Freebie Nvidia XFX 9800GT - Lines on screen

[tom66]

My dad's Acer PC, which is one of my brother's older computers (the other sensible brother) started getting lines from the top to the bottom of the screen on the image. I immediately diagnosed it as a graphics card issue, and as my dad doesn't really play any games (uses it for eBay, email, Skype, etc.) I moved it over to the integrated graphics and took out the faulty card, works perfect now. Actually, I do remember that this card *always* had some lines on the Vista bootup screen, but never had them in the desktop - only recently (and suddenly) have the lines appeared on the desktop, and now on all boot stages.

So now I have a faulty "XFX" 9800GT and a known good PNY 9800GT so I'm thinking of some kind of awesome SLI setup, but first I need to fix the broken card.

I'm suspecting bad BGA joints; did the Nvidia cards of this generation have this problem, or is it something like DRAM? Anything someone like me can fix at a reasonable price?

[ratdude747]

yup, probably bad joints on the GPU or DRAM.

To fix it, what you should do is: (oven reflow method)

-remove the heatsink/fan, case bracket, and any lytic caps.
-loosly wrap anything plastic (usually connectors) with aluminum foil.
-add some flux under anything BGA (includes GPU, DRAM
-add some aluminum foil to a baking tray (to catch stray/dripping flux)
- place the tray in the oven and the card on the tray, GPU up.
- set the oven to 220-230 degrees C. Close the oven's door.
- open some (glass) windows; the flux may put off some nasty fumes.
- once the oven is up to temp (preheat over), let the card sit for 1 min, then carefully turn off the oven and open the oven door. Do not touch or move the card; the BGA joints are still molten!
-let the card sit in the opened oven for a while while the card cools down. once cool, remove any excess flux on hetsinked chips using your preferred method.
-while the card is cooling, clean any dust from the heatsink and fan. If you have the materials to do so, I suggest cutting and installing new thermal pads on the heatsink (esp. if they are the thin white kind, those need to go).
-reassemble the card using fresh thermal compound on the GPU.
-test the card!


(momaka taught me this method when my $5 8800GTS bombed)

[Koda]

Hey RD how long since you did the reflow? Does it still hold?

[Uranium-235]

an ideah!

why not put the card in there with a small vise putting some slight pressure on the gpu? pushing it down more, might spread the solder out some more make it harder to fissure again

[ratdude747]

Originally Posted by Uranium-235

an ideah!

why not put the card in there with a small vise putting some slight pressure on the gpu? pushing it down more, might spread the solder out some more make it harder to fissure again

no

all that will do is crack the die... even if you did spread the solder, it would short... the balls are too close to eachother.

as for the reflow holding, it wasn't long ago and the card is shelved until I have a case and HDD for the rig.

[momaka]

Originally Posted by ratdude747

yup, probably bad joints on the GPU or DRAM.

To fix it, what you should do is: (oven reflow method)

-remove the heatsink/fan, case bracket, and any lytic caps.
-loosly wrap anything plastic (usually connectors) with aluminum foil.
-add some flux under anything BGA (includes GPU, DRAM
-add some aluminum foil to a baking tray (to catch stray/dripping flux)
- place the tray in the oven and the card on the tray, GPU up.
- set the oven to 220-230 degrees C. Close the oven's door.
- open some (glass) windows; the flux may put off some nasty fumes.
- once the oven is up to temp (preheat over), let the card sit for 1 min, then carefully turn off the oven and open the oven door. Do not touch or move the card; the BGA joints are still molten!
-let the card sit in the opened oven for a while while the card cools down. once cool, remove any excess flux on hetsinked chips using your preferred method.
-while the card is cooling, clean any dust from the heatsink and fan. If you have the materials to do so, I suggest cutting and installing new thermal pads on the heatsink (esp. if they are the thin white kind, those need to go).
-reassemble the card using fresh thermal compound on the GPU.
-test the card!

That's how you do it

The only remark I would make is not to preheat the oven (or if you do, no more than 100C), otherwise you might give the PCB thermal shock, which can cause all sorts of nasty stuff such as internally damaged traces and whatnot. Most PCBs are quite tough, so they will survive that more often than not.
On the other hand, if your oven heats up very slowly, you should preheat the oven to about 100 to 130C. The reason for that is because some fluxes tend to bake/harden/burn off rather quickly and you don't want the flux to do that before the solder reaches the melting point (187C for leaded solder and 217 for lead-free). Also, if you're not using flux (note: this is not recommended), the longer you have the card in the oven, the more likely you make it possible for tin whiskers to form between the solder balls.

So overall, it's good to slowly let the card get to 220-230C, but not too slow. Once at that temperature, let it sit for 30 seconds to a minute, then cool it down.

Also, keep in mind that if there are any electrolytic caps, you *WILL* need to remove them if you want to do the oven reflow method. Cooked caps stink (quite literally)!

[ratdude747]

Originally Posted by momaka

That's how you do it

The only remark I would make is not to preheat the oven (or if you do, no more than 100C), otherwise you might give the PCB thermal shock, which can cause all sorts of nasty stuff such as internally damaged traces and whatnot. Most PCBs are quite tough, so they will survive that more often than not.

It's better to slowly let the card get to 220-230C. Once at that temperature, let it sit for 30 seconds to a minute, then cool it down.

Also, keep in mind that if there are any electrolytic caps, you *WILL* need to remove them if you want to do the oven reflow method. Cooked caps stink (quite literally)!

hence why I only used the term "preheat" with the word "over" after it. I mean that the preheat cycle is over and the oven is at the desired temp. review the procedure, as I have setting the temp (and implied, turning on the oven) AFTER putting in the card.

Hope that clears that up.

[momaka]

Originally Posted by ratdude747

hence why I only used the term "preheat" with the word "over" after it. I mean that the preheat cycle is over and the oven is at the desired temp. review the procedure, as I have setting the temp (and implied, turning on the oven) AFTER putting in the card.

Hope that clears that up.

If only I could learn to read properly . Sorry about the confusion.

BTW, I was editing my post above while you were typing yours, so your quote is "outdated" .

I see. (mods can view edit histories of any and all posts).

[tom66]

There are some purple Sanyo "SVP" polymers on the board, and some light blue ones, not sure who makes those (but they're not Sacons or anything, they might be Nichicons?) Also one FJ and one FK Panasonic cap. I suspect the FJ will not survive a reflow because through hole are supposed to be wave soldered after SMD, but the FK should because it -is- a surface mount cap and it would normally be reflowed. Overall, it seems better than my PNY which from memory has some poor caps on it. And it was bought very cheap in Thailand by my brother.

[momaka]

FJ and FK are electrolytic, so remove them if you can. IIRC, the wave soldering doesn't usually expose them to that high of a temperature (or rather I should say it's too quick for them to heat up that much). The oven, on the other hand, will have plenty of time to bring them to 220-230C. Loosely wrapping them in foil and with many layers might save them or it might not. If you feel experimental, leave them in .

[tom66]

FJ datasheet (Through hole):
http://www.ndb.com.tw/files/panasoni...ies%20Spec.pdf

Makes -no- mention of reflow resistance other than 210°C for 10 seconds soldering. So I will definitely pull that one.

FK datasheet (SMD):
http://www.panasonic.com/industrial/..._fk_series.pdf

Makes a mention of the reflow profile - one graph has 235°C for 60 seconds.

I'm pretty poor at removing SMD caps, so I want to avoid doing that if possible. Especially on an 8-10 layer graphics card. I'm used to single layer PCBs or maybe two layer ones...

[tom66]

In fact this is my card: http://images.bit-tech.net/content_i...n-review/5.jpg

Although that might be a later version. Mine only has one through hole cap (replaced by blue Nichicon(?)) and the upper most pad has an FK cap.

Maybe I can try wrapping all except the GPU and RAM in foil and use the grill instead of the oven to reflow only that area.

[ratdude747]

it's not too hard removing smd caps.

What I would do:

1. melt the joint on one side of the cap. while molten, bend the cap to the side away from the molten joint.
2. melt the other joint and while molten pull the cap.
3. carefully use pliers to straighten the contacts on the cap.

when you go to reassemble it, use some braid to clean the pads on the PCB. then add flux to the pads, place the cap on the pads, add a drop of solder to your iron's tip, and solder the first pad. repeat for the other pad.

[momaka]

Originally Posted by tom66

Maybe I can try wrapping all except the GPU and RAM in foil and use the grill instead of the oven to reflow only that area.

That may or may not give you a very good reflow.

Yes, removing SMD caps on multilayer boards sucks, but if there's clearance to reach the pads on both sides of each SMD cap, it's doable. Judging by the picture you posted, there should be enough clearance.

I do it pretty much like RD mentioned, except I do it in multiple phases.
1) Put flux on the pads for the cap you'll be removing
2) Heat cap on one side and wiggle sideways (perpendicular to pads) just a bit to loosen the cap. Repeat for the other pad.
3) Heat cap on one side and bend the cap towards the pad that you are NOT heating with the iron. Now repeat for the other side of the cap. Do not bend the cap too much, though.
4) Repeat step 3 above 2-3 times. When the cap is sitting about 1-2 mm off the board, you can finally repeat step 3 one more time and detach each side from the pad.
This method should give you completely undamaged caps. I just used this yesterday (yet again) to remove some SMD caps on a bunch of Radeon 9700 video cards. On one card, I had to remove the SMD caps 3 times (since I thought the first and second reflows didn't work properly). All of this using nothing more than a cheap 35W Radio Shack iron.

For putting the caps back on.
1) Put flux on pads on the board where you'll be putting the cap.
2) Hold cap in place (make sure the polarity is right!), then heat one pad until the solder melts. Repeat for the other side of the cap
NOTE: The only place where my method differs from RD's is that I do not clean the pads with braid or anything else (I may sometimes just reheat when the caps are off just to spread the solder evenly on the whole pad). It's actually easier if you leave the solder on there. The flux will do the magic to make sure the remaining solder on the pad wicks to the cap's legs. It will also make the cap sit a bit higher - this makes it easier to remove if you ever need to do it again.
....
Ah crap, it's light outside again. I think I better go to bed soon .

[tom66]

Oh - and I also have a Radeon X1200 which has major graphics issues after 5 minutes - stretched polygons/spikes mostly - I will probably try reflowing that too. It came from a friend, but its performance isn't as good as the 9800GT so it's been shelved... for now.

[ratdude747]

Originally Posted by momaka

Ah crap, it's light outside again. I think I better go to bed soon .

same here

[tom66]

So, I tried pulling the caps off the board, no dice. I ruined one FJ cap in the process - pulled the lead out of the cap, although I did get it off eventually. I am using a temperature regulated soldering iron and the same method I've always used for though hole caps, but that only works on single layer and *maybe* two layer boards, not 8-10 multilayer beasts... I will replace the FJ with an SMD cap as there are pads for it - the through holes are blocked. Anyway, I will reflow it with the electrolytics on it and watch the flames. If any test bad afterwards, I guess I will have to replace them.

[tom66]

I'm going to replace the FJ with a Panasonic FP:

http://www.farnell.com/datasheets/370232.pdf

This is the best capacitor I can find, but I'm not sure how it compares to the FJ because I can't find a datasheet for the FJ's!

[RJARRRPCGP]

Originally Posted by tom66

My dad's Acer PC, which is one of my brother's older computers (the other sensible brother) started getting lines from the top to the bottom of the screen on the image. I immediately diagnosed it as a graphics card issue, and as my dad doesn't really play any games (uses it for eBay, email, Skype, etc.) I moved it over to the integrated graphics and took out the faulty card, works perfect now. Actually, I do remember that this card *always* had some lines on the Vista bootup screen, but never had them in the desktop - only recently (and suddenly) have the lines appeared on the desktop, and now on all boot stages.

So now I have a faulty "XFX" 9800GT and a known good PNY 9800GT so I'm thinking of some kind of awesome SLI setup, but first I need to fix the broken card.

I'm suspecting bad BGA joints; did the Nvidia cards of this generation have this problem, or is it something like DRAM? Anything someone like me can fix at a reasonable price?

Ironically, I had video corruption and coloured letters on the BIOS screen with a 9800, too, but it was an ATi Radeon 9800 Pro. (A gazillion moons older than GeForce 9 series) Seems that ATi possibly had a bad case of solder issues.

[Th3_uN1Qu3]

Originally Posted by tom66

because I can't find a datasheet for the FJ's!

Here you go.