5970 problems - seeking suggestions

[yochdog]

I have a couple 5970's that keep crashing my systems.  I am able to get one core rocking at 350 MH/s....but as soon as I start up the second core, everything freezes, the driver crashes, and sometimes there is a BSOD. 

Anyone have a potential solution? 

Win7 64-bit, 750W PSU, Sempron

[1713548670]

1713548670

[1713548670]

1713548670

[DeathAndTaxes]

I have a couple 5970's that keep crashing my systems.  I am able to get one core rocking at 350 MH/s....but as soon as I start up the second core, everything freezes, the driver crashes, and sometimes there is a BSOD.  

Anyone have a potential solution?  

Win7 64-bit, 750W PSU, Sempron

Need more info.

IF you startup only the second core is there any problems?
IF you startup both at stock is there any problems?

Core temps look ok?  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

I assume this is a clean install.  What driver version?  What miner?

How many total cards in the system? Just one?

Is there a monitor connected to this system? If so which GPU drives it (you can find it out by playing HD video on youtube and looking at GPU load monitor in GPU-Z)?

[jamesg]

I have a couple 5970's that keep crashing my systems.  I am able to get one core rocking at 350 MH/s....but as soon as I start up the second core, everything freezes, the driver crashes, and sometimes there is a BSOD. 

Anyone have a potential solution? 

Win7 64-bit, 750W PSU, Sempron

Don't over clock the second core.

Also, your core voltages should be 1.05v.

I have some OEM 5970s that are like this. It's is infinitely better to at least have them running then to overclock too much just to have them crash all the time.

[yochdog]

I have a couple 5970's that keep crashing my systems.  I am able to get one core rocking at 350 MH/s....but as soon as I start up the second core, everything freezes, the driver crashes, and sometimes there is a BSOD.  

Anyone have a potential solution?  

Win7 64-bit, 750W PSU, Sempron

Need more info.

IF you startup only the second core is there any problems?
IF you startup both at stock is there any problems?

Core temps look ok?  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

I assume this is a clean install.  What driver version?  What miner?

How many total cards in the system? Just one?

Is there a monitor connected to this system? If so which GPU drives it (you can find it out by playing HD video on youtube and looking at GPU load monitor in GPU-Z)?

Yes, if only the second core, it all goes haywire.
Yes, everything is at stock at the moment. 

Core temps look great.....60-70 range.  I will have to install GUP-z to get the VRM info.

Only one card.....this is my test bed system. 

It is an existing install.  Using the latest ATI CCC version.  Using GUIminer. 

The monitor is connected to the 5970. 

[Inaba]

The second core on your 5970 is toast... have a seen a few like that.  Not much you can do about it.

[DeathAndTaxes]

The second core on your 5970 is toast... have a seen a few like that.  Not much you can do about it.

Have to agree that seems likely.

To confirm you can try doing a clean install of OS/drivers. Leave card at stock, run only 2nd core.  If you get the same hard locks & BSOD then the core is bad.

The way that AMD designs the cards is almost like 2 independent "sub cards" connected with an internal PCIe switch.  This means you can have one of the two GPUs fail and the other one work properly just like you can have 1 of 2 physical cards (say 2x 5870 in same rig) fail and the other work perfectly.

[mtminer]

The second core on your 5970 is toast... have a seen a few like that.  Not much you can do about it.

It is toast, I am in the process of sending a 5970 back, because one of the cores died.

[yochdog]

The second core on your 5970 is toast... have a seen a few like that.  Not much you can do about it.

Have to agree that seems likely.

To confirm you can try doing a clean install of OS/drivers. Leave card at stock, run only 2nd core.  If you get the same hard locks & BSOD then the core is bad.

The way that AMD designs the cards is almost like 2 independent "sub cards" connected with an internal PCIe switch.  This means you can have 1 or 2 GPU fail and the other one work properly just like you can have 1 of 2 physical cards (say 2x 5870 in same rig) fail and the other work perfectly.

I am going to try this tonight.  Clean install from, the ground up.

Crossing my fingers that will fix it. 

[Proofer]

...  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

Is there a Linux tool to show VRM temps?

[bulanula]

...  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

Is there a Linux tool to show VRM temps?

Was wondering about this as well.

[sadpandatech]

...  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

Is there a Linux tool to show VRM temps?

Was wondering about this as well.

a port of gpu-z would be the ideal. the hitch is the cards that don't support reading those values from software. The few options I am familiar with are limited
read VDDC via command line
aticonfig --pplib-cmd "get activity"
Sensor 0 temp
aticonfig --odgt
You can try Radeonvolt which atleast will work with cards that support reading the values from software.
thread here; https://bitcointalk.org/index.php?topic=10228.0

Its funny that Bitcoin is poking the Linux world for such uses, which without it are almost non existent. ;p

[DeathAndTaxes]

...  VRM temps?  (GPU-Z gives detailed VRM and other temp info)

Is there a Linux tool to show VRM temps?

Was wondering about this as well.

Sadly I don't think so.  My understanding (which may be incorrect) is those supplemental values are NOT available via AMD drivers (no idea why not).  GPU-Z queries the GPU BIOS for raw data and parses it.  Those supplemental values are used for things like thermal throttling.  This is why you can see thermal throttling (usage drops from 99% to say 90%) despite cgminer (which is looking at core temp) shows a "normal" temp.

This is the only area that windows is superior to linux.  Maybe someday there will be a GPU-Z version for Linux.  My understanding is that it "could" be done.  

[cuz0882]

I just had a similar problem. I set the cgminer to start on boot up, soon as it loaded I got a yellow screen and it shut off. Now if I try to mine with one of my gpu's the whole computer crashes. Pretty sure mine just got overloaded and ruined. Would not be surprised if the same thing happened to you. To bad i bought mine on ebay last week  

[DeathAndTaxes]

I just had a similar problem. I set the cgminer to start on boot up, soon as it loaded I got a yellow screen and it shut off. Now if I try to mine with one of my gpu's the whole computer crashes. Pretty sure mine just got overloaded and ruined. Would not be surprised if the same thing happened to you. To bad i bought mine on ebay last week 

If the card was defective when you bought it use ebay buyer protection program.   Open a claim w/ ebay that "product is not as adevertized".  Since nobody has a clue what Bitcoin mining is try to see if you can replicate the problem w/ a game or some game benchmarking/stress testing utility like Furmark.

[yochdog]

I have managed to get them stable at stock clock speeds within a new installation of windows.  Crossing my fingers that it continues working for a long time!

Thanks for all the suggestions everyone!

[sadpandatech]

I have managed to get them stable at stock clock speeds within a new installation of windows.  Crossing my fingers that it continues working for a long time!

Thanks for all the suggestions everyone!

what versions of Catalyst, etc did you settle on?  And please tell me you are leaving out CCC? ;p  There are much friendlier OC tools available that aren't so picky or risky.

cheers

[DeathAndTaxes]

I have managed to get them stable at stock clock speeds within a new installation of windows.  Crossing my fingers that it continues working for a long time!

Thanks for all the suggestions everyone!

what versions of Catalyst, etc did you settle on?  And please tell me you are leaving out CCC? ;p  There are much friendlier OC tools available that aren't so picky or risky.

cheers

Agree on Catalyst I have found it causes nothing but problems.  The default install includes it but you can safely remove it or install the driver-only and then SDK runtime seperately.

cgminer + driver + sdk is all you need.  KISS.

[sadpandatech]

I have managed to get them stable at stock clock speeds within a new installation of windows.  Crossing my fingers that it continues working for a long time!

Thanks for all the suggestions everyone!

what versions of Catalyst, etc did you settle on?  And please tell me you are leaving out CCC? ;p  There are much friendlier OC tools available that aren't so picky or risky.

cheers

Agree on Catalyst I have found it causes nothing but problems.  The default install includes it but you can safely remove it or install the driver-only and then SDK runtime seperately.

cgminer + driver + sdk is all you need.  KISS.

aye, with the downloaded self installer you can extract the whole thing and then just run the 'Install Manager', which will allow you to selct just 'display driver' and 'SDK'(forget what it is called in the manager. Or just run the seperate driver and sdk installers which you will find within the extraced files.