Burnin test show 2D 3D GPGPU test fails on 23H2 MTL platforms

[LiCheng]

Hi
We met 2D 3D GPGPU test fail on our platform. We test MTL ultra 135H CPU on 23H2 OS. detail information as below. Please help to check log and if there is any solutions.

CPU:
CPU manufacturer: GenuineIntel
CPU Type: Intel(R) Core(TM) Ultra 5 135H
CPUID: Family 6, Model AA, Stepping 4
Physical CPU's: 1
Cores per CPU: 14
Threads per CPU: 18
P-Cores: 4
P-Threads: 4
E-Cores: 10
E-Threads: 20
Hyperthreading: Enabled
CPU features: MMX SSE SSE2 SSE3 SSSE3 SSE4.1 SSE4.2 DEP PAE Intel64 VMX SMX AES AVX AVX2 FMA3
Clock frequencies:
- Measured CPU speed: 4384.5 MHz
Cache per CPU package:
- L1 Instruction Cache: Unknown
- L1 data cache: Not applicable
- L2 cache: Not applicable
- L3 cache: Not applicable

Memory
Total Physical Memory: 32437MB
Available Physical Memory: 27208MB
Memory devices:
Controller0-ChannelA:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller0-ChannelB:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller0-ChannelC:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller0-ChannelD:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller1-ChannelA:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller1-ChannelB:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller1-ChannelC:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Controller1-ChannelD:
- LPDDR5, 4096MB, 7467MHz,
- Samsung,
Virtual memory: D:\pagefile.sys (allocated base size 32768MB)

Memory SPD:

Graphics
Intel(R) Arc(TM) Graphics
Memory: 1024MB
Driver provider: Intel Corporation
Driver version: 31.0.101.4725
Driver date: 9-25-2023
Monitor 1: 2560x1600x32 120Hz 144 DPI (Primary monitor)

Disk volumes
C: Local Drive, \\?\Volume{9480d9d2-20a6-4fd7-8abf-a5aeb49b2e8d}\, Windows, NTFS, (200.49GB total, 157.57GB free)
D: Local Drive, \\?\Volume{fd596efe-ba13-445e-8b10-24d871289cf8}\, Data, NTFS, (752.00GB total, 719.88GB free)

Disk drives
磁盘驱动器: Model: SCY S5000 1TB Serial: DW230810A000031 (Disk: 0, SSD, Size: 953.86GB, Interface: NVMe, Volumes: C D)

Optical drives

Network
Intel(R) Wi-Fi 6 AX201 160MHz (Speed: 866Mb/s) (MAC: DC:46:28:9A:0E:57) (IPv4: 192.168.5.59) (IPv6: fe80::f8ad:bf78:3d6d:817d)
Bluetooth Device (Personal Area Network) (Speed: 3Mb/s) (MAC: DC:46:28:9A:0E:5B)

Ports
Keyboard Port: PS/2 connector
Mouse Port: PS/2 connector

USB
符合 USB xHCI 的主机控制器
符合 USB xHCI 的主机控制器
- Rapoo Rapoo Gaming Mouse
- Blestech Betterlife Fingerprint (SN: 1.3.16.0)
- (R) Bluetooth(R)


**************
RESULT SUMMARY
**************
Test Start time: Tue Nov 14 17:17:26 2023
Test Stop time: Wed Nov 15 08:45:51 2023
Test Duration: 015h 28m 25s
Temperature HDD 0 (SCY S5000 1TB) (Min/Current/Max): 0.0C / 25.0C / 32.0C
Disk (C Speeds MB/s (Read/Write) 1209.7 / 1682.3

Test Name Cycles Operations Result Errors Last Error
CPU 3570 731 Trillion PASS 0 No errors
GPGPU 34 4.016 Trillion FAIL 1 No operations reported in timeout period
Memory (RAM) 11461 451 Trillion PASS 0 No errors
2D Graphics 129 117 Billion FAIL 497852 Error writing to GPU memory
3D Graphics 52 1.564 Million FAIL 1 An error occured during the DX12 3D test
Sound 496 1.106 Billion PASS 0 No errors
Video Playback 4246 54354 PASS 0 No errors
Disk (C 741 3.386 Trillion PASS 0 No errors

[LiCheng]

Test Name Cycles Operations Result Errors Last Error
CPU 94 25.425 Trillion PASS 0 No errors
GPGPU 47 257 Trillion FAIL 1 GPGPU unable to access graphics resource required for verification
Memory (RAM) 269 10.596 Trillion PASS 0 No errors
2D Graphics 101 92.523 Billion FAIL 51 Error writing to GPU memory
3D Graphics 1 52319 FAIL 1 An error occured during the DX12 3D test
Sound 12 28.224 Million PASS 0 No errors
Video Playback 100 1290 PASS 0 No errors
Disk (C 118 529 Billion PASS 0 No errors
TEST RUN FAILED

[Richard (PassMark)]

If you have the app Microsoft APP OpenCL™ and OpenGLŪ Compatibility Pac installed, you will need to disable it. The issue is with Microsoft APP OpenCL™ and OpenGLŪ Compatibility Pack, the app is an emulation layer to run OpenCL code through DirectX 12 supported hardware on system that may not have OpenCL installed. However, if OpenCL is supported on the hardware natively, and the App is installed as well, it can cause duplicates entries when querying for available OpenCL devices.

BurnInTest runs a OpenCL tests on all reported OpenCL supported hardware. So, it is out of resources because they all likely share the same GPU memory and each instance of 2D/GPGPU test requires certain amount of memory to run. To prevent the error, you'll need to disable the Microsoft APP OpenCL™ and OpenGLŪ Compatibility Pack.

There may be more information in the Level 2 Trace Logs or by running BurnInTest in Debug mode.

[LiCheng]

Hi Richard,
We can find there is display TDR in event view. We use intel intergrated VGA for this project. We test the items with 100% loading. Is there any information about the failure and TDR? Debug mode log also uploaded.

[David (PassMark)]

We test the items with 100% loading

See this FAQ
https://www.passmark.com/support/bur...uty_cycles.php