News AMD warns of card-cooking Crimson driver bug

Ouch, more fuel on the AMD inferno. Not been a great year or few for them. Shame really, I recall having a Sapphire 9800 (correct me if I am wrong on the number!) several years back that was awesome, without a shadow of a doubt the best bang for the buck at the time, akin to today's GTX970s maybe.

It was good when Nvidia has decent competition to keep them on their toes.

 

Here's hoping my 3 week old Fury X doesn't inherit this card destroyer.
Having said that, my card is constantly monitored so any high temps and I'd shut that down immediately, though I suppose even a few seconds at exceedingly high K would be very very bad for it.

 

TBH I'm skeptical of claims that user have had GPU failures, isn't there plenty of hardware based safeguards to protect against overheating?

 

Kind of what I thought really.
If there was a card breaking bug, surely it'd affect only certain GPU models, or core types? Either that, or it'd break absolutely every one of them. This way, it'd be pretty easy to find a correlation.

The claims I've seen come from loads of different card models so there's no consistency. Unless, of course, it's down to a software conflict or something.

I'm not sure what Crimson's overheat protection is tbh, never run into it as yet.

 

I could be wrong but i thought overheating protection was hardware based (maybe in the firmware), otherwise you could risk killing the hardware before the software (drivers) got going.

 

The only experience I've ever had with overheat was running my old XFX 7870. Hot hot hot card as it was, it would shut down the display driver altogether for a few seconds until the fans got the core temp down, then it'd reinitialise.
Since I've dumped that card, I've not seen a single figure out of place, especially in Crimson.

 

Probably a good thing I took the fan off my Fury X and used two of my own connected to the motherboard.

 

I have not had a chance to update yet thankfully

 

Oh goody

 

Are AMD really warning of "a card cooking bug" that will kill your card?

Or a bug that causes the fans to potentially underspeed and (by extension) overheat, but it should throttle/shutdown before damage as mentioned above?

It seems the only mention of the bug killing/cooking a card is one person on twitter.

It's pretty normal to overheat chips bench/stress testing them with insufficient cooling/fan profiles. This is because they have hardware/firmware protection against damage.

 

Regardless of what the problems are or not, it just shouldn't have happened at all. Wasn't Crimson supposed to be a step forward, to leave exactly this kind of negative GPU driver/software experience in the past? AMD have once again been left very red faced... Crimson even.

 

That's what I thought, unless of course the bug in Crimson can override the hardware interlocks some way.

 

I've come across a couple of *claims* by people saying this has killed their card, just in 1 games sub-reddit, whether there is truth in the matter, and likewise the claims they're using to suggest its happening to tons of people also.

My understanding is that the fan simply doesnt go above idle speed, is it possible the performance throttling isnt extreme enough to account for dead fans or idle speed fans while running intensive programs for extended periods? The majority of people ive seen responding to these issues have been saying they've not had this issue, largely because they're using something other than the AMD software to control their fan speed, such as MSI Afterburner.

Its not an excuse, but i think instances like this are a good example of why its important to have OSD stats showing. Instances like this should never happen, and that it has is entirely AMDs fault, but we can help protect ourselves too.

 

I'm not going to say that's impossible but it would be highly unusual (IMHO), to get the temperature to spike quicker than the throttling gubbins could react would probably need a bare GPU, i.e without a heatsink, at least that's my experience when i had trouble with a badly mounted heatsink.

 

Yeah, it does seem unlikely.

I had to return my 290X for RMA after xmas as both fans were extremely slow and their RPM at 100% was drastically lower than the 270X i was using in between to provide info for OCUK to RMA.

https://dl.dropboxusercontent.com/u/20670693/ocuk_rma/bf4_screendump.jpg
Thats what i sent to OCUK, temps hitting 94'c very quickly, fan speed hitting 100% but only reportedly spinning at around 600rpm for the majority of that (peaked at 1200 initially, but it fades off).
The RPM should hit 3000 (its what the replacement is hitting at 100% fan speed now), but in a stroke of luck its actually achieving about 20%, which is roughly what the AMD fault seems to be doing (maintaining idle speed of 20%, when on full GPU load and regardless of temps).

So my GPU was essentially doing the same thing due to faulty fans (the fans themself were roasting hot even 10-20min after removing the card, while the heatsink was room temp or close), and yet the performance throttling was keeping it alive and running to be able to run tests and get to the bottom of the issue, and i continued using the card for about a month, because i didnt fancy RMA'ing it before xmas because of holidays at OCUK and the manufacturer, busy post etc. It didnt die in a couple of days, which is what is being suggested for people days after the new drivers.

Doesnt make it impossible, but it doesnt match my experience, which while through different reasons, inadvertently created a very similar (near identical) scenario.

 

Yep, I would have thought the combination of passive cooling from a correctly mounted heatsink and limited active cooling from the fan would easily avoid any potentially damaging spikes even if you suddenly ran a stress test or something. With throttling and decent case airflow it might even run relatively normally.

 

The only real card killing fan bug was the Gtx driver which turned off the fan. A good number of gtx 480 and 580 died

 

I once used some Omega drivers with my 9800pro. I came home from work to find my flatmate 'playing' CS:Source with massive polygons all over the screen. The card was never the same again.

 

And 280s... I should know