AMD Radeon HD7970M possible failure issue - need to diagnose

Hi guys.

First, sorry in advance for eventual languages mistakes, i'm french-native speaker.
Also, sorry if i should have posted this thread in the "hardware-component" section of the forum (with my main problem being 'GPU' related, i thought this section was more appropriate).

I will need your help and your knowledge for a small diagnosis, to see if my GPU is "dead" and if no (i am praying really hard for this scenario), why i have the issues described below.
I purchased a Clevo P150EM (from Malibal) approx. 1.5 years ago with i7-3610QM proc, 8GB RAM and AMD Radeon HD 7970M as GPU (there is also a useless integrated Intel HD Graphics 4000) running with Windows 7SP1. I have thermal paste for both CPU and GPU.

Yesterday, I started having issues while playing games. After 5-10mins in game, the screen freezes (i mean nothing moves, not just a big lag) and i have no solution other than harbooting the pc. When the pc freezes, I can not go back to windows desktop or anything else.
My first thoughts were "F.... i hope it's not the GPU" (for the short story, Malibal did replace a previous defective HD7970M just few months after my purchase).
I tried several games with always the same result.
I tried different drivers (I was with catalyst 13.9 when the 1st "crash" happened, and I'm currently using catalyst 13.11beta v9.5) with no better results.

I decided to run 3DMark11 demo just to see if the card is still working but the demo crashed after like 1second and i was stuck in the exact same way i was stuck when playing games.

The laptop is not overheating, I did clean the vents and i was monitoring major stats (temp, fps, etc. using HWINFO, MSI Afterburner and OSD) during gaming sessions.

I'm a little bit novice and thus am stuck in the situation with not knowing what is happening. Given the price of the GPU and the whole laptop, it would be sad if the GPU is really dead.
The problem is that i'm not even sure the GPU is dead, the computer seems to correctly detect the GPU (it 'correctly' appears in Computer>System properties>Device manager>display adapters).


I don't know what to do and am looking for any idea, any advice for understanding what is happening. Do you have any tools for testing the GPU before i have to open the laptop?


Thanks in advance to all 'helpers' and to the whole NBR community.

Pim

 

When at the Alienware bois screen: tap the F12 key, choose diagnostics, and allow it to "test" the video card. Reply with results.

 

Hi Wwallender, I may be wrong but i don't think there is a diagnostic tool in my clevo BIOS. I think diagnostics are specific to dell/alienware BIOS (the Enhanced Preboot System Assessment -ePSA- diagnostics).
I see nothing like this in my clevo BIOS.

 

Some news:

I am still trying different drivers with no more success.

When monitoring GPU usage in-game, i'm at >95% most of the time wathever the game, when i m playing with a 1920.1080 screen resolution and then the screen is freezes after few minutes.
However, i just figured out that I stil can play games by using lower resolution and lower graphics.
With my settings in catalyst control center ("quality" for all games), games should work with the 7970 (and not the discrete HD4000), whatever the ingame settings (if I lower details and resolution directly in the game), right?
If i am not mistaking, it would mean that the GPU is still "able" to work and not 100% dead.

I still don't know where the problem is. I guess I will soon have to try a windows repair or a full formatting (using the 'special' procedures for SSDs). After that, if it is still not working, i guess i wil have to open the laptop and check the GPU in details.

I welcome any idea for fixing this.

 

I totally missed that you have a clevo XD. Even though you said the gpu is not overheating what are your temps?

 

Lots of new "observations" here.

Chronologically:
- I played again with drivers (using driver sweeper etc.) -> didn't work
- I tried to disable/enable the GPU in the device manager -> didn't work
- I performed tons of tests (Windows Memory diagnostics tools, dxdiag etc) -> everything looked strange
Windows Memory Diagnostics tools, in "complete" mode, was stuck at 20% -> I had to hardboot again (Windows Memory tests did work when set with "basic" and "regular" settings).
3DMark11 crashed in few seconds, and even gave me strange error messages like - "Workload single init returned error message: DXGI call JDXGISwap Chain: Set Fullscreen Sate failed"
- "The requested functionality is not supported by the device or the drivers: DXGI_ERROR_CURRENTLY_AVAILABLE"
- I tried to play a game and i got a beautiful "bluescreen" BCCcode 'a' meaning that I probably have hardware issues (I had the same kind of bluescreen when Malibal changed my 1st 7970M GPU).

Thus, I opened the laptop, removed the GPU and reseated it again (everything looked ok on the GPU, physically speaking, no burning or anything looking strange)
Strangely, after that, I was able to play a game (i tried with a non-demanding game: League of Legends) with lower resolution and details, and it worked well.
I then tried 3DMark11 which had not worked a single time since then, and the test did not crash.
I tried different configurations, and depending on if i am performing the test fullscreen or centered, and if my laptop is on "energy saver" or "balanced" mode, the test works ...or not.

The 7970 seems to be recognized by 3DMark11 but the score is really low ( http://www.3dmark.com/3dm11/7812042). It may be because the driver is not working properly however i'm using the latest "official" one from he AMD website and did a "clean installation).
Maybe it is because the score is for the integrated Inteal HD4000 Grpahics, but it seems pretty high for an integrated GPU.

Now, GPUz seems to recognize the 7970. It is a good point because it was not the case before I reseated the 7970.

I'm still investigating, but it seems that the 7970 does not accept to be "stressed" really hard. Each time i am trying to perform 3DMark11 test with power supply settings on "balanced" mode in Catalyst control center, it crashes (whatever it is centered or fullscreen test). When I am in "Energy saver" mode in catalyst control center, I can perform at least the test "centered".
I am wondering if the3DMark11 score i get is really for the 7970M or the integrated GPU. In an other hand, when I disabled the 7970 in the device manager, the integrated GPU was not even able to launch a single game (message error: "can't reach D3D" or something like this). So i don't think it could score anything on 3DMark11

I attach some pictures here so you can have a look on the GPU results and stats. I still don't understand were the problem is and am not even absolutely sure that the issue is "hardware" based. Sometimes the 7970 seems completely dead but sometimes it seems to work (with limitations though). It gives me some kind of "hope".

 

Clean out the dust in your vents and fans.

 

Your card might be reckognized by different software: Windows, GPU-Z and such.
But the silicon might be unstable, hence it runs on 300MHz instead of 850MHz. There is a thousand things that can go wrong with the GPU. It could be the power supply, it could be the silicon, it could be the VRAM.

Anyhow, to find out if it even runs at normal speed and if its able to substain it for a while,
Run 3DMark11 while keeping GPU-Z open.

After 3DMark11 is finished, in GPU-Z under "Sensors", click on the "GPU Core Clock", "GPU Memory clock" and "GPU Temperature". You should see "High", "Low", "Average" by the numbers,
What do you see as the "high" and "average" on all those three parameters?

 

Hi guys and thanks for your help.

To hockeymass: the fans are all clean

To Cloudfire: My 7970M seems pretty unstable. I cannot provide a "clear and simple" reply.
I did try few things with Catalyst Control Center "Power Play (PP)" and "Switchable graphics Global Settings (SWGS)".

*With PP on "Maximizing Battery Life"
-and SWGS on "Force power-saving graphics" -> computer use the integrated HD4000 -3DMark11 score P700 approx.
-and SWGS on "Optimize Power savings", "Optimize Performance" and "Maximize Performance"
-> computer use the 7970M with GPU core clock 450 MHz and GPU Memory bus Clock 300 MHz
-> 3DMark11 score 3400 approx.

*With PP on "Maximize Performance"
-> computer use the 7970M with GPU core clock 850 MHz and GPU Memory bus Clock 1200 MHz
In that precise case, the computer crashes (screen totally frozen, or even bluescreen) after 2-3 seconds in 3DMark. Impossible to obtain a score or monitor anything over time.



In all my tests, the 7970M does not maintain the GPU Core and Memory clocks (whatever it is 450 or 80 and 300 or 1200) and all the clocks fall to 0 MHz after approx. 5 seconds (the P3400 obtain in 3DMark11 with the 7970M is obtain with this problem occurring) . This seems to be linked with the fact that GPU Bios version (and the default clocks) are not detected (by GPUz).




..... However, in the middle of my tests, after a shutdown of the computer, gpuz did manage to recognize a GPU BIOS version, default clocks etc. In this case, it seems that the 7970M did manage to maintain the clocks for pretty long time. When this happened, I did try to change the clocks from the PP-"Maximize Battery Life" (450MHz core / 300MHz Memory) to the PP-"Maximize Performance" mode (850MHz core / 1200MHz Memory). The temps when pretty high quite quickly and the laptop shutdown. This is no good, but at least it seems that my 7970M was working properly (at least it was properly recognized and able to maintain clocks) if the temp issue did not create the shutdown (not 100% it is the temp that is responsible for the crash though). I did not have temp issue before and this may be caused by the fact that i did remove/reseat the GPU.




I don't really know what caused GPUz to recognize the BIOS again for this short time but it gives me some (not a lot i have to admit) hope, maybe it is not over yet with my 7970M. After the crash however, i' m back with my "strange" BIOS and my 7970M unable to maintain clocks.




I'm sorry for not giving "simple" reply and hope these infos can help you understand "the issue".
Thanks again for your help.