Inference error while using Multiple NCS2 devices with RaspberryPi 4

[Krbism]

Greetings everyone,

I'm working on a project where I'm required to run three different deep learning models concurrently where each of models take frames from a live video feed as input. I'm using the Rpi 4 for post processing and some control system jobs and I'm using the NCS2 stick for the models' inferencing. Since I couldn't fit more than one model in a single NCS2's memory, I decided to inference them independently on 3 NCS devices. [Note: The 3 DL models that I'm using function independently] I'm loading each model on a separate NCS and using them simultaneously.

The three NCS2 sticks are connected to the Rpi4. 

In this setup, I'm noticing that after ~20mins, one of the NCS sticks runs into the error as shown below.

Is this an error regarding the power supply to the NCS? or does the NCS has a tendency to crash when used for longer duration of time.  

Any help into understanding the reason behind this error would be a great help.

----------------------------------------------------------------------------------------------

E: [global] 785959] [Scheduler01Thr] dispatcherEventSend:54 Write failed (header) (err -4) | event XLINK_WRITE_REQ

E: [xLink] I

785959] [Scheduler81Thr] sendEvents: 1150

E: [global] I

Event sending failed

785959] [Scheduler01Thr] dispatchereventsend:54 write failed (header) (err -4) I event XLINK_WRITE_REQ

E: [xLink]

735959] [Scheduler91Thr] sendEvents:1150

[global]

Event sending failed

E: [global]

735960] [python3] addEvent:359 Condition failed: event-›header.flags.bitField.ack != 1

E: [global]

785960] [python3] addEventWithPerf:373 addEvent(event, timeouts) nethod call failed with an error: 3

7859691

Ipython3] XLinkReadData: 156

Condition failed: (addEventlithPerf(&event, Goptine,

E: INCAPIT

735960] [python3] tGraphMonitorResponseValue:1901

XLink error, re: X_LINK_ERROR

OXFFFFFFFF))

E: [nCAPI]

7859601

[python3]

ncGraphQueueInference:3453

Can't get trigger response

[Peh_Intel]

Hi Krbism,

Thanks for reaching out to us.

Yes, you are right. It might be due to a power delivery issue.

I would like to share this previous discussion that the similar issue was solved by unplugging all other USB devices except of the NCS2.

Regards,

Peh

[Krbism]

Hi Peh,

Thank you for your response.

I tried disconnecting other devices from the USB port and kept only the 3 NCS devices plugged to the Rpi4. But still I seem to get the same error after some inferences. 

Is there any other reason behind this?

[Peh_Intel]

Hi Krbism,

Could you share the files (models, inference application) with us, if possible, for us to replicate on our side?

Regards,

Peh

[Peh_Intel]

Hi Krbism,

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.

Regards,

Peh