[BOUNTY] Help: AMD Drivers have stopped responding

[fourks]

UPDATE: BOUNTY FOR PERSON WHO SOLVES THIS. (not including: insufficient power).

So, after signing up and looking for the post new thread section for 10 minutes i saw the message saying new members can only post in the newbies section.

I have been mining for about two weeks now, the rig i am currently having troubles with is;

Asrock 890GX
2x Corsair 500w (1 for mobo and 1 card and other for 2 cards)
3x sapphire 5850s (Not crossfired)
AMD Sempron 140
2gb ram

Catalyst 11.6
Don't know how to check OpenCL version.

Windows 7 32-bit.
Guiminer.

So, cards 2 and 3 are running nicely around 330MHash/s with the -v -w 128 flags. However I am constantly having troubles with the card #1. Currently in GUIMiner it's showing 323.9MHash however it is frozen. (Other two cards are running fine). After I stop then start again the card will be fine for a few minutes and then the screen which is connected to it will go black an once it turns on a message saying the AMD drivers have stopped working and have since recovered. Though the hashing just isn't there.

Any help would be appreciated.

[shamen]

My advice is to install GPU-Z and GPU Caps viewer

maybe something got overheated? You can check driver versions, SDK version, bios versions, temperatures and fan speeds in the utilities mentioned. Thats the first place to start

If you remove the two 'good' cards, does the freeze still happen on the remaining card?


Also, a small overclock and RAM downclock and you can squeeze more juice out of the 5850's.. mines up at 366MH/sec (though thats with the 5870 bios)

[fourks]

Currently at 876/1000 but they're running a little hot. 77 - 60 - 63

GPU caps viewer says OpenCL 1.1 APP SDK 2.4

Also, it only seems to be the card in slot one that is doing this. I've tried moving them around.

[shamen]

temps seem completely fine. I run mine much warmer than that (81 to 82 - 950Mhz/600mhz @ 1.163v with fan settings tweaked to keep it at that temp)

If its only happening in slot one no matter which card is in there, you could try a mobo bios upgrade. Im assuming you have tried just one cardin the mobo in slot one and the problem still remains? If thats the case and a bios upgrade does not fix it, I'd RMA the motherboard as something is not right

[fourks]

Okay, Just have one card in there at the moment. Running fine 310Mhash/s.

edit: Asrock 890GX Extreme4 Only doesn't have a bios update on their website. -.- Won't be trying that then.

[shamen]

So slot works fine with only one card.

What PSU are you running? Have you monitored the amps and volts on the PSU with all three cards running? Does GPU-Z show the same(ish) volts and amps running on all three cards?

[fourks]

I'm using 2 500Watt Corsairs.

PSU 1 Connected to: Motherboard & GPU #2.
PSU 2 Connected to: GPU #1 & #3

At the moment:

GPU1: 191MHash/s, 1.088V
GPU2: 312MHash/s, 1.088V
GPU3: 132MHash/s, 0.950V

Just by looking at it, seems like the 500watt is quite enough?

[shamen]

500w dedicated for two 5850 GPUs should be fine IMHO. I would monitor the PSU voltages and amps to make sure they are within tolerance under load. Perhaps even swap the PSUs round to see if the problem remains?

Personally I run 700w for mobo+hdds + one 5850.

Not sure on multiple PSU setups myself as I always just get enough PSU wattage for the need at hand. It may be irrelevant nowadays, but I recall from years ago having to ensure the correct floating ground when using two PSU's. but as I say, that may be irrelevant in this case.

Anyone else have any thoughts?

[fourks]

I've been looking around for some info whether i have to have the correct grounding floor (no idea what it is) and can't seem to find much information. From what i've seen, most people with dual power supply set ups are just using a paper clip to kick start the power supply (like me) or purchasing a dual power supply adapted, which seems to do the same thing. Might have a look into swapping out my other rigs power supply if I can't find a solution.

[fourks]

No one has any additional thoughts?

[fourks]

Dual cards were working overnight just fine. 12 hours and no problems. All cards are fine when they run solo.

I do have a voltage meter for the powerpoint, power supply 1 (card two, mobo, hdd) according to the power meter is drawing about 650watt, 2.8Amp :s

Power supply 2 (gpu #1 & #3) 270watt and 1.2Amps.

[fourks]

Sorry for my inexperience but with what would I measure the voltage differences between the grounds? I understand that the grounds are the black connector pins but how do i measure it?

[fourks]

UPDATE: BOUNTY FOR PERSON WHO SOLVES THIS. (not including: insufficient power).

[InstaGx]

So the card in slot #1 always gets hotter than the rest?
You can't really say 77°C is OK on all cards or manufacturers. I suspect your airflow to be in a way that the card in slot #1 is getting all the hot air from the other 2 cards.
Are the fans at 100%? Also try underclocking the card in slot #1 and see if it gets any better.

I've experienced a lot of freezes with default clock and default autospeed of the fans with temps of about 70°C. Runs way more stable at 65°C max.
Temperature sensors aren't always reliable. They only measure at one specific point.

[Keichtech]

I had that problem with a card once, it turned out to be a faulty card.
I was able to get around it for a little while by overclocking my processor. AMD Phenom Quad core.
Using CrystalCPUID

The Program is CrystalCPUID (http://crystalmark.info/download/index-e.html#CrystalCPUID)
You have to run the CrystalCPUID and then run the TLB Which can be downloaded (https://rs90cg2.rapidshare.com/#!download|90cg|99264216|PhenomTLB_v_1.04_with_built_in_DLL.rar|264|R~0)

I don't know why that worked but it did. There might be a version out for yours.

I replaced the card as soon as possible, if yours is under warranty I would look into that.

-Keichtech