anyone?)

That was a problem with the BGA solder balls cracking, not the Silicon chip itself. A manufacturing defect, in other words.

I've also worked in the electric industry (Noratel) and have designed and produced in more than your just named example fields.

I'm an experimental physicist by education, with the work on my M.Sc. thesis in semiconductor technology, specifically the properties of impurities and defects in SiGe alloys. Please do not hesitate to pour it on. I ought to be able to take it.

But this discussion doesn't help the OP.

It helps me, though. Sounds like I am missing something important in my upbringing.

[Evildeffy]

You have got to be kidding me.

As much as it may (or may not) surprise you most are actually either soldered or welded on for heat transfer, including those I've seen for cars.

You mean completely unlike the power semiconductor devices in an electric car, or the output devices on many radio transmitters, just to mention two examples?

Not to the same extent, how many radio transmitters are you aware of (not the massive ones that are the size of a house) that have active cooling requirements and work on a CPU that is as complex as that of a PC?

Never heard of this being an issue. Google only mentions PC builder concerns, and thermal expansion cycles should do nothing, considering how soft the material is. Which mechanism(s) do you propose is responsible for the degradation of the thermal conductivity? The goo is basically a slurry of very fine metal oxide particles, like BeO, suspended in a mixture of synthetic silicone oils.

A lot of thermal pastes actually dry up or even crumble over time as the actual goo (which are oils, fats etc. every paste has a different build-up and mixture) mostly after 4 - 5 years, thermal pads generally tend to be the worst of them all (pads as in pre-applied goo, not the actual heat transfer pads that are foam-like) since they have to be solid for the application where others are a viscous liquid type.
The foam-like pads literally crumble due to the heat over time (this shit makes you wonder WHERE the fuck it comes from when you see it first time).
Others are based upon silicone, carbon or the rare few on metal .. like Arctic Silver 5 or the liquid metal one I forgot the name of.

Oh yes, no argument. It just takes a lot longer than a few years, if the devices in question are run within specs. We would see absolutely insane infant mortality rates on just about every piece of electronics, if the average MTBF on a particular device was only measured in a few years.

It depends upon usage, if a CPU/GPU isn't running 80 - 90 degrees C continuously over overclocked it lasts give or take 10 years before it starts showing defects.
This is counting proper usage and not 1 hour a day, MTBF for CPUs (like SSDs) are completely meaningless.

That was a problem with the BGA solder balls cracking, not the Silicon chip itself. A manufacturing defect, in other words.

That was one of the issues, it also cracked the die itself due to heat expansion and contraction.

I'm an experimental physicist by education, with the work on my M.Sc. thesis in semiconductor technology, specifically the properties of impurities and defects in SiGe alloys. Please do not hesitate to pour it on. I ought to be able to take it.

OK! *prepares large unnamed object* I'MA GOING IN DRY!
Ever seen a 2,3MVA transformer's cooling? Even though there's no silicone involved it requires massive heatsinks which are welded on, and we made this for all types of industries like Medical (though not that big for them) automotive, shipyards, oilrigs etc.
Though nothing of nature in full silicon involved production though.

It helps me, though. Sounds like I am missing something important in my upbringing.

We all miss some in our upbringing.. like my sanity... it has been lost to the dark reaches of the galaxy with the family I have! But it's all good though, my friend called Fred, whom apparently no-one can see but me, is always there to help me.

[notanmmmonewbie]

I made a post before, untill i realised your temperatures actually got to 85 degrees wich are kinda high for that card. Your cpu is absolutely fine, dont worry about that one.
Your card is spinning up to near 100% because the temperatures are reaching near 90 degrees as evidenced by msi afterburner wich is too much, the only way to fix this is to take the card apart and clean it for dust (maybe reapply thermal paste like the people above me said)

Your card should still be fine but some maintenance is required.

uhh yah so anyways...

I went ahead and cleaned and reapplied the thermal paste on my GPU. Even with the heat in my room it still does spike to 85-91 occasionally but it doesnt last for long, the fan is on for only 5 seconds and it starts to wind down. I wonder if its possible with MSI afterburner or some other GPU program to set my GPU fan to ramp up sooner to keep temps lower. Or if i can just set my fan to run at a constant speed like 40 or 60% at least just during the summer time.

[Deleted]

Sorry about the long delay in responding.

As much as it may (or may not) surprise you most are actually either soldered or welded on for heat transfer, including those I've seen for cars.

That doesn't surprise me, since both techniques have been used in car AC alternators since forever. However, that is not what I was getting at.

There is a lot of high power industrial grade electronics out there, where the per-unit area heat transfer is comparable or higher that what we encounter in modern CPUs. For instance I am sitting here with an old Siemens kA(?) level SCR with a M12 mounting thread, intended to be bolted down (using a torque wrench). I happen to know these were intended to be mounted with thermal grease/goo or whichever. I have never heard anyone mentioning in writing or speech that this needs to be reapplied every X years.

If this goo fails after a time, then a *lot* of factories and similar industrial installations should fail with predictable regularity. Pulling working electrics and electronics apart to replace thermal goo isn't part of any maintenance scheme I have ever heard about. I have a friend, who owns a company performing installation and maintenance of industrial electric/electronic systems, so I guess I should ask him.

Not to the same extent, how many radio transmitters are you aware of (not the massive ones that are the size of a house) that have active cooling requirements and work on a CPU that is as complex as that of a PC?

Every medium power (100W to low KW class) intermittent duty semiconductor power amplifier, specifically quite a lot of amateur (ham) radio amps, maritime mobile and aeronautics amp falls under this description. There are a *lot* of these out there, and 100W+ amps have been common even among commercially manufactured radio amateurs transceivers since the early/mid 80'ties.

The first viable 1KW push/pull linear RF amplifier design, using two RF power BJTs, was published by none other than the legendary Helge Granberg of Motorola Semiconductors fame, back in the 70'ties. His reference articles were published in Motorola's old Big Blue Bible on their RF devices, complete with every conceivable mechanical and electrical detail needed to replicate the designs. No mention ever of periodically having to replace the goo.

Today we are much higher in power from a pair of RF MOSFETs, each of which can - if memory serves - have upwards of 500W of allowable dissipation. Note this is for linear RF operation, where efficiency on a good day hovers around 60%. So Pd of this order is not just for fun. These devices can and do dissipate a lot of power in a small volume.

I have never seen a semiconductor RF amp, where the active devices are either just press fitted nor brazed to a heat spreader. The technique certainly isn't commonly used from what I have seen, if at all.

Many of the hundreds of thousands of RF amps built in the 80'ties are still with us, I have one here. Again, never heard of the requirement for periodic goo replacement on any of these.

A lot of thermal pastes actually dry up or even crumble over time as the actual goo (which are oils, fats etc. every paste has a different build-up and mixture) mostly after 4 - 5 years, thermal pads generally tend to be the worst of them all (pads as in pre-applied goo, not the actual heat transfer pads that are foam-like) since they have to be solid for the application where others are a viscous liquid type.
The foam-like pads literally crumble due to the heat over time (this shit makes you wonder WHERE the fuck it comes from when you see it first time).
Others are based upon silicone, carbon or the rare few on metal .. like Arctic Silver 5 or the liquid metal one I forgot the name of.

Yes, many thermal pads, like those horrible, blue SilPads(TM), deteriorate fairly quickly in high power/high mounting torque applications. No argument. Thus why it is wise to use ye olde Mica/Kapton plus goo for equipment, where longevity is a factor. Messy and time consuming to assemble, but it works.

As for the drying goo part, then by coincidence I looked into this in some details about a decade ago, in connection to a fairly expensive piece of IBM hardware utilizing Xeon CPUs. An IBM tech conjured up a datasheet for Intel's grey goo, and as you might have guessed, it was a Dow Corning product. I cannot say whether the grey stuff Intel uses these days is the same compound. But at least the stuff I saw the datasheet for specifically emphasized the longevity of the stuff. Absolutely no mentioning of the need to replace it at regular intervals.

The same applies to other technical Dow Corning Silicone goo products I have been near. One of these is DC-4, which is a transparent Silicone grease with rather exactly the same electrical properties as solid Teflon. One of its uses is for filling in air gaps in Teflon/metal interfaces in high power RF environments, to prevent corona abrasion of the Teflon dielectric. DC-4 *never* dries out, as it would be a disaster if it did due to its intended purpose.

All this leads me to a curious idea: Imagine for arguments sake that both Dow Corning and Intel both has all their technical ducks in a row. *If* this is the case, then maybe - just ... maybe - Intel's grey thermal goo is as good as it gets in a non-drying thermally conductive Silicone grease/compound. If you need a lower specific thermal resistivity, then you absolutely need to use a mixture, which has components that can evaporate. Once those compounds has evaporated, their place in the matrix is replaced with air, and the thermal impedance skyrockets.

So why would this specifically be a problem for enthusiast level PCs? Because people generally seem to not know about the razor blade technique, so they use way, *way* too much goo, necessitating the lower thermal resistivity. Intel's non-drying goo (which you at least used to be able to buy separately from their heat sink assemblies) wouldn't work nearly as well, *unless* you applied it properly. Which almost nobody does.

It depends upon usage, if a CPU/GPU isn't running 80 - 90 degrees C continuously over overclocked it lasts give or take 10 years before it starts showing defects.
This is counting proper usage and not 1 hour a day, MTBF for CPUs (like SSDs) are completely meaningless.

OK, this is much more reasonable. I read something entirely different into your previous posts.

That was one of the issues, it also cracked the die itself due to heat expansion and contraction.

Interesting. Electro-migration at work, or something entirely different? I couldn't easily find an online reference, so a link would be appreciated if you have it.

OK! *prepares large unnamed object* I'MA GOING IN DRY!
Ever seen a 2,3MVA transformer's cooling? Even though there's no silicone involved it requires massive heatsinks which are welded on, and we made this for all types of industries like Medical (though not that big for them) automotive, shipyards, oilrigs etc.
Though nothing of nature in full silicon involved production though.

Yes, I have seen similar transformers. They use liquid oil cooling to transfer the heat to the external welded-on steel heat sink fins, thus the literal spectacular fireworks if you somehow do manage to brutally overload them.

I may have missed your point in referencing them, though.

[Lathais]

Why do people think thermal compound gets bad as it gets older?

Because it does. It has happened to me. On some older PCs I used to have with AMD Phenom II x4 965 BEs I had running always ran fine and at great temps. After they were about 4-5 years old, I started having issues with them shutting down under heavy loads. Checked temps and CPU temps were bad, causing the system to shut down to avoid damage. Replaced Thermal Paste and problem went away and temps were once again acceptable.

Depending on what stuff you use, it does "dry out" and therefore shrink. Once it shrinks a bit it is no longer properly filling in the are it should and you lose surface area between the CPU and the heat sink causing less heat to be able to be transferred to the heat pipes. It does happen.

[Evildeffy]

Sorry about the long delay in responding.

No worries, I'm not a stickler for these types of responses which aren't required with any haste.

<Lots of info Part 1>

There are plenty of different types of paste, the ones used in the computer industry dry up after a time because they cheap this stuff out.

Example:
If you were to de-lid an Intel CPU (done after Intel wiped the floor with AMD after the Sandy Bridge series) you will find a standard grey paste inside.
Then they moved to the original Haswell CPU's which actually used a silicone base white paste in their early line and then back to some grey paste after the refresh.
(I honestly do not know which brand was used by Intel) But prior to that Intel actually soldered their CPU dies to the IHS which was far more effective in heat transfer, but if you de-lid an Intel CPU (not the <Prefix>-E series, those ARE soldered) and replace the paste with any aftermarket one you would have noticable gains in temperature decrease.
If you used Coollaboratory's liquid metal "paste" it would drop a drastic 20°C.

The paste used in today's computers, even supercomputer GPUs which only crunch numbers, are fucking horrid in performance and over time (4-5 years, normally, depending upon abuse of the devices and thus temperatures) need replacing.

There may be some industrial type paste I'm unaware of which doesn't require this but so far in 21 years in the computer business (of which paste was a rarity the first 10) I have yet to find any paste that didn't need replacing after a certain time, indicating that all pastes use a mix of materials that evaporate over time.

<Lots of info Part 2>

Looks like we have different standards of what is referred to as high, medium and low power devices.
100W+ here is considered high-power, consumer grade not industrial grade.
Most I've seen were all passively cooled with large heatsinks which none had any form of thermal paste or pads, I HAVE seen some disgusting ones which looked like they used a silicone glue to an extent that it looked like a large ball of silicone yellowish rubber.

<Lots of info Part 3>

Ah but 10 years is an insanely long time in the computer world where everything is changed, updated, etc. etc. in under 6 months time.
I couldn't tell you if they use the same stuff as prior but I can tell you that Intel's thermal paste rather horrid in what they use both in connecting the CPU die to the IHS as well as what they use on their own heatsinks.

The problem is that to make a profit they have cheapened out a lot in all areas and not just Intel either.

There is a reason why builders tend to buy stuff like the <Insert high quality brand here, I personally use Arctic Cooling MX-4> to replace the crappy paste they get with the manufacturers

OK, this is much more reasonable. I read something entirely different into your previous posts.

It depends fully upon environment and abuse, in a normal scenario where a computer is used in a house just to browse the web, read e-mail, watch youtube etc. etc. it'll take a long ass time before it needs replacing, but it WILL at some point.
But if the user is a gamer whom overclocks, has limited thermal space, is a cheapskate, lives in a country where it's 35-40°C ambient temp etc. etc. it will evaporate VERY fast and as you said when that happens you get air pockets and what are air pockets good at? Retaining heat and not transferring.
If you can think back to the P4 Prescott era the CPUs were running standard at 80°C and overheated A LOT after a year of use because the paste went bye bye.
Let's not go into the fact that they actually also damaged themselves over time adding to the whole overheating and shutting down bit.

Interesting. Electro-migration at work, or something entirely different? I couldn't easily find an online reference, so a link would be appreciated if you have it.

If I did I would've posted it last time, I actually looked for it specifically when I posted but I cannot find it anymore.
This was a technical article post dated in 2009 or 2010 so finding that is no easy task, the most common problem were cracked solder joints but there were plenty of cracked dies as well.
If I remember correctly the entire PCB where the die rested was a poor choice then because it too expanded and contracted, it's well possible the main cause of the die cracks were due to the PCB ... but it's been a while so if I knew with certainty I'd tell you but I simply do not.
All I remember is having read it and visually experienced it especially on a lot of HP Pavilion line of notebooks.

Yes, I have seen similar transformers. They use liquid oil cooling to transfer the heat to the external welded-on steel heat sink fins, thus the literal spectacular fireworks if you somehow do manage to brutally overload them.

I may have missed your point in referencing them, though.

Not all transformers use oil cooling of that since, plenty use "AN" cooling or "AF" or a couple even "SWF" but yeah .. you don't want to be near those if you fuck up the connections, create a short or have a construction fault.
I referenced them because of your sentence which I quoted, I kinda read it as an asshole-ish comment so I responded in a sense of that the world is far bigger with different solutions used than just the ones we often days work with.
What happens in country A because they find it sensible does not work in country B, think of Nikola Tesla's A/C and Thomas Edison's D/C.
They hated each other's guts but both contributed to society with their inventions but Edison is more well known rather than Tesla who's A/C devisement (is that even a word) is powering almost every single device today.

What is used in a different industry but is the same in function and form does not have to function or even be used in another industry is the moral of the story.