XFX 6950 1GB Randomly shutting down

[Arkan]

Hi,
I've recently swapped out an 4850 for the mentioned HD6950 1 GB.

The 6950 card however seems to cut out (signal to monitor lost) & the monitor goes black after a few minutes playing a game. The previous card wasn't (& still doesn't) do this!

I'm a CompTIA student & have been building computers for years, so I'm no newbie when it comes to this kind of thing. I wonder if anyone else has any ideas however?

In Afterburner the GPU temp is reading about 77*c with the fan on full. Left at normal can rise to about 88*c at last glances before cutting out. I'm not sure It's an overheat problem with the GPU as the side of the case has been taken off & still the problem persists?

The CPU temp is fine & is not throttling. All overclocks have also been brought back down to stock speeds. CPU & GPU @ stock.

My 1st thought was it must be the PSU, so I swapped it out for an Powercool 80+ 750W PSU, it still happened. Ok, I thought maybe the Powercool is also a duff? So I then swapped it for an Corsair TX850M, still the same issue.

So PSU is 'not' the problem!!

I now suspect it is now a problem with the BIOSes with the card & the motherboard. Since my P55 Deluxe doesn't seem to be supported (strangely) any longer by ASRock. That means it's flashed with the most up to date BIOS already. The card is not unlocked to 6970 spec!! & the bios is checked as the original XFX BIOS here on the database.

There is no artifacting or screen tearing taking place just sudden hang, bump fan restarts on GFX card (change in fan speed / tone). There is sometimes a hang in audio also, then I have to reset the PC myself.

There are 2 monitors connected, the 24" 1080p monitor & a TV. I've disconnected the TV & still it happens.

The problem seems to be similar to the trouble the guy with another 6950 card is experiencing here, & albiet:
http://www.techpowerup.com/forums/showthread.php?t=183235&page=3

 

[OneMoar]

does it crash if you run furmark
It should't be shutting down at 77c or even 90c for that matter you could check and make sure the heatsink is making contact with all the parts of the card *VRMs*
its a xfx so it should be covered under there double lifetime warrennnnnnteeee id rma it with them sound like its just defective for the record a 1GB 6950 is a bit of a waste that card really needs 2GB

 

[Arkan]

I did reseat & repaste the cards heatsink, however I have not thought it might be down to the VRM getting too warm which it could well be. Some stock heatsinks are not always the best of contraptions.

What made me think it was the mobo's BIOS was that I once had an ASUS XONAR pci-ex card & it was incompatible. It would make the screen blank with no signal but the pc would still boot whenever it was installed, you just couldn't see anything. So that had to be removed.

 

[Arkan]

*Update bump*

I changed the cards cooler for an Gelid Icy Vision REV2 cooler. This did visibly bring temps down, although they were in spec before.

I played through Far Cry 3 for about an hour & thought 'hey, this has sorted the issue', however to no avail as it once again crashed / shut off & has been since.

One more thing I noticed was the one DIMMs slot of RAM was showing as not being there in BIOS. But in windows that module was recognised in CPUZ etc but in Window's exp specs it was only allowing 3.66 GB to be usable (while 8 GB was installed). I previously know this elsewhere to be an issue usually with the memory controller voltage dropping with all DIMM slots being in use. *I'm running 64-bit Win 7*!

However, I have since changed the card back to the old HD4850 & the memory problems have also vanished & everything is back to normal.

Problem with the XFX was removal of the warranty void (if removed) sticker, when I re-applied some better themal compound & re-set the old cooler previously. I cannot return the card!

I can test if it really is an Macaroni (faulty) card in another machine, however I will have to strip that & run it outside it's case with this card as it's in a tiny Micro ATX case (an extra hassle I could do without).

If the card turns out to be fine in another machine. I will almost certainly know it's a compatiblity issue with the p55 Deluxe mobo. Which will cause me even more headaches, as how do I know other newer AMD & NVidia cards won't also have the same problem with the board? The platform 1156 is still fairly relevant & up there still with the modern systems performance wise (especially from AMD) when it's overclocked. I could really do without needless spending on Sandybridge etc when I've a perfectly good system here. Impossible to get another type skt 1156 mobo here, as they seem phased out over 1155.

Might be worth a gamble for an Nvidia based card like an GTX660? I have heard of many similar shutoffs with 78XX & 79XX AMD cards in my pursuit of this trouble. Could be an AMD issue with some systems? Maybe going NVidia might be a wiser choice this time?

 

[OneMoar]

the memory issue makes me wonder if you need to reseat the CPU

 

[Arkan]

Hmm..because the memory controller is on board the CPU?

Thing is everything seems fine now the card is removed. In my searching I heard a problem with GPU core Voltage (& stability) being an issue with some 6XXX & 7XXX series cards shutting down in a similar way to this. This would usually result in the card needing to be returned but some able to up the voltage & get rid of the problem. I'm able to up the power limit / draw in Afterburner for the 6950 but not the Cvoltage. This obviously did nothing much.

Some seem to think it's just an AMD driver issue.
Could maybe try flashing to a card BIOS with a higher Cvoltage value?

 

[OneMoar]

Hmm..because the memory controller is on board the CPU?

Thing is everything seems fine now the card is removed. In my searching I heard a problem with GPU core Voltage (& stability) being an issue with some 6XXX & 7XXX series cards shutting down in a similar way to this. This would usually result in the card needing to be returned but some able to up the voltage & get rid of the problem. I'm able to up the power limit / draw in Afterburner for the 6950 but not the Cvoltage. This obviously did nothing much.

Some seem to think it's just an AMD driver issue.
Could maybe try flashing to a card BIOS with a higher Cvoltage value?

noooo no bios flashing
if you think the clock scaling is as fault there are better and safer ways to get around it such as using ATItray tools to force a consistent clock speed

 

[Arkan]

I'll check with ATITraytools' voltage tab after I get it set up on the other machine (if it's still not playing ball then).

I guess I had better see if that really is the problem or not 1st & eliminate whether the card itself is faulty , causing the same issue on both machines or not, then update on here.

I kind of hope it is in a way as cutting a loss ($£) on a GFX card instead of having to mothball the whole rig due to immature MOBO BIOS would be worse.

 

[jaggerwild]

P55 are famous for pin burn check the CPU socket for burnt pins CPU pads...........

 

[Arkan]

Ok, thank you.

I will check for this if the card shows signs that it is actually fine & dandy in the other machine & the problem points to something else.

I also realised I could underclock the card in regards to the voltage / core stability theory. If I underclock it by say 50mhz the current voltage should be sufficient for that even if not for the stock 800Mhz. If the card happens to run fine then we know it's likely a voltage issue (I wish I had thought of this before removing it from this machine now).

After spending on this card, 2 new PSUs & an £30.00 aftermarket cooler I also realised I could have likely bought a 7950! (or equivalent) ..lol. Part of me wishes I did now.

 

[Arkan]

*update*

Ok, the card's issue is down to a voltage one, & it needs more than the 1.1v tops available in Tray Tools' drop down menu.

I'll admit that at the minute I'm not sure whether it's down to the core or the memory. Memory has been clocked back 10mhz (just to make sure) & the core has been walloped down to 750Mhz (from stock 800Mhz). Although, I think it's fair to say the core is the best guess. It's working just fine no problems whatsoever at that speed. I'll try to edge it up & see how far it goes until problems persist again.

I reflashed the motherboards BIOS (just for good measure) & this time no errors have appeared on that front. It says it's making use of the full 8GB.

I would say down to dodgy binning or lower clocking cores from binning used in a more cheaper version of the board for release? Either way it's a poor clocking core.

So it looks like I have got some naturally occuring SE or LE cut down version of the card unless I can get more voltage into it in someway?

*I'll also add I ran AMD's driver cleaner & tried to reinstall Catalyst however that will no longer install after the driver cleaner was used. I am able to just install the device driver from device manager for the card.

 

[OneMoar]

*update*

Ok, the card's issue is down to a voltage one, & it needs more than the 1.1v tops available in Tray Tools' drop down menu.

I'll admit that at the minute I'm not sure whether it's down to the core or the memory. Memory has been clocked back 10mhz (just to make sure) & the core has been walloped down to 750Mhz (from stock 800Mhz). Although, I think it's fair to say the core is the best guess. It's working just fine no problems whatsoever at that speed. I'll try to edge it up & see how far it goes until problems persist again.

I reflashed the motherboards BIOS (just for good measure) & this time no errors have appeared on that front. It says it's making use of the full 8GB.

I would say down to dodgy binning or lower clocking cores from binning used in a more cheaper version of the board for release? Either way it's a poor clocking core.

So it looks like I have got some naturally occuring SE or LE cut down version of the card unless I can get more voltage into it in someway?

1.150 or 1.175 is the max *default* voltage
750 to 800Mhz is the Nominal clock for those cards
if setting a fixed clock and voltage didn't correct the issue then RMA the card with XFX

 

[Arkan]

I'm unable to RMA the card, I removed a warranty void sticker to check & reseat the heatsink 1st time around.

1.100V is all that is available within ATI Tray Tools. Any other apps give more voltage? What about Trixx?

I doubt I can hardlock the Mhz that 'do' work into BIOS either? It seems to be a locked ROM.

& these Catalyst issues. I think next time around it's going to be Nvidia's outing this time after 4 or so years with ATI / AMD.

 

[OneMoar]

I'm unable to RMA the card, I removed a warranty void sticker to check & reseat the heatsink 1st time around.

1.100V is all that is available within ATI Tray Tools. Any other apps give more voltage? What about Trixx?

I doubt I can hardlock the Mhz that 'do' work into BIOS either? It seems to be a locked ROM.

& these Catalyst issues. I think next time around it's going to be Nvidia's outing this time after 4 or so years with ATI / AMD.

had you told xfx that you where going to re-seat the cooler you could still rma it
you can try msi afterburner and see if it will give you anymore voltage options but I highly doubt it has any programmable voltage control
you aren't losing any performance at 750 or even 700Mhz vrs 800 Its a 1GB card you will run out of vram before you run out of core power

 

[Arkan]

My monitor is only 1900 x 1080.

I thought it only really started to make a difference gaming at higher resolutions than this at the moment? (which was why I went for the 1 GB version for now). It should last me fine until the new year. Then I'll see what Nvidia has to offer in the £140 - £170 range. Or I'll just get 2x 2nd hand GTX660s cheaper & SLI them.

 

[OneMoar]

My monitor is only 1900 x 1080.

I thought it only really started to make a difference gaming at higher resolutions than this at the moment? (which was why I went for the 1 GB version for now). It should last me fine until the new year. Then I'll see what Nvidia has to offer in the £140 - £170 range. Or I'll just get 2x 2nd hand GTX660s cheaper & SLI them.

I doubt it will be a issue just saying that a 1GB card is something to be avoiding now adays

 

[Arkan]

Sure. I think most likey it will be 2-3 GB available by the in the ranges I'm looking for anyway. But even then a higher GB will most likely be available across the board.

I noticed bad reviews of this XFX card on New Egg & Amazon recently, people RMA ing the cards for cutting out & intermittent errors similar to mine. I guess like I said, they kept poorer binned samples for the release of the cheaper cards. So poorly binned some wouldn't run at their rated speed set in BIOS. Shame I can't edit the BIOS so I can sell it later on, without having to explain how it needs to be underclocked etc in software.

Oh well, thanks for the help on the issue. Apart from the new Catalyst problem I think the issue is now solved.

*Just for the record Sapphire's Trixx utility. (The beta V3.0.4) has options that will change the voltage of the GPU!: https://www.sapphireselectclub.com/ssc/TriXX/

It is possible newer versions of the program will also do this but, since it says that is where gpu over volt support was added for 69XX cards on a beta I thought it a logical start, since someone said voltage options had been removed for their 2Gb version Sapphire 6950.

 

[Arkan]

Just to add further for anyone not able to overvolt there card & preferring Afterburner, running Sapphire's Trixx utility together simultaneously with Afterburner seems to automatically unlock Afterburner's core voltage options (without need for altering .cfg files)!

Even though I was not able to alter my .cfg files even though I had enabled the permissions to be able to do so, this is a good find for me.

 

[SaltyFish]

I used to run into such a problem. The video signal would get cut off randomly and get restored in anywhere from 30 seconds to a few minutes. It turned out my old cable was shoddy. Swapped in a different cable and never had the signal drop out. Do you have another cable you can try? If you need to buy one, under no circumstances should you get "boutique" brands like Monster or AudioQuest; any regular cable that obviously wasn't questionably put together should do. Monoprice is a good place to go for sanely-priced cables.