I've read all of the discussions regarding operating the compute stick headless, and have nearly (but not quite) achieved a state where it is operating adequately.
First off, I have the Linux version of the device, so the discussion pertains *strictly* to the linux version and as it ships.
Second, I have TWO units, and they exhibit identical behavior, so it isn't a defective unit.
Since the device is marketed towards connecting to a television, it is probably important for it to actually *work* on a television. I've tried it on TWO televisions, one a 48" Toshiba, another about 24" Vizio. The compute stick fails in the same manner on both. And that is, if the television is unplugged, or powered off, or even just switched to a different INPUT when the compute stick is powered on, then the compute stick will fail to boot.
I have what appears to be the latest bios referenced as fixing problems operating headless, bios version is reported as FCBYT10H.86A.0024.2015.0522.1757
I've managed to *partly* fix the problem by adding kernel parameters as "video=HDMI-1:e drm_kms_helper.edid_firmware=edid/1920x1080.bin"
To explain those parameters, the first one is supposed to switch on HDMI-1 output, even if it is detected as not-connected. The second one feeds it a default edid for 1920x1080 resolution, matching most televisions.
The problem is that the first one doesn't seem to actually do anything. I'm not sure if its because it isn't booting all the way to the kernel, or if the bios is setting it up so that it won't output even if we force the output on.
What the kernel parameter *did* fix, are the cases where the television is plugged in but turned off, or turned on to the wrong input. I.e., cases where the compute stick can detect that the television is *present*, but the television does not communicate an EDID. So it is able to ignore that and set up a proper output mode.
This state would be fine, except that the television has to be plugged in for a certain amount of time (whether it is on or off is irrelevant, its probably around 5 seconds) before the compute stick is able to detect that it is present, and that time is longer than the time between the compute stick being plugged in, and it testing if the television is present.
WHICH MEANS, that it cannot recover following a power failure, since the monitor and compute stick will be powered up at the same moment.
Possibly related, I also observe that on certain occasions, it will also fail to boot when the television is plugged in, turned on, and even switched to the proper input, until I plug in a keyboard and hit ctrl-alt-delete. It happens *consistently* when exiting from the bios setup mode. The failure is NOT cleared by unplugging and plugging back in. The failure is NOT cleared by powering off and on with the power button. It is ONLY cleared by ctrl-alt-delete. This also occurs sometimes after an attempt to boot headless. Being connected to a television, it is very inconvenient to have to plug in a USB keyboard since it is hard to reach -- I control the units over wifi and bluetooth.
- Intel® Makers
Thank you for contacting the Intel® Communities.
I would like to know, what is the software that you are currently using to remotely use the Intel® Compute Sticks?
I look forward to your reply.
You are focusing on irrelevant details and clearly misreading. The device does not boot. I am *NOT* requiring remote access -- that is not the objective.
When it does boot, it does so in about 30 seconds.
In order for the device to be deemed as working correctly, it should be possible to plug in ONLY the power, wait 30 seconds for a full boot, then plug the HDMI in, and observe output on the monitor consistent with a fully booted system. That behavior should be possible with the kernel parameters I have indicated, but it does not work.
Thank you for the clarification on this.
I performed the test mentioned.
Connected the unit to the power only.
-Waited around a minute>plugged the device to a HDMI port in a monitor.
-Then the main screen of Ubuntu 14.04 appeared.
-BIOS version in the unit: 0031
If you are not being able to perform this, and the unit is with all the default settings it came with, you can check the warranty of the unit.
Note: If you purchase the unit less than 30 days ago, the warranty would be handled by the place of purchase.
If not, contact Intel Customer Support (http://www.intel.com/content/www/us/en/support/contact-support.html Contact Support)
If you require any further information, feel free add it in this thread.
Like I said, I have *TWO* units, and am testing by connecting to *TWO* different monitor/televisions, so this is *not* a defective unit problem.
I don't know what "bios version 0031" refers to.
I have indicated that the bios version showing on both of my units is "FCBYT10H.86A.0024.2015.0522.1757".
Ok, that helped. A lot. Thank you.
It does start up now, even when unplugged.
One flaw remains, not a deal breaker though... when it starts up without the HDMI plugged in, the sound doesn't initialize. It thinks that the sound output device is something it calls "dummy". Presumably the sound equivalent to /dev/null. I'd guess that this would be on the OS side of things, and maybe something simple like restarting pulseaudio when the HDMI hooks up might fix it.
At least the device is actually online now though, which means I can just reboot it rather than having to replug it.
Looks like the Intel alsa driver probes the hardware to see what is and isn't connected, so when the HDMI is unplugged, it figures that the sound is also unplugged. The good news is that it looks like there are overrides available for it, so I can probably muddle my way through forcing it to turn the sound on anyway, or reprobing it when the HDMI is plugged in. Reprobing the sound on an HDMI plug event is probably the way is *should* be implemented.
That is correct, also make sure the audio plugins are installed in your system.