r9 290 problem child

[DrMrLordX]

Hello! I'm about ready to throw in the towel on this r9 290, though I wanted to run it by youse guys to see if you can think of anything I haven't yet tried.

Here's the situation:

I grabbed a pair of reference Sapphire r9 290s off fleabay for semi-cheap with the intent of using them as mining cards. Both of them had horrible TIM applications. One would throttle even when severely undervolted and underclocked, but the other wouldn't (it would run at 92C but not throttle). It's the latter card that we're discussing here.

I popped the cooler off the non-throttling card and found that capacitor C105 had sort of popped off and was rolling around loose inside the cooler. No idea how I managed that, or even if I was the reason why it came off. Anyway I fixed the TIM application and put the stock cooler back on the card.

For a short while, everything was fine. Temps were lower, and I actually had the thing mining stably with 1120 MHz GPU clocks and a -100mV undervolt (memory was at around 1250 MHz since anything higher does not help). That lasted maybe 10 hours until it crashed. Now all it wants to do is bomb out and cause the driver to crash. Sometimes it'll take the driver down so hard that it'll bluescreen Win10 with THREAD_STUCK_IN_DRIVER. I've done multiple re-installs of the drivers (both 15.12 and 16.2.1) with no relief.

I even tried downclocking the memory to 1100 mhz, the GPU to 930 mhz, and lowering power limit to -50% to see if it would mine under those conditions. Nope!

So while I suspect I'm just going to have to take a loss on this thing . . . what else should I be trying here to save this card? If I can get it up and running without a soldering iron (I lost cap c105 anyway, shucks) then that would be great.

 

[sandorski]

It's probably beyond repair if you ran it without a Cap.

 

[Rvenger]

Yeah, if it was a very bad TIM job to tweak the board like this one I had, could be why the cap fell off.



hotobucket.com/user/rpisano01/media/WP_20160425_013_zpsip0bft8p.jpg.html]








If a pea size drop is too big then what is this?

 

[Elixer]

How the heck did that even pass QC?

 

[Shmee]

Unless they were for parts, you should be able to return them as not working properly.

 

[DrMrLordX]

The TIM job on mine was the opposite. Too little, with extremely poor contact with the die. I could still see shiny parts of the die immediately after removing the cooler on one of them.

Not sure if I'll be able to return the non-working one but I'll look into it.

 

[DrMrLordX]

One thing that mystifies me about this card is the constant thread_stuck_in_driver crashes. Well not quite constant, but near-constant. I wonder why the card detects and "works" but crashes like that so often?

 

[AntonioHG]

Had that thread stuck in driver and could never figure out what the reason was. Temps were okay, not overclocked on core or mem, TIM cleaned, reapplied and still would crash.

At first, it would crash after hours of playing then it became so bad I could launch a game and get hit with the thread stuck in driver.

I sent it in for an RMA and got it replaced.

 

[Killrose]

maybe can someone measure c105 and see what capacitance it has? then replace it and see if it fixes it. Would look very closely at the same area and surouning components to make sure others aren't cracked or missing too.