TC4400, Power led remains on with intel N 6230

[kvas]

I just received a mini pcie Intel N 6230 and with it installed the power led and the caps lock led remain lit when the tablet is turned off.
With the card removed or the older intel 3965 installed there is no problem.
Wireless works great on Windows 7, reaching close to 11Mbytes/sec on file transfer.
Another problem is that the BT of the card is not even showing in device manager but I can live without it.
Anyone have any ideas for fixing it or suggestions for a tried out alternative wifi card?

[Hattori Hanzo]

Could you do me a favor?
Shut down the laptop, disconnect all batteries and then put the main battery in again. What happens?

You should also notice that your battery gets drained at about 5% per hour in that not really off state, no matter wether Wake on LAN is enabled or not.
Have you also noticed display artifacts after long times in a suspended or shut down state with the lighting LEDs?

[kvas]

With the 6230 installed, when you remove and put back the battery, the charge led, the power led and the caps lock led light. After around 5 secs the charge led stops.
With the old card or no card, after putting back the battery, only the charge led lights for 1 sec.

[Hattori Hanzo]

That's interesting.
I'm using a Wifi Link 5100 and for me it's only the power LEDs (both) that maintain a certain glow. If you suspend it, you still see the flashing as a brightness jump.
Another funny thing is that if I insert the battery it kind of switches on, screen staying black, fan running. I can shut it off by keeping the power switch pressed, but switching it on into a usable state is impossible. I have to insert the battery while running and connected to AC.

I think this must have something to do with the BIOS, either my mods to whitelisting or some flaw that was originally present.

[kvas]

Happy update
As you had a similar problem, I thought something must be shorting out.
Got an idea to look at the mini pcie pinout and saw the many reserved pins.
Started covering with electrical tape the pins in small section and after some tries the leds stopped.
Right now I have covered with tape pins 41,43,45,47,49,51 and everything works as expected.Full speed wireless,on/off wifi key and no leds when shutdown.
Will see how it works the next days and I have to still see if I can enable the BT.

[Hattori Hanzo]

That's what I wanted to try out next, too, but haven't yet found the time to do. I have a strange misbehavior with my screen as a side effect. When the TC is connected to power and being "off" or at sleep for some time, I have sort of a display burn in from the edges. It goes from single lines of subpixels being overly bright to the whole outer area. Strangely complete lines only occur at the left and right side, in the middle it's only less than an inch.
I gotta make a picture before fixing it. No one would ever possibly consider that the side effect of a wifi card

Thanks for trying this fix and listing the pin numbers
I'll follow you with this in some days when I got time again for tinkering.

Btw the link is slightly broken. For anyone else: PCI Express Mini Card (Mini PCIe) pinout and wiring @ pinoutsguide.com

[Hattori Hanzo]

Ok, just opened my TC and put some tape on the uneven pins from 37 to 51. Works a charm

[kvas]

Nice to hear about it.
Everything works here too, except BT.
Hattori, could you please tell me how you undervolted and if you see any difference.

[Hattori Hanzo]

First I'd try to see how low you can go with the CPU.
Get either RMclock (google the name) or CrystalCPUID (also google). Both are able to change the VID in software. CrystalCPUID is a bit easier for quick changes, but the voltages it displays are 0,1125V too high, which is no concern when you only do software based undervolting.
You also need something to stability test your settings. You don't want random crashes. Get Intel Burntest for that purpose (google again).

In your power plan set minimum and maximum CPU performance to 100%. Otherwise Windows will also change the P-states on its own and this interferes with those tools.
Then open CrystalCPUID, choose Function and the Intel SpeedStep management menu item. You have to tick a checkbox to enable changing voltages.
Switch to Intel Burntest and choose a custom stress level and set repetitions to something funnily high. For a quick start you can choose something like 256MB. The tool essentially uses Intel's linpack libraries to stress the CPU by making silly calculations. It displays the result for each pass and if this changes your CPU miscalculated something, so it is instable.
Run the test and switch back to CPUID. Set the voltage to 1.263 (highest one available) and lower it step by step after some successful passes. Sooner or later the test will fail or you'll get a bluescreen. Note the settings for multiplier and voltage and test the next lowest multiplier, starting with the voltage the next higher one crashed.

Once you're done you get a rough curve of how your CPU handles different clock speeds. Which multipliers to choose and how to make the laptop change between those depends on that curve. The higher end CPUs usually need a lot more voltage for the highest multipliers and hence should be kept off those as much as possibly, as they are inefficient. The mid level and low level CPUs of a series usually have a very flat voltage/clock curve and don't need to be kept off the higher multipliers.
E.g. the better one of my T7600s can do 1.66GHz (multi 10) at 0.925V absolutely stable. For the full 2.33GHz it needs 1.125V. While clock speed increase roughly increases power consumption linearly, voltage affects it a bit more than squared. So with the voltage for the highest multi it roughly needs 50% more power for the same work. Not to be preferred unless necessary.
On the other hand the T2300E that's now in my T421x only needs 0.1V more to be stable on its highest multiplier at 1.66GHz than it needs on its lowest. So making it clock up rather soon is beneficial in most cases.

Depending on how low you can go with the voltage it might make sense to drop the idea of undervolting. If you can barely get below what the CPU uses on its own the decrease in regulating granularity will likely use up a lot of the benefits you get. In addition all those undervolting tools use some CPU cycles for themselves you also need to compensate for. SpeedStep and the Windows low level driver have a big advantage here.

Once you are sure you wanna still do it, test the multipliers you wanna use with one or two steps more of voltage than they crashed or failed at and set IBT to as much memory as you can give it. Give it ten runs and obstruct the air vents enough to keep the temp in the nineties, which you can conveniently check with RMclock.
If the test fails or the system crashes, further increase voltage until it is stable. If you wanna be absolutely sure, double check with other torture test tools.
Repeat for all multipliers used.

For automating the undervolting both tools offer options. I prefer CrystalCPUID and call it from the Windows task planning with the options /resi /cq /hide, with admin rights of course. It only offers you three possible multipliers, though, which is imo enough as managing them with such tools is terribly crude anyways.


If you have a CPU that is able to run even at rather high clock speeds at the lowest voltage (0.95V, 1.063V in CPUID), you might feel like going even lower on voltages. That needs hardware modifications.
Voltage signaling for the CPU is done by several Voltage ID pins. Their signals form a binary number that is the voltage in mulltiples of 0,0125V and the resulting voltage is this number subtracted from 1.5V. So the lowest pin reduces the voltage by 0.0125V when it's high, the second lowest by 0.025V and so on.
You can not set them low by connecting them to ground, you can only set them high by connecting them to VCC or to a different VID pin that is high. If you interconnect two VID pins, you create a logic or, meaning that when one is high the other will be high, too. This way you lose 50% of all voltages you could signal.
Intel has nice documents detailing all that, the appropriate one for Merom based CPUs and the older Socket M used in the TC4400 should be this: ftp://download.intel.com/design/mobi...s/31407804.pdf

In the end the only really viable option is shorting VID3 and VID1 which sit at AF 4 and AF 5 in the pinout. Take a blade and cut a small V between the pin holes, then take about half a centimeter of a single copper strain from a Cat5 wire, bend it and position it to interconnect those pins.

The downside of this is that it affects even voltage combinations you are not being able to set via software as they're used for lower power states. So this is potentially unstable and hard to test. VID3 which only reduces 0.1V always worked for me, but VID4 has led to sometimes erratic behavior especially on resuming from standby.


The results, well, quite good. Some time ago I used a T5200 that worked between 0.8 and 0.95V. This way the absolute maximum power consumption of the entire laptop went down from about 55W to slightly below 30W. Those numbers are with Wifi transfers, ethernet transfers, max brightness, a graphic benchmark and IBT running. Practically power consumption never even spiked above 20W, even with wifi on. As a result the battery life was almost static and only influenced noticeably by the wifi switch and screen brightness. Whether I'd just read PDFs, hold a presentation or use OneNote intensively or even work with really demanding software was barely a difference.
Thanks to the at least thermally well designed case of the T4200 series the fan was not spinning about 95% of the time. If it was, it was the lowest level and only during summer, when it was really hot.

My current T7600 is not as extreme and has the potential to suck down the battery with its highest multiplier. For doing stuff that only midly tickles the CPU there's a definite increase in battery life, though. It's hard to guess but I'd estimate nearly an hour for things like extensive notetaking, which the CPU does on the middle multiplier (1.66GHz) at just 0.925V.
If there's nothing CPU intensive that makes it use the highest multiplier the fan only kicks in rarely on the lowest speed for some seconds. With stock settings it's on most of the time and occasionally going up to it's second speed level.


Hmm, this became quite long. Any questions left?

[NikolaTesla]

mine does this with the intel advanced n 6200 is it the same pin layout? i need some help with this ASAP please