Alienware m17x R3 won't start and randomly shuts down - intermittent

About a year ago, I had to replace my damaged 580M graphics card to solve a problem where my system would simply power-off very frequently, sometimes even during Windows boot or very soon after logging-in. Since then I've stuck to the 314.22 drivers and gamed a fair amount, the system has been very stable. Details of the previous issue in this thread: http://forum.notebookreview.com/threads/alienware-m17x-r3-random-screen-blank-system-hang.759244/

Alienware m17x R3
This is a 64bit system with Windows 7 SP1. I'm running with the latest BIOS (A12) and Video Bios (70.24.2B.00.06, A01)
RAM: 16GB
HDD: 250GB SSD + 1TB SATA (replaced my 750GB Momentus which failed a month ago)
Graphics: Nvidia 580M with 2GB VRAM (no onboard card)
Screen: 120Hz 3D

• After a good run for about a year, the problem has now returned, in a far worse form. It started with a few random reboots during regular use (not while gaming) and worsened last week to simply powering off as before.
  
• Right now, if the system is shut down, pressing the power button only results in the alien head at the back of the case lighting up for a quarter of a second before going off again.
  
• The fans do start and the backlit keyboard sometimes lights up, but everything shuts down just like the alien head does - in under a second.
  
• Don't think the screen even comes on, so definitely not even getting to POST.
  
• I've noticed that sometimes the power button is lit (I see some light around the button), but most of the time it isn't. However, this state of the power button light does not affect the behavior, it still powers off in under a second.
  
• The blue ring on my power cable remains lit, always.
• After trying to start the machine repeatedly, it actually does start, eventually - could be after 5 minutes of trying or after an hour. I've managed to keep it on for hours in safe mode with networking and in normal mode too, but after I un-installed the graphics drivers (with normal mode + drivers, it would shut down soon after log-in, most times).
  • However, shutting it down in this state could result in inability to start again, even though it was working just a minute ago, without powering off itself.

Is my graphics card toast / on its way out again? Can't understand why it doesn't even make it to POST. I did keep a hardware logger on last week, just a couple of times, logging temps every second or so - the last records around the time at which the system went down did not indicate any excessive temps of CPU or GPU - I'm not even doing anything intensive on the machine.

The machine is almost exactly 4 years old now and therefore out of warranty. It performs really well otherwise, so I'd like to keep it alive for at least another year.

Dell India is pretty useless at diagnosis, so if I have to pay for a new graphics card, I can (last year I think the replacement was around $80), but would like your thoughts first. Thanks!

 

Hi there, it's sounding very much like another card failure - I went through three of them! I can't say for sure because you can't just pull the card and test the machine without it .

The 580 (and 675m) are among the worst cards as far as reliability goes I have ever seen. The 680m would be a much better bet and give your machine a new lease of life but it's not as easy as using the original card. You will need an unlocked bios and modified driver but there are plenty of threads on a '680m in the R3' to follow .

 

No No this sounds far more like a power supply issue. My old m17x R4 did that over and over again. The powersupply would get hot. Sometimes tell me it was running on battery power and just shut off. The unit would then stay off until the power supply had time to cool. (sometimes 10 mins, sometimes an hour.) I'd test it with another power supply long before I bought another GPU.

 

I plus this one Micky. I have been recently victim of not one but two in a row GPU failures so sadly it sounds the most likely cause of your issue. But if you can put one hand on a second PSU then you can test and I will be happy if loafer is right. See here for my story it can perhaps help you.

 

Thank you everyone for the quick replies.

A) I don't think it's the power adaptor; the system has also refused to start after being completely disconnected overnight with a very cool adaptor. I don't have another 12 A adaptor around, but I do have a 4 A one from another Dell machine (not ideal, I know - I do get a POST level complaint, I think, about a adaptor not supplying enough power) but I will try that when the system is in its happier state and is starting. I would love it if this were the problem, of course!

B) Unfortunately, the 680M is out of the question - it's still around $400 and shipping to India is a non-starter; more delays, higher costs, more uncertainty If it comes to that, I will probably buy a new machine.

What I can do, hopefully, is pay Dell $80 for a new 580 M, use that until it dies and keep paying for new cards (hopefully not more than once a year) until I move to a new system. $80 / year to keep this otherwise fine system going is not too bad and hopefully Dell can keep supplying them at that price (I will know more when I call in tomorrow).

I have a few questions:

1. Our theory last time around was that overvolting the card (which the later drivers did) was killing it slowly leading to ultimate failure. With my replaced card, I observed the voltage with the 314.22 driver and it was at 0.8, not 0.92 (but can't be sure it was *always* 0.8), and stuck with these drivers for a year. Now the card has apparently still failed in about 14 months.
   1. If not the voltage, are these cards just bad in a way that they will fail over time? Micky has gone through 3 ..
      
   2. Does pushing the card with games affect longevity of the card? For instance, if I get a new card and don't play games at all, are there chances that the card could last longer? Or is it likely to fail after a given amount of time anyway?
2. The situation this time is a lot worse than the last time - then I could boot even into Windows and it would power-off quickly, but now it won't even power on for more than a second - can a graphics card failure cause this sort of behavior?
3. When the system does start, I have been able to run for 5-6 hours in safe mode or in normal mode if the graphics drivers are not installed (I guess this points to the graphics card), any other deductions from this behavior?

Of course, I understand there isn't a way to definitely answer these questions, but definitely value your opinions.

 

I really don't know the answer. From my own experiences and that of many others I have worked with I came to the conclusion that something (probably the VRM circuitry) was built with components that are 'borderline'. Up the voltage or push them hard and a quick failure is on the cards (sry no funny intended!).

One fairly consistent trigger has been a driver update and I always thought that they messed around with voltages too much - and these cards do not support Boost. Now I wonder if it is just the way (poor quality) silicon gets burned in? Unused pathways degrade over time and then a driver update triggers problems? A real 'tinfoil-hat' idea.

My last failure (random shutdown) would run OK in safe mode for ages but after about a week it started to crash in safe mode. Until then I was thinking it may be power related, running in safe mode has the card in it's lowest power state. Replaced GPU under warranty.

Hopefully you can find a cheap 580m but although history points at the card it's by no means certain .

 

I have been running in safe mode and driver-less normal mode for over 12 hours today, but I guess my days are certainly numbered

 

Very probably. Driverless the card is in VGA compatibility mode and far less stressed. Good luck .

 

Thankfully, Dell does have the card, but this time the cost is $105, including an engineer visit to replace the card. A little more than I'd like, but OK - I've initiated the replacement process.

The bad:

• New card has a 90 day warranty only - I thought it was at least twice that.
  
• Old card will be taken away.
  
• Service request could take upto 21 days! Wonder if my machine will make it.

The machine does start with the alt power supply I have when it's in "happier state" (I do get a warning that I'm using a 90W adapter instead of 150W). When it's refusing to start though, the alt power supply doesn't help either - so not a power adapter problem

 

Just wanted to let everyone know that I ultimately did get the card replaced and it worked flawlessly ... for about 50 days, after which the same crap started happening. Since this was within the 90 day warranty period on the card, I scheduled a replacement, only to have this one fail in about 10 minutes after replacement. The odd part about this replacement is that it was showing a voltage of 0.92 volts quite frequently, whereas I never noticed this in the other cards (was using the same 314.xx driver).

Still, I noticed the video BIOS was a bit dated, it was 70.24.33.00.02 instead of 70.24.2B.00.06, so I updated the video BIOS and went back to Dell's original driver (version 269, from July 2011) - which made no difference to the voltage behaviour. After this, the system worked for a day and then failed in the same way. I assumed the newer drivers were responsible for overvolting the card, so was surprised that it frequently touched this level, even with such an old driver.

I'm getting my next replacement on Monday, with which I hope I have better luck. Any other suggestions on how to increase the lifespan? Could I perhaps overclock the card but actually just fix the voltages to 0.8 instead? (or a max of 0.8?) instead of actually overclocking it? Any other suggestions?

 

That 0.92v overvolt was introduced around v 314.22 which was the go-to driver for the 580m/675m. It's some sort of bug that the Boost 2.0 feature is kicking up the voltage when it is not supported on that core. After two 580m's and a complete replacement with an R4 - but still with a 675m which also failed, I went with a 680m and that thing is bulletproof !

It;s not even a big OV but with such a flaky card it doesn't seem be able to cope.

I never saw it myself but I saw here a couple of cards that refused to drop the voltage to the stock 0.87v once they had been overvolted, even with a different vbios - weird, and no help I'm afraid . NVidia Inspector allows you to change the voltages but I don't think it will help if it is a faulty boost action? Give it a go maybe?