R5 vs R3 problems on GPU object_detection_sample_ssd

[Stefano_M_]

Before R5 I was using R3.343 on Intel® Celeron® N3160 with Windows 10 Enterprise LTSB (version 1607 - Build: 14393.693)

Intel HD Graphics driver version: 20.19.15.4531 - 29/09/2016

object_detection_sample_ssd in R3.343 was working both in CPU and in GPU mode (only in FP32 precision)

this was the output for initialization:

[ INFO ] InferenceEngine:
        API version ............ 1.2
        Build .................. 13911
[ INFO ] Loading plugin:

        API version ............ 1.2
        Build .................. cldnn/GEN_GPU_clDNN_ci-main_cldnn-main-03988_artifacts.zip
        Description ....... clDNNPlugin
 

Now, with R5 (same S.O. and GPU drivers) the GPU mode is not working (both in FP32 and in FP16).

The sample gives me this error: [ ERROR ] Device lookup failed - unsupported device id: 0x22B1. Note: HD5xx+ devices are supported

this is the output on the initializazion:

[ INFO ] InferenceEngine:
        API version ............ 1.4
        Build .................. 19154
[ INFO ] Loading plugin

        API version ............ 1.5
        Build .................. 19154
        Description ....... clDNNPlugin

Is there something that I can do to run the sample with R5 in GPU mode?

Thanks

[nikos1]

Most likely similar issue to https://software.intel.com/en-us/forums/computer-vision/topic/800219

but could be also that was supported before and not supported now due to new clDNN requirements for newer OpenCL extensions.

Can you post clinfo output please?

[Stefano_M_]

How can I get clinfo output?

Thanks

nikos wrote:

Most likely similar issue to https://software.intel.com/en-us/forums/computer-vision/topic/800219

but could be also that was supported before and not supported now due to new clDNN requirements for newer OpenCL extensions.

Can you post clinfo output please?

[nikos1]

>  How can I get clinfo output?

On Ubuntu just install clinfo and run from command line.

On Windows it is a bit harder but please search for a binary or build from source.

A clinfo.exe binary for example is attached in https://boinc.mundayweb.com/wiki/index.php?title=Determine_OpenCL_capability_of_GPU_and_CPU