Author Topic: R9 290X problems  (Read 6382 times)

0 Members and 1 Guest are viewing this topic.

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
R9 290X problems
« on: February 02, 2018, 04:29:41 am »
Hello,

First time posting here. :)
So i received a bad Sapphire R9 290X reference from a friend. The video card is dead, no display on any output and it`s not recognized on windows (verified with teamviewer).
The fan spins, but the LED on the PCB is not on at any time.
I`m sure my PSU is ok and able to properly power the R9.
So i took the R9 apart and started measuring. Both the fuses were blown ( 1 - 2 ohms) and all the caps after the mosfets are shorted (0.03 ohms), including the capacitors under the GPU.
Since this is the first time i measure a video card, i have no idea if the values i measured are correct or not.
I checked for bad capacitors or missing components but i couldn`t find any.
Any idea on what to do next?  ;D
Any help would be greatly appreciated.

PS. While trying to boot the R9, the card was cold. Nothing warmed up.
« Last Edit: February 02, 2018, 04:32:29 am by LowRez »
 

Online Rasz

  • Super Contributor
  • ***
  • Posts: 2616
  • Country: 00
    • My random blog.
Re: R9 290X problems
« Reply #1 on: February 02, 2018, 05:09:59 am »
Capacitors are most likely not shorted, but your multimeter sucks. GPU resistance on main power line nowadays is in the milliohm range. 200W at 1V, do the math.
At the same time its possibile one of the mosfets shorted to ground.
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 
The following users thanked this post: LowRez

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #2 on: February 02, 2018, 05:51:49 am »
Ok, then i`ll get a new multimeter..
Any other ideas? I`m really stuck.
 

Offline llkiwi2006

  • Regular Contributor
  • *
  • Posts: 105
  • Country: nz
Re: R9 290X problems
« Reply #3 on: February 02, 2018, 06:56:40 am »
You should check if the on-board dc-dc converters are working. Is there anyway you can measure the output of them?

Also what fuses are you referring to?
 

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #4 on: February 02, 2018, 09:00:38 am »
Thanks for the reply.
I will check when I get home.
In the red squares are the two fuses.

LE.: I will take down the inductors one at a time and see if the short disappears.. Tried desoldering a mosfet and it`s a pain without heating the PCB too.
I`m thinking of applying some (0.9V) voltage and measure the voltage drop. Is 0.9V safe?
PS. the card is mine, the friend I wrote about earlyer wanted to throw it away.. So I can risk anything without a problem. :D
« Last Edit: February 02, 2018, 09:12:34 am by LowRez »
 

Offline Mr. Scram

  • Super Contributor
  • ***
  • Posts: 9810
  • Country: 00
  • Display aficionado
Re: R9 290X problems
« Reply #5 on: February 02, 2018, 09:06:49 am »
Capacitors are most likely not shorted, but your multimeter sucks. GPU resistance on main power line nowadays is in the milliohm range. 200W at 1V, do the math.
At the same time its possibile one of the mosfets shorted to ground.
Very good quality multimeters have trouble measuring these tiny capacitances too. It's typically the territory of LCR meters.
 

Offline llkiwi2006

  • Regular Contributor
  • *
  • Posts: 105
  • Country: nz
Re: R9 290X problems
« Reply #6 on: February 02, 2018, 10:04:08 am »
I`m thinking of applying some (0.9V) voltage and measure the voltage drop. Is 0.9V safe?

The chip might need a certain power supply sequencing, so putting any voltage on a single rail could be unsafe.
 
The following users thanked this post: LowRez

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #7 on: February 02, 2018, 10:19:22 am »
Quote
The chip might need a certain power supply sequencing, so putting any voltage on a single rail could be unsafe.
Ok then, I`ll stick to taking the inductors off the pcb and see if the short "moves" to the mosfets. :)
Also I`ll check the dc-dc converters.
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: ca
Re: R9 290X problems
« Reply #8 on: February 02, 2018, 06:29:49 pm »
HDMI and DVI ports have a +5 V output to supply a small amount of power to some devices like repeaters/equalizers etc. The fuses are for those 5 V lines, and they are not blown. They are likely poly fuses, and the resistance you measure is normal. A blown fuse is always an open circuit, not 1 or 2 ohms.

Measure the voltage across the power supply outputs, across each of those blue/silver caps stacked vertically near the RAMs. They should have the GPU core or RAM voltage on them. I don't know those off-hand but likely about 1 V and 1.5 V respectively. If you are "lucky", you have a dead power supply.
 
The following users thanked this post: LowRez

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #9 on: February 03, 2018, 02:51:48 am »
HDMI and DVI ports have a +5 V output to supply a small amount of power to some devices like repeaters/equalizers etc. The fuses are for those 5 V lines, and they are not blown. They are likely poly fuses, and the resistance you measure is normal. A blown fuse is always an open circuit, not 1 or 2 ohms.

Measure the voltage across the power supply outputs, across each of those blue/silver caps stacked vertically near the RAMs. They should have the GPU core or RAM voltage on them. I don't know those off-hand but likely about 1 V and 1.5 V respectively. If you are "lucky", you have a dead power supply.

Just measured, 0V on all capacitors you said i should measure.
 

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #10 on: March 12, 2018, 02:57:57 pm »
Hello again,

Sorry for the late reply.
The short circuit was on the GPU.
Searching for a spare part now..
Found one in China for 130USD. Is it worth it?

 

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2289
  • Country: us
    • medpants.com
Re: R9 290X problems
« Reply #11 on: March 12, 2018, 04:39:43 pm »
Definitely not.  Unless you have specialized tools and a lot of experience, the success rate for replacing a large, sensitive BGA part like that is basically zero.  There's also a good chance that the chip itself isn't the problem - an above post did mention that a typical resistance between rails for the GPU's power supply is typically quite low because of the massive amount of current they draw.... so it's probably not a simple short either.  0V on the bypass caps doesn't necessarily mean the chip is shorted, it can just as easily be that the things supplying the power are dead.


If you can't fix the problem by messing with external components, it's unlikely it can be fixed without a specialist, and it's extremely unlikely it can be fixed for less than the price of a new one.
« Last Edit: March 12, 2018, 04:41:34 pm by DaJMasta »
 

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #12 on: March 12, 2018, 06:05:00 pm »
I made a few reballs on iphones successfully.. I guessed it just requires some higher wattage tools because of the larger components, thickness of the board etc.
I thought 0.3 Ohms was a short-circuit because my multimeter shows that exact same value when i touch the testers together..
I measured with the same multimeter my GTX960 and a RX470. The lowest value i got was 0.9 Ohms on the rails (both cards are working). That`s three times more than the R9 290x.
 

Online Rasz

  • Super Contributor
  • ***
  • Posts: 2616
  • Country: 00
    • My random blog.
Re: R9 290X problems
« Reply #13 on: March 13, 2018, 04:43:33 am »
I made a few reballs on iphones successfully.. I guessed it just requires some higher wattage tools because of the larger components, thickness of the board etc.
I thought 0.3 Ohms was a short-circuit because my multimeter shows that exact same value when i touch the testers together..
I measured with the same multimeter my GTX960 and a RX470. The lowest value i got was 0.9 Ohms on the rails (both cards are working). That`s three times more than the R9 290x.

as I said in first post:
>your multimeter sucks. GPU resistance on main power line nowadays is in the milliohm range. 200W at 1V, do the math.
GTX960 is 1/2 to 1/3 that.
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #14 on: March 13, 2018, 10:57:12 pm »
Yep.  It sucked big time.  :palm:
Went and bought a cheapo Mastech MS8040.
This one confirmed there was a dead short on the CPU. 0.04 ohms.
 

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2289
  • Country: us
    • medpants.com
Re: R9 290X problems
« Reply #15 on: March 14, 2018, 05:47:54 am »
Which is 40 milliohms, which is consistent with what rasz is saying is fairly standard.


Basically, the random variation from the contact resistance on your probes may be significant enough as to be greater than the actual resistance of a working chip.


And still I think the likelihood that it's a power element for the chip that has failed or a bad solder joint/trace that would not be fixed with a GPU replacement, provided you could actually do the replacement well, is higher than it being a fault in the GPU itself.


Have you checked every phase of the board power supply?  It would only take one for the whole thing not to work - either just not meeting the current requirements of the chip or being auto shut off for lower than allowed voltages.
 

Offline llkiwi2006

  • Regular Contributor
  • *
  • Posts: 105
  • Country: nz
Re: R9 290X problems
« Reply #16 on: March 14, 2018, 10:57:24 am »
Have you checked every phase of the board power supply?  It would only take one for the whole thing not to work - either just not meeting the current requirements of the chip or being auto shut off for lower than allowed voltages.

Genuinely curious, how would you go about doing that short of cutting the traces coming out of every phase?

Also for the resistance thing, I thought cmos chips have very little idle (as in when not switching) power consumption? So a low resistance across the rails would still be suspicious even if the chip is rated to dissipate a lot of power? Then again I don't know much about GPUs, maybe they have some weirdness like intel cpus have on silicon switching regulators. My experience with microcontrollers and other small cmos devices is that you will measure a high resistance (mega ohms) between power and ground rails.
 

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #17 on: March 14, 2018, 12:17:10 pm »
Have you checked every phase of the board power supply?  It would only take one for the whole thing not to work - either just not meeting the current requirements of the chip or being auto shut off for lower than allowed voltages.

Genuinely curious, how would you go about doing that short of cutting the traces coming out of every phase?

Also for the resistance thing, I thought cmos chips have very little idle (as in when not switching) power consumption? So a low resistance across the rails would still be suspicious even if the chip is rated to dissipate a lot of power? Then again I don't know much about GPUs, maybe they have some weirdness like intel cpus have on silicon switching regulators. My experience with microcontrollers and other small cmos devices is that you will measure a high resistance (mega ohms) between power and ground rails.

I unsoldered the coils. On the power (with mosfets) side I had no short.
On the GPU side I had a very low resistance (0.3 Ohms with my old multimeter).
I forgot to mention in the previous post: the new multimeter has REL? button. Without that REL?, The multimeter shows 0.04 Ohms with testers touching.
What I did: shorted the testers, pressed the REL? button, the multimeter now showed 0.00 Ohms.
With the coils still unsoldered, measured between GND and each V+ rail (GPU side).
Multimeter showed 0.00 ohms on each rail.
I did manage to take the GPU off the PCB and I cleaned the PCB and GPU.
Found GND and V+ rails "under" the gpu. No more short-circuit. Same on the PCB, no more 0.00 Ohms, even after I soldered the coils terminals back.
Measured on the GPU directly, same method with REL? , 0.00Ohms.
I believe that this multimeter is a little better than my old UNI-T.
On the other side, I did order a stencil for this GPU and a GPU from China. They will arrive next month probably.
If this still won`t fix the card, lesson learned. But I see it as experience.
 

Offline MyHeadHz

  • Regular Contributor
  • *
  • Posts: 165
  • Country: us
Re: R9 290X problems
« Reply #18 on: March 14, 2018, 01:31:56 pm »
This may or may not pertain your exact problem, but I have seem many similar threads about GPU repair, mostly reballing.  I also once went down the rabbit hole checking all sorts of problems, and eventually failing... just to have reballing as my last resort and it work.  Feel free to ignore this post if you want, but I really do think it could help this type of topic to have someone test out the below.

A bit of back story and rationale: These type issues have annoyed me a lot.  This is another of many such threads.  More often than not, people just get into arguments, leaving the problem unsolved or half-assed.  There are some who insist that nothing less than spending hours on a full reverse-engineered schematic to find the exact cause is the only way, others who go straight for the reball because most of the time the problem is ultimately an open/short from heat stress on the BGA, and the third group who say no matter what you do it's temporary so don't mess with large BGA 's.  I think all three methods have their logical flaws based on assumptions that aren't necessarily well-founded.

GPU's, especially the r9 series (I have one myself) and similar AMD cards, run quite hot and oxidize the solder from repeated reheating and cooling, esp with the lead-free solder that they use for ROHS compliance.  I've "reballed" several GPU's and have had mixed success.  I've had more success where I added pressure and increased cooling/airflow as part of the repair process.

These cards are manufactured professionally using solder, so it is possible to do it properly.  The question is, HOW to do it properly (without professional equipment)?  Anyone have any tips?

Perhaps he is not getting it hot enough (lots of thermal mass to heat up)?  Perhaps the solder is too oxidized already... so how would one go about taking care of the oxidation?  Would it be possible to just add a lot of flux around the BGA before reballing, thus chemically taking care of the oxidation?  Would it be possible to do a Rossman-style BGA removal and reinstallation with more standard tools?  By that I mean a basic hot air gun and some kapton tape.    is the type of technique I'm talking about.  That and some solder paste.

It seems perfectly doable, with the main question being thermal capacity needed vs what your hot air system can produce.  I've reballed GPU BGA's with my 300w Radioshack heat gun before with success, but I don't know if it would take more heat to remove the GPU and resolder with fresh solder paste.  Any input?  I usually get the BGA quite hot (and protect other stuff with kapton tape) when I reball, and so far everything is still holding up... in other words, I don't have any non-functional boards to test with.

It would be very valuable service to the forum if someone with experience trying those things (on consumer setups, preferably) would chime in, or if someone would be willing to try those things and report the results.

Or, should I just assume that people will argue no matter what and nothing will ever change.  :p

(I've checked the code and I don't see any errors...  I don't know why that one video is embedding itself instead of being displayed properly as a link.)
« Last Edit: March 14, 2018, 02:02:04 pm by MyHeadHz »
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 735
  • Country: de
Re: R9 290X problems
« Reply #19 on: March 14, 2018, 01:48:42 pm »
Have you tried measuring V+ and GND in reverse with the diode-test?
If you see 0V, there's a short in the Chip. Otherwise you should see a drop of about 0.2V across the parasitic diodes.


Btw. resistance of the GPU is NOT in the mOhm-Range. The 200W are dissipated when the GPU is doing stuff. When idle (Windows-Screen), thy draw about 30 to 40W. And without any oscillator running, they are a static CMOS-Chip that draws very little power.

Have you measured the voltage across the capacitors on the back of the card? And the Gates of the MOSFETs of the OnBoard-PSU?
If you see 0V across the caps, but activity on the gates: GPU dead (short circuit). If there's no activity on the gates: OnBoard PSU defective (either the main control chip, or the one that distributes the PWM-Signal to the different stages).
From my experience, failures of Graphics Cards are caused by these defects (sorted by probability):
- Ceramic Cap fails Short Circuit - Often quite spectacular und many times not repairable due to excessive PCB-Damage
- GPU fails partially (most of the time GPU still works kinda, but either produces a ton of glitches on screen, or fails once text-mode is exited and some graphical-mode is activated) - This is sometimes correctable with a reball or reflow for some time. With certain older boards (Built around 2008 to 2011) and leaded solder, this problem can even be completely repaired in some cases.
- MOSFET fails Short Circuit (easy to locate: just look for burn-marks :) ) or Open Circuit (hard to locate, card often still works, but gets unusually hot, or develops some crazy quirks that mislead you big time)
- GPU fails completely (Depending on the Series, this failure can be much more or less likely - The R9 290 seems to be a little more likely to develop this kind of problem)
- PSU Control-Chip fails and the OnBoard PSU doesn't regulate anymore.

Of these problems, the capacitor, MOSFET and PSU-Control Chip failure are good to repair. If it's the GPU itself, there's no point in trying to repair the damn thing. It WILL fail on you again. Maybe in 3 months, or in 3 minutes, but it WILL fail again (one exception to this rule: GPU with bad solder-joints due to wrong soldering-profile with lead free solder - Older Cards that don't deserve the attention anyway :) ). The only thing you can do if the GPU is the problem is to replace it, but the chips are expensive enough that it's better to just buy a new card if the chips can be sourced at all.
« Last Edit: March 14, 2018, 01:57:10 pm by SaabFAN »
 
The following users thanked this post: LowRez

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #20 on: March 14, 2018, 02:19:53 pm »
@MyHeadHz

Lots of flux around the chip. That worked for me.
Well I used a 2KW air gun lol.. But from a safe distance, and a temperature probe was right next to the chip so I could read the temp in real time, precisely. Did not go over 240*C, and that was not for a long time.
Before that, preheated the whole board at 70*C for about one hour (I know, not enough) so that the moisture would disappear.
Had in mind for some time to build some sort of rework station using recycled parts.
4x1KW halogen bulbs in series for the bottom heater, and I`ll find some top heater locally.
All controlled with a raspberry pi with picoreflow software installed.
Maybe I`ll do it before the chip and stencil arrive.  :-//

@SaabFAN

Nope, did not do the reverse measuring with the diode test. Too late now.
I did measure the voltage with the coils off and I had 0.2V on all rails. But I was thinking that there could be a feedback limitation..
I found the chip here: http://betteryoyo.com/index.php/Home/Goods/goodsInfo/id/565181765705.html
About the PSU control chip.. Isn`t that programmable? Just curious. :)
 
The following users thanked this post: MyHeadHz

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 735
  • Country: de
Re: R9 290X problems
« Reply #21 on: March 14, 2018, 02:47:43 pm »
@SaabFAN

Nope, did not do the reverse measuring with the diode test. Too late now.
I did measure the voltage with the coils off and I had 0.2V on all rails. But I was thinking that there could be a feedback limitation..

That was due to no connection from the power-MOSFETs to the Chip. Without the coils, there was only a capacitive coupling there. Surprising that there was any voltage at all :)
The OnBoard-PSU is a Buck-Converter. https://en.wikipedia.org/wiki/Buck_converter Both the switch and the diode you see in the simplified schematic are implemented with MOSFETs.

About the PSU control chip.. Isn`t that programmable? Just curious. :)

Well... Kind of. There are some with i2c-Interface, but most use simpler methods to set the output-voltage. Dedicated pins to encode Voltage IDs for example. Depends on the manufacturer of the card and the revision, which one is used.
More sophisticated ones can even adjust their frequency depending on the load.
But they also work without any input from a control-chip by defaulting back to some preset configuration.

Btw. Some cards (My R9 390 for example >:( ) refuse to work if you try to adjust the voltage even a little! Which severely limits overclocking-potential. Especially if there's tons of room in the thermal budget because of good heatsinks and airflow inside the case.

Online Rasz

  • Super Contributor
  • ***
  • Posts: 2616
  • Country: 00
    • My random blog.
Re: R9 290X problems
« Reply #22 on: March 14, 2018, 04:36:30 pm »
Btw. resistance of the GPU is NOT in the mOhm-Range. The 200W are dissipated when the GPU is doing stuff. When idle (Windows-Screen), thy draw about 30 to 40W. And without any oscillator running, they are a static CMOS-Chip that draws very little power.

you would think that, right? now go ahead and measure one. Pull out your R9 390 and check resistance - Im genuinely curious, I never measured any AMD cards, only Nvidia.
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Offline llkiwi2006

  • Regular Contributor
  • *
  • Posts: 105
  • Country: nz
Re: R9 290X problems
« Reply #23 on: March 14, 2018, 08:32:04 pm »
Well I used a 2KW air gun lol..

I have to say you are a brave soul, I'm glad it worked for you. Personally I wouldn't touch bga rework, especially not with multi-thousand ball bga. With my luck the board will delaminate and/or the pads will get ripped off.
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 735
  • Country: de
Re: R9 290X problems
« Reply #24 on: March 15, 2018, 01:03:58 am »
I got pretty much mixed results: Depending on where I measure first, I get either 0 Ohms (200 - 300 mOhms, to be exact, but that is the resistance of the leads) or 2,4 Ohms. A LOT less than I was expecting ;D
I have measured across the caps on the back of the GPU on the card.

Offline LowRezTopic starter

  • Contributor
  • Posts: 11
  • Country: ro
Re: R9 290X problems
« Reply #25 on: March 15, 2018, 11:51:49 am »
Well I used a 2KW air gun lol..

I have to say you are a brave soul, I'm glad it worked for you. Personally I wouldn't touch bga rework, especially not with multi-thousand ball bga. With my luck the board will delaminate and/or the pads will get ripped off.

Well I did happen to break some pads.. But not on this board.
That`s fixable, you just need steady hands and a good camera/microscope.
And this too: https://www.zipy.ro/p/ali/0-01-0-02-mm-pure-copper-wire-line-for-fix-repair-iphone-ipad-phone-motherboard-logic-board-pcb-fingerprint-sodering-point-tools/32767228143/     8)

But that happens when you force the chip off the PCB (for various reasons), or when the pad is "pre-broken".
And about the delamination... Never happened to me yet.
But, as I wrote, I used the temp probe from my multimeter, and didn`t go above 240*C and that was for a short time.
Now I`m wondering if I should use solder paste (same I used on phones) or solder balls...
I guess it will depend on the thickness of the stencil.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf