Need help with my PC graphics card

[dougal]

I got this graphics card, GTX 260 1.7GB RAM. I got it super cheap as the seller did not know if it works or not.

Well, i plugged it in the PCI-E slot and fed it the two PCI-E power cables and booted up.

Problem is that the screen is full of artifacts.. Really bad ones and although Win7 detects the card it won't start it up and says there is a problem with the device. The artifacts show everywhere even on the bios screens.

I am guessing the problem is hardware no ? I've tried cleaning the contacts and tightening the screws of the GPU from the back of the card but it doesnt help.

Sometimes the computer wont even boot with it.

Its a shame because my current graphics card (GT9400) is crap and this would be a nice upgrade.

Any help appreciated

 

[Harrison]

Artefacts normally mean overheating. I'm guessing the heatsink isn't correctly seated/mounted on the GPU. You will need to get some thermal compound, remove the heatsink from the graphics card, clean up the GPU, apply the thermal compound and remount the heatsink.

Also make sure the fans are spinning on the card.

 

[dougal]

Artefacts normally mean overheating. I'm guessing the heatsink isn't correctly seated/mounted on the GPU. You will need to get some thermal compound, remove the heatsink from the graphics card, clean up the GPU, apply the thermal compound and remount the heatsink.

Also make sure the fans are spinning on the card.

The fans are spinning fine and there isn't that much dust that it would cause problems.

The first time I tried this card it seemed to be artefact free for a few minutes. But after that even after a reboot or off for 5 - 10 minutes it would show artefacting right away.

Should i still go ahead and re-apply the arctic thermal paste ?

 

[Justin]

only thing you can really try is to bake the card, this card this reflowing

 

[Harrison]

Reapplying the thermal compound is fairly quick to do so you won't be wasting too much time or money giving it a go.

If that fails then Justin's suggestion to reflow the card is the only other option really left to you because the original solder on the card may have become brittle due to heat from the card over time (card heats up during use then cools down when off meaning the solder keeps expanding and contracting), causing dry joints within the solder. Reflowing will fill all the solder back in and reconnect everything on the board (if you are lucky).

 

[dougal]

Ok... I removed the plastic and fans and heat sinks. I cleaned the GPU and the other smaller GPU thingy from thermal paste. The paste seemed dryish.

I pre-heated the oven fan asissted for about 7 minutes at 190 degrees centigrade and then switched the fan off but left oven on and put the card in a tray lined with foil GPU side down and two foil balls on the side where there are no ports to support it and put it in the oven for 8 minutes.

Timer done and i switched off oven and removed the tray and now the card is cooling down for 30 minutes or so.

Afterwards i will re-apply thermal paste and put the heatsinks and fans on assembling the card and will test it.

Wish me luck.

 

[Sardine]

good luck hope it works for you, same on my 480gtx baked and lasted 2 more weeks, got it reflowed lasted 8 days,

 

[bebek]

Not a good idea sticking electrolytic caps in the 180 for this long - they are boiled .

 

[dougal]

It is still the same. It made no difference

I've given up.

I also tried my XFX 8800 GTS which is dead. Did the same procedure and it is still dead.

 

[Sardine]

as was explained to me 190c is not hot enough to melt the solder and any hotter would melt the DVI plastics.

you would need someone with a hot work station to apply an accurate 390c - 480c to reball the GPU and or ram chips.

 

[dougal]

as was explained to me 190c is not hot enough to melt the solder and any hotter would melt the DVI plastics.

you would need someone with a hot work station to apply an accurate 390c - 480c to reball the GPU and or ram chips.

Well in the case of the GTX 8800 a component (buzzer) actually fell off because the solder melted.

 

[Stedy]

Ok... I removed the plastic and fans and heat sinks. I cleaned the GPU and the other smaller GPU thingy from thermal paste. The paste seemed dryish.

I pre-heated the oven fan asissted for about 7 minutes at 190 degrees centigrade and then switched the fan off but left oven on and put the card in a tray lined with foil GPU side down and two foil balls on the side where there are no ports to support it and put it in the oven for 8 minutes.

Timer done and i switched off oven and removed the tray and now the card is cooling down for 30 minutes or so.

Afterwards i will re-apply thermal paste and put the heatsinks and fans on assembling the card and will test it.

Wish me luck.

To reflow lead free solder you need a controlled ramp up to 260C, more than most domestic ovens can goto. If you put the card in the pre-heated 190C oven, you severely thermally shocked the card, potentially damaging the PCB.

The original issue you describe was caused by issues with the Framebuffer interface, which consists of the GPU framebuffer and the GDDR RAM devices, which are also BGA. From experience, the smaller GDDR RAM devices are more troublesome than the GPU when graphical glitches occur, a proper SMT reflow can fix the issues, if the device has not died.

Ian

 

[sprcorreia]

as was explained to me 190c is not hot enough to melt the solder and any hotter would melt the DVI plastics.

you would need someone with a hot work station to apply an accurate 390c - 480c to reball the GPU and or ram chips.

What? 390c-480c?? Damn. That would melt the damn card...
It would make sense if you were talking Fahrenheit, not Celsius.

 

[ShambleS1980]

artifacts "when the gpu actually works to boot with" IMO are more often than not ram related (gpu ram)
i would underclock the ram speeds on the gpu to start it.
i doubt they have adequate cooling, if underclocking the ram seems to help then you may want to throw a few £ towards an after market cooler that comes with good ram heat synks. (or just buy them seperatly If the current gpu cooler would let you mount them (unlikley)

like i said though in my expirience when there are artefacts and the system dosent simply hang on boot or bsod instantly its more often than tan not the ram on the gfx card thats either dead dying over heating or running at the wrong speed. Gddr runs prety damned fast as standard. any reall factory or end user OC can easily cause issues. even at stock speeds you can have problems. usually you would rma the card, but i would start off underclocking the GDDR ram and possibly the gpu see if that fixes it and work from there.

 

[dougal]

I'm gonna have another go at it. But first a couple of questions:

If the previous owner overclocked it, would it still be overclocked now ?

I'm downloading the software from Palit. Since the computer boots up but with artefacts including on the PC bios etc, but Windows sees the GTX260 but with an exclamation mark saying it could not start the driver up, would the tuning software work ?

 

[ptp170]

as was explained to me 190c is not hot enough to melt the solder and any hotter would melt the DVI plastics.

you would need someone with a hot work station to apply an accurate 390c - 480c to reball the GPU and or ram chips.

What? 390c-480c?? Damn. That would melt the damn card...
It would make sense if you were talking Fahrenheit, not Celsius.

I think he means 390F - 480F (~200-~250C)

- - - Updated - - -

Ok... I removed the plastic and fans and heat sinks. I cleaned the GPU and the other smaller GPU thingy from thermal paste. The paste seemed dryish.

I pre-heated the oven fan asissted for about 7 minutes at 190 degrees centigrade and then switched the fan off but left oven on and put the card in a tray lined with foil GPU side down and two foil balls on the side where there are no ports to support it and put it in the oven for 8 minutes.

Timer done and i switched off oven and removed the tray and now the card is cooling down for 30 minutes or so.

Afterwards i will re-apply thermal paste and put the heatsinks and fans on assembling the card and will test it.

Wish me luck.

To reflow lead free solder you need a controlled ramp up to 260C, more than most domestic ovens can goto. If you put the card in the pre-heated 190C oven, you severely thermally shocked the card, potentially damaging the PCB.

The original issue you describe was caused by issues with the Framebuffer interface, which consists of the GPU framebuffer and the GDDR RAM devices, which are also BGA. From experience, the smaller GDDR RAM devices are more troublesome than the GPU when graphical glitches occur, a proper SMT reflow can fix the issues, if the device has not died.

Ian

What Ian said... It's also best to use a more accurate temperature gauge and pay attention to that instead of the oven settings. Don't leave the card to 'cool down' with the door shut either, open the door and leave it to cool down that way

 

[Bastich]

artifacts "when the gpu actually works to boot with" IMO are more often than not ram related (gpu ram)
i would underclock the ram speeds on the gpu to start it.
i doubt they have adequate cooling, if underclocking the ram seems to help then you may want to throw a few £ towards an after market cooler that comes with good ram heat synks. (or just buy them seperatly If the current gpu cooler would let you mount them (unlikley)

like i said though in my expirience when there are artefacts and the system dosent simply hang on boot or bsod instantly its more often than tan not the ram on the gfx card thats either dead dying over heating or running at the wrong speed. Gddr runs prety damned fast as standard. any reall factory or end user OC can easily cause issues. even at stock speeds you can have problems. usually you would rma the card, but i would start off underclocking the GDDR ram and possibly the gpu see if that fixes it and work from there.

I totally agree with this
All problems I have had with similar symptoms have been related to fried graphics card RAM.

 

[ShambleS1980]

if the previous user edited the bios files to change the clock speeds (would morst likley also edit the voltages and so on if they went to that much trouble) then yes it would always be over clocked..
if they used a overclocking application like msi after burner or similar then it would only be over clocked whilst the application was able to overclock it...

a lot of people will test over clocks and adjust voltages and keep doing that till it gets stable. and then edit the bios so it is always at those speeds. (i wouldnt do that my self)
Cards do get older and will usually start being unable to achieve the over clocks they used to, so to have to edit the bios to lower the clock speeds when needed would be a choire, but again most people just run the cards for 2 or so years then get the next great version.
I prefer to run them at a stable speed then keep a hold of them as an emergency card incase i have to RMA one (my boards dont usually have on board gpu's)

but the long and the short of it is.
if its dead then its probably not coming back to life.
if its close to dead then you can probably get it up and running if you slow it down. (whether your slowing it down from over clocked, or stock speeds makes no difference)
and if its still not working after you slow it down as much as you think is reasonable then its dead.

remember cooling is a huge factor with gpu's they run hot and they run at a high work rate. heat helps ions bounce about and break the little traces and stuff which messes up systems. i guess you could think of it as electrolisis (im sure thats not its correct name) but heat + high power + teenie tiny connections = ions bounce out of line taking some of the atoms that make up the wire with it, untill it creates a break or a short or even just enough damage so you dont get quite enough power to where it needs to be.

but as i said initially however. i would think the ram is to blame. whether its just running to fast due to an over clock or has degraded due to any nmuber of factors like esd or that electrolisis type thing i was just on about, or even its just over heating.. i would start by working on the ram. Slowing down the gpu core by 5% wouldnt hurt either.

 

[dougal]

I can't mess with any options because although windows finds it it can't start the driver, so the cards software won't work to tune it.

 

[ShambleS1980]

you could try msi afterburner. but it may not work either if it has an error 10