STK1AW32SC BIOS v0029 Failure

[idata]

I am a digital signage provider which standardizes around use of Intel NUC and Compute Stick hardware. We recently sent out an order of compute sticks with BIOS v0029 and we are getting reports that the sticks are not "powering on" when there is a loss of power and BIOS is set to always on. I am able to recreate the behavior in my office by leaving a stick on for several hours and rebooting. The blue light comes on but the device does not post. When this happens the fan does not run and the screen is dead. The solution is to remove power and reapply it -- several times until the device finally posts -- which means the boot screen is shown and fan starts.

I need to know what happened and if this is a known issue. I can call to ask that we RMA 26 Intel Compute Sticks -- but what would be better is us establishing if there is a known issue with V0029 BIOS and whether BIOS v0030 specifically resolves the behavior we are seeing.

Thanks,

Shaun

[n_scott_pearson]

I suggest that you call Intel Customer Support directly and open a ticket regarding this issue...

...S

[idata]

Hello ShaunT,

I have not seen similar issue with this unit, I'm not saying that the issue is not there but we have not gotten this report before, have you tried BIOS 0030 and try to recreate the behavior at your office? I'm asking just to make sure. Thanks, Ivan.

[idata]

Thanks Ivan for the help. I now believe it may be thermal related. If I put one of the affected devices in a refrigerator to cool it down for five minutes, and then reconnect it boots fine. So the question is now -- is there a thermal regulation process that prevents the hardware from POSTing? None of these compute sticks exhibited this behavior during testing and burn-in at our shop. After we mailed them out to a client they started failing after 1-3 weeks. They are being used to play 1080 HD video in a loop. I haven't received all of the hardware back yet -- it is coming in waves. After artificially 'cooling' the device I can update the BIOS to v30 but the behavior is unchanged. I am going to keep testing to see if it is thermal related and produce a video and reproduction steps if I can.

[idata]

Thank you ShaunT for this information

Could you please tell us how you have the unit connected, is it connected with the power adapter or through some other connection method? I will appreciate if you can explain to us how you have it set up. Regards, Ivan.

[idata]

Hi Ivan -- I am using the power adapter that came in the box. (I'm aware of the issues with using TV USB for power or non-standard adapters). Interestingly, it does not appear to be thermal related. I have a video that I'll be able to share shortly.

[idata]

OK, here is a video of the behavior: https://youtu.be/GJZvUlr_-cg Intel Compute Stick - Will it Post? - YouTube

I believe it is not thermal related.

[idata]

Thank you for this information and for the video you have kindly share with us.

I will share this information you have provided to further investigate the issue. Best wishes, Ivan

[idata]

Thanks Ivan. Just to clarify the following:

• The players in the field are on v29.
• The player in the video was on v29 this morning and I upgrade it to v30. I see no difference in behavior between v29 and v30.
• Yes, I am using the supplied power adapter.
• The Compute Stick in the video was manufactured 08 JUN 2016. S/N GESC622009VZ
• I am seeing failures in approximately 20 out of 30 units. I don't have the exact number yet because I am receiving shipments back from the client in groups. They are only sending back broken ones right now.
• The players loop HD video 24/7. Everything works fine during initial setup and for approximately 1 to 3 weeks after they enter into use.
• When they "fail" in production the screen is black. Attempts to power cycle the stick result in the behavior shown in the video above.

If you have any additional questions please ask.

Thanks,

shaun

[idata]

Thank you for all this additional information you have kindly provided us.

After I shared this information with my co-workers, one of them had a case similar to this one; the issue was resolved when the customer downgraded the BIOS to version https://downloadcenter.intel.com/download/26270/BIOS-Update-SCCHTAX5-86A-?product=91065 0026. Would you mind trying BIOS 0026 and see how it goes? Since you will be downgrading your BIOS, I suggest you to do a BIOS recovery to that BIOS version, please follow http://www.intel.com/content/www/us/en/support/boards-and-kits/intel-compute-stick/000005902.html BIOS Recovery Instructions for Intel® Compute Stick Ivan.

[idata]

Thanks Ivan. I downgraded to v26 and there is no change from the video shown above. I have now tried the following BIOS:

• v29 (Original)
• v30 - No Change
• v26 - No Change

[idata]

Thank you for trying that, I have sent you a private message.

Ivan.

[idata]

I am working with Intel Support and I wanted to share my current understanding on what I think the cause may be. Here is a clue -- the Compute Stick shown in the video above failed to boot after repeated attempts. However, I came in this morning and it booted reliably. This behavior is highly suggestive of an environmental cause for the behavior documented in the video above. The evidence points to a thermal issue.

To test this theory I cranked the heat in my office to 85F and ran HD video in a tight loop on the Compute Stick. I used a Microsoft UWP app to do this -- which means both the GPU and CPU is used. After running the player for two days I disconnected and reconnected the power cable. Guess what happened? It failed to boot!

Here are the reproduction steps:

1) Operate the compute stick in a room with an ambient temperature of 85F or warmer.

2) Run content that is both GPU and CPU intensive, such as a series of HD videos in a loop

3) Wait 8 to 24 Hours

4) Unplug Compute Stick and plug back in.

5) ... It will fail to boot.

These details explain why our players work in an office or lab setting but fail when they are deployed in the field. The space behind a television is usually warm -- 80F to 90F. It takes hours or even days for the unit to internally heat up. Once power is lost and restored the compute stick boot process will not post until internal temperatures on the unit decreases AND a day passes.

Since Intel Compute Stick's are marketed as solutions for digital signage I have two thoughts here...

1) The Intel Compute Stick has a physical design flaw which prevents adequate cooling in warm to hot ambient environments.

2) Or... The Intel Compute Stick's BIOS and/or firmware has an overly stringent thermal envelope for allowing operation. This threshold is more stringent during the boot phase of the device compared to the run time phase.

[idata]

Thank you ShaunT for sharing this valuable information with us, I suggest to send this information to Intel Support as well for them to be aware. Also any findings that you can share on this thread it will be very much appreciated.

Best regards, Ivan.

[idata]

Thanks Ivan. The original stick shown in the video now magically works -- i.e. it will post. A different stick (which I have sent to Intel) consistently does not work. There seems to be three different scenarios:

1) The blue light is on and the stick won't post or there is no display. It stays broken. (This is what I sent Intel)

2) The blue light is on and the stick won't post -- but if you try unplugging and plugging in several times, eventually it comes back.

3) The blue light is on and the stick won't post -- but eventually for some unknown reason it starts working again and posts normally.

I thought this was thermally related -- but I can't get the "failure" scenario to occur consistently. I am at a complete loss. I give up.

[idata]

I mailed off three sticks to Hillsboro, Oregon. I included the following letter within:

I received a shipment of five sticks back from a client (they are rounding up all thirty of them now and they are coming back in waves). From this batch of five 2 were immediately dead with the "solid blue light but no post" behavior and the remaining three booted fine for me. Of the three which worked I could not get them to fail.

I am sending you the two bad sticks for your investigation and one of the sticks which appears to work but is reported to have failed at the client site with the blue light on behavior.

SA# : H93326-102. This one powers on with a blue light (non-blinking) but won't post. Out of thirty sticks this is what happened to approximately 20 out of 30 of them either permanently or intermittently. If you disconnect power, wait, and then turn it back on they will sometimes "come back". This one seems permanently stuck in a bad state or I am just very unlucky and I can't get it to boot. This one may be your best bet at reproducing the behavior since it appears stuck in a bad state.

SA# : H93326-101. This is my development stick. It is the one shown in the video. It worked fine until I decided to leave it on for a week to see if it would fail like those in the field. It failed with the same

failure signature and it wouldn't boot except for about 1 in 20 boots as shown in the video. During one of the successful boots I upgraded the bios to v30. It still would only boot about 1 in 20 tries (similar to video) after I cycled the power connection. I downgraded to v26 per guidance from support and the problem remained. I came back two days later (to mail this one to you) and I tried rebooting it. ... and it works fine now but I have no idea why.

SA# H93326-102. This one is like a rock. It doesn't fail and it works fine. It has our signage software on it and you may try it out to see how things work normally.

Please note that all three compute sticks have 128GB MicroSD cards inserted. I felt that if I removed themit could alter our test and result in a different outcome.

Thank you!

[idata]

Thank you ShaunT for that information, hopefully you will get some results some from Hillsboro.

Best regards, Ivan.