Does Inference engine support Intel Core i5-8500 with Intel UHD graphic 630?

[pham__hai]

Hi every one,

I am using opencv/dldt version 2019 to optimize my model.

I have executed sample object_detection_sample_ssd with CPU. But when trying to run it with GPU, I cannot find any IGPU device, as below error:

[ INFO ] InferenceEngine:
        API version ............ 1.6
        Build .................. custom_2019_c585b530c1b969930df61252057ccea2f72dfc76
Parsing input parameters
[ INFO ] Files were added: 1
[ INFO ]     D:\Develop\Surveillance\Test\000000509403.jpg
[ INFO ] Loading plugin

        API version ............ 1.6
        Build .................. custom_2019_c585b530c1b969930df61252057ccea2f72dfc76
        Description ....... clDNNPlugin
[ INFO ] Loading network files:
        D:\git\open_model_zoo\model_downloader\Retail\object_detection\pedestrian\rmnet_ssd\0013\dldt\person-detection-retail-0013.xml
        D:\git\open_model_zoo\model_downloader\Retail\object_detection\pedestrian\rmnet_ssd\0013\dldt\person-detection-retail-0013.bin
[ INFO ] Preparing input blobs
[ INFO ] Batch size is 1
[ INFO ] Preparing output blobs
[ INFO ] Loading model to the plugin
[ ERROR ] failed to create engine: No OpenCL device found which would match provided configuration:
    Intel(R) Core(TM) i5-8500 CPU @ 3.00GHz: invalid device type

I'm using windows 10 with core I5-8500 with Intel UHD graphic 630. I'm not sure Inference engine does not support my device or It cannot find the device.

Can anyone give me some advice?

Thank you!

[Shubha_R_Intel]

Dear pham, hai,

Please find your processor here:

http://mark.intel.com/

And OpenCL Runtime support here:

https://github.com/intel/compute-runtime/blob/master/README.md

Hardware

6th-8th Generation Intel® Core™

Intel® Xeon® v5 family

Intel® Xeon® v6 family

Intel® Pentium® processor N4200/5, N3350/5, N3450/5 with Intel® HD Graphics

Is mentioned LInux Installation Guide .

Also the System Requirements Doc is helpful.

Intel UHD graphic 630 is 8th Gen according to my research so it should work. 

I noticed that you are using an extremely old version of OpenVino. Can you kindly download OpenVino2019R2 and install the latest drivers and try again ?

Thanks,

Shubha

[pham__hai]

Hi, Shubha

Thank you for you helping.

I've updated Intel Graphic Driver and now I can run inference on Intel graphic. But it seem that the performance on GPU is slower than CPU.

Thank you very much!

[Shubha_R_Intel]

Dear pham, hai,

Glad to help ! Well, GPU can be faster than the CPU too. You can't generalize about one or the other. It really depends on the particular inference application you are running as well as your model.  Please experiment with different models using The benchmark app and then draw your conclusions.

Thanks for using OpenVino !

Shubha

[pham__hai]

Hi Shubha,

I tested Benchmark app with pre-trained model person-detection-retail-0013  on my computer as mentioned above.

The command is:

-i <path-to-image> -m <path-to-model>/person-detection-retail-0013.xml -d CPU

And out put result for CPU and GPU is as followings:

CPU
Latency: 34.32 ms
Throughput: 52.77 FPS

GPU/FP32
Latency: 66.68 ms
Throughput: 28.96 FPS

GPU/FP16
Latency: 63.29 ms
Throughput: 29.67 FPS

Regards,

[Shubha_R_Intel]

Dear pham, hai,

Thanks for trying out benchmark app. Your results certainly look reasonable to me ! I hope you agree ?

Thanks,

Shubha