Announcement

Collapse
No announcement yet.

Burn in test 10.2 1004 (64bit) 2D/3D Graphic and memory (RAM) fail

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Burn in test 10.2 1004 (64bit) 2D/3D Graphic and memory (RAM) fail

    Hi Passmark Team

    We use BurnInTest 10.2(1004) to test , and found that most of the machines will have 2D/3D Graphics/Memory errors when the burn-in time is more than 160 hours.
    Could you please help us to clarify some question?

    1.if we observe happening sequencing, sometimes it is memory (DDR4) and sometimes it is 2D Graphic(Arc A770m).
    Why do you confirm it is Arc A770M driver issue?

    please reference Passmark David reply:
    "My guess would be that it is a device driver bug for the Arc A770M video card.
    Likely there is a small resource leak causing a exhaustion of resources after 70 hours of load. Which then causes a cascade of 2D and 3D errors.
    These cards are new from Intel, and it has been widely documented that they had driver issues. If you report it to Intel they might be able to get it fixed"​

    2.Intel team don’t have a way to trace log for issue which needs long time replication (and only shows error message by tool but no crash/TDR dumps)
    could you please share specific reason why RAM will report such message and see if there’s any way to generate crash dump when RAM starts.


    BurnIntest LOG and machine specifications are detailed in the attachments​
    Attached Files

  • #2
    For the 2D test the specific error code (-5) is the OpenCL error code CL_OUT_OF_RESOURCES, specifically when BurnInTest tried to use the OpenCL function clEnqueueReadBuffer this function failed and returned that error code.

    The "No free memory for buffer" message is likely due to low available RAM on the system, if you enable the trace logging in the logging options to at least level 1 then the .trace log should contain the amount of RAM the memory test is trying to allocate at this time.

    So yes it does sound like the system has a resource leak that after a few days will cause the system to run out of available RAM.

    Comment

    Working...
    X