[Ben]

“ME” is the Intel Management Engine, which is basically a hardware-based spyware.  Useful for companies to keep track of their electronics and make sure no one is installing pirated software on company-owned devices, but for an end-user it is just one more thing that might be accessed remotely.

“ME FW update failed” means that something is trying to change its behavior… maybe as innocent as a Windows update, maybe malicious, no way for you to tell at this level.

If you don’t have a use for the management engine, protect your privacy by turning it off in the BIOS settings.

[Ben]

The tablet has a combination headphone/microphone jack (TRRS style).  Unless you have a headset/mic combo cable, it is likely that your microphone cable has a mono plug and is contacting the wrong circuit in the jack.  (The jack is designed so that a stereo headphone being plugged in will “just work”, by putting the microphone on the additional contact)

 

You can buy a TRRS <-> 2x TRS adapter cable that will make the right connection between your microphone cable and the tablet.

Example: https://www.amazon.com/SuperWhole-Earphone-Headset-Microphone-Adapter/dp/B0114AB0QW/

[Ben]

I did a clean install of Windows 10 build 1607 (the new one with enhanced Ink features).  Before formatting the old install, I ran the Export-WindowsDriver command in PowerShell to make sure I could recover any drivers that aren’t available for download.

 

Export-WindowsDriver -Online -Destination d:\drivers

 

After that, it’s attaching a USB hub through the included OTG cable, so you can have both the install flash disk and a wired keyboard hooked up simultaneously.  Hit “Delete” to get into the BIOS, change the boot order, boot from the USB stick.  You can use the https://wudt.codeplex.com to make the USB bootable, it says Windows 7/8 but also works fine with 10.

I did have some problems during the “Preparing files” phase of the install due to data corruption, but that was likely just the particular USB drive I was using which got overheated.  A couple retries later, it went through fine.

SSD, Touchscreen (both display and touch input) and wifi worked out of the box, which means you don’t actually have to inject drivers into the install image, setting them up later is perfectly feasible.

The following drivers I installed using download from the websites of the companies making those parts: Wacom driver (the download for “Tablet PC”), Intel GPU, Intel WiFi, Intel Bluetooth.

Notably, the Wacom driver captured by Export-WindowsDriver installed fine but didn’t result in a working digitizer.  Use the proper installer from the Wacom website instead.

Then I went through in Device Manager and for every unrecognized device, pointed it at the directory with the drivers from Windows-ExportDriver with “search subdirectories” enabled.

Audio still wasn’t working, so I did the same for the Realtek audio device, even though it wasn’t unrecognized, the Windows 10-supplied driver didn’t work and the archived driver did.

 

It seems like a success, I now have the Windows 10 edition of my choice and no worries about some SuperFish-style exploit placed there by anyone along the supply chain.

Now I need a thermal mod and a better pen than the tiny one borrowed from my old Galaxy Note 2.

[Ben]

Chris G wrote:

Hi, this kind of mod always increases surface temps. Internal CPU max temp dropped 20 degrees, but the outer housing when gaming or encoding video gets very hot to the touch. This is the trade-off in transferring heat to the rear housing. I use HWinfo to monitor CPU & GPU temps.

 

This is ignorant of thermodynamics.

The CPU is generating heat at a fixed rate.  Under steady-state conditions, all the various junctions (die – package, package – heatspreader, heatspreader – back cover, back cover – surroundings) must be carrying heat outward at the same rate at which it is generated.  The heat lost from the back cover depends on the temperature difference between it and the air or desk against it.  So that won’t change.  If anything, because adding a thermal pad transfers heat from the CPU package to a greater area of the back cover, the cover temperature ought to be more even, with hotter cold spots and cooler hotspots.

Knowing that the back cover temperature stays the same explains why the internal temps decrease.  Because the conductivity between heatspreader and cover is better, the temperature difference doesn’t need to be as great.  The difference between the die and package and heatspreader isn’t changing, so they drop the same amount.

Now, if you aren’t running at the same CPU load, because the CPU is no longer throttling, then the heat generation will increase, and then the temperature difference between cover and environment must increase also.  But that does not mean that the mod made the back cover hotter, it means that driving the CPU harder did.  You might equally well say that the thermal mod decreases battery runtime, which is also an expected result of reducing throttling.

What will happen is that the back cover will heat up to its steady-state temperature more quickly.  But that steady-state won’t be any hotter than before unless the CPU is running at a higher average clock frequency.

[Ben]

maus wrote:

The bad news: I noticed that the battery discarging rate that the tablet provides is not accurate at all, it seems to be only about 82% of the real discharging rate! For example, during the video test with brightness set at 25%, BatteryBar told that the battery was discharging at about -4.600 mW, but the real discharging rate was -5.600 mW. In fact, if you divide the battery capacity (about 41 Wh) for 5.600, you get about 7, that is the number of hours you can squeeze from the battery in that scenario. And it also seems that Windows 10 calculates its estimation of the battery runtime by using the average discharging rate, so if your battery is going to last 7 hours, it will tell you that you have about 10 hours left! That’s pretty annoying, I don’t know if it can be fixed by software (maybe with a BIOS update) or if it is an hardware issue.

 

Good observation but wrong conclusion.  Actually the capacity of the battery is nowhere near 41 Wh (reality: 4200 mAh * 7.6V = 32 Wh).  Use the correct capacity in the equation, and now you see that the discharge rate of 4600 mW matches closely with your observed runtime of about 7 hours.

It would be very surprising if the discharge rate were incorrect, since that is actually *measured*.  BatteryBar should eventually recalibrate the capacity based on actual power released and replace that ridiculous 41 Wh estimate by the real value.

(For those who care — a Li-ion cell ranges from about 4.3V at full charge to 3.6V at about 2% charge remaining, with two batteries in series the voltage is twice that.  The voltage also doesn’t decrease linearly with energy remaining, which results in the “average” voltage being a somewhat below the midpoint between 3.6 and 4.3V.)

[Author]

Posts