Exact temperature sensor locations for A31 series found

[Milos]

Hello all,

I am sensitive to fan noise and have been using Shimodax's tpfancontrol and NHC fan control on an A31 without clue which temp sensor means what. So, I finally had to take my laptop apart, take freeze spray to the motherboard and find (fairly exactly) all the temperature sensors, so we know which is which.

Here they are, and their corresponding EC addresses:

Code: Select all

Label     Address     Typical Temp        Notes
CPU        0x78         36-40		CPU for sure
BAT1a      0x79         32		   Battery (main) - heats up when laptop on battery power, cool when AC
PWR        0x7a         40-46		Power diodes   - sensor left of CPU, top side
BAT2a      0x7b                    (probably) Ultrabay 2000 Battery (main?)
BUS        0x7c         48         Northbridge - sensor below and left of Northbridge, bottom of MB
PCM        0x7d         25         PCM/ambient - near VGA 9pin plug (LM75 single temp sensor chip); seems always lowest temp (room temp)
BAT1b      0x7e         31         Battery (ambient) - 0 when battery removed, but doesn't heat up under battery load
BAT2b      0x7f                    (probably) Ultrabay 2000 battery (ambient?)
-          0xc0           0        none of these (below) have any values
-          0xc1           0            "
-          0xc2           0            "
-          0xc3           0            "

I have added this information and photos of the top and bottom of the motherboard tested with marked sensor locations onto the Linux Thinkpad wiki, http://thinkwiki.org/wiki/Thermal_Sensors.

Now that the sensor labels are right, I'll upload an NHC control script I have been using, which allows fan control inside NHC.

Hope this helps. One curious fact is that the thermal sensor chips appear to have more or less total ports than the temps identified depending on whether you assume batteries are measured by them or not (CPU tells its own?). So if I didn't miscount it's possible that some temperatures may be hidden elsewhere in the EC memory.. Anyone any ideas?

Milos

[Milos]

Ok, I just uploaded a fan control file for NHC that I've been using, with the correct temperature sensor labels, here: http://nhc.justred.info/. Fan RPM not shown properly due to NHC issue soon to be fixed (as discussed in forums), but control works ok.

This script reproduces Shimodax's tpfancontrol utility in NHC. I see a script done by Spawn has better features, so I may try to copy those in later. Anyone else feel free.

Milos

[christopher_wolf]

Awesome. That is one Thinkpad model down, more and more to go.

[Milos]

Sadly, in searching for the temp sensors, I found that the hottest parts (Northbridge/GPU and "power diodes" sensors) are hotter than the CPU, yet the CPU is the only thing attached to the heat sink. The Northbridge and GPU have those pink thermal pads and contact some thin aluminum on the bottom of the machine which seem not so great.. no real heat sink. So really to keep the thing quiet someone would have to retrofit a decent heatsink.. e.g. make a copper piece that replaces the memory bay cover, has fins on bottom and reaches in to cover the Northbridge and diodes (from underneath)..

Milos

[christopher_wolf]

Yeah, I seem to remember one forum member actually did something like that to his T4X Series machine with copper but it was from the top. Strange how everybody seems to think that the CPU will always be the hottest component and yet it turns out to be either one of the bridges or GPU after all. Even the engineers at the time seemed to be more concerned with the CPU enough so to put a major heatsink on it.


A passive system might cutit for awhile, but if there is enough heat in a short enough period of time, it will spread to the rest of the system quickly and then make the fan kick in...again. Suprising when you need something like a really efficient Peltier cooler with good airflow, isn't it?:lol:

Good Luck with your Thesis by the way.

[Milos]

Yeah, I saw that earlier thread, I was thinking basically something like that but more solid. It's really annoying that all these non-heatsinked parts are the ones cooking. A decent heatpipe to a little heatsink radiator that's part of the case might hopefully delocalize the heat enough to spread it so at least on low use the fan doesn't go on..

Re: Peltier coolers, dude, that would be awesome, but I think they're only about 15% efficient or so . Of course they're just a "heat pipe", so you still need a heatsink somewhere with a design able to radiate all the extracted heat to the outside... anyway, not this year..

Thanks for the good wishes, better get back to that..

[Thinkerer]

A decent heatpipe to a little heatsink radiator that's part of the case might hopefully delocalize the heat enough to spread it so at least on low use the fan doesn't go on..

If you're OK with a low fan speed but don't want the fan to step up to loud levels, there's another approach: pipe the heat internally into a small radiator on the machine's air intake path. On the T4x, this means a thermal connection to the PCMCIA slot.

Advantages: No externally visible modding. No hot radiating surfaces on your lap. Unaffected by other hot components.

I guess the reason the thermal designers care mostly about CPU (and sometimes GPU) is that, while its idle temperature may be lower than other components due to aggressive power management, its 100%-use temperature is by far the highest.

[Milos]

A decent heatpipe to a little heatsink radiator that's part of the case might hopefully delocalize the heat enough to spread it so at least on low use the fan doesn't go on..

If you're OK with a low fan speed but don't want the fan to step up to loud levels, there's another approach: pipe the heat internally into a small radiator on the machine's air intake path. On the T4x, this means a thermal connection to the PCMCIA slot.

Yeah, I thought about this as a simpler alternative, but unfortunately 1) the fan in an (or my?) A31 has no "low" speed setting, it can go 0, 3200 or 3400 RPM (or 4000 RPM on disengaged mode); 2) half the hot parts are on the bottom of the motherboard so it would be hard to wrap a copper heatpipe around to the other side to thermally join to the CPU heatsink and heat exchanger. The RAM slot was closer on the bottom, so I was thinking it would be easiest to go there (e.g. machine a copper ram slot cover with ribs - dreaming of course, not happening any time soon), but not sure if that would actually work well enough since there's no airflow except passive.

Re: crappy fan control above, this is why I'm very interested in mg's thread on the EC and your suggestion to look into whether the fan is directly PWM controlled from the EC (and that it might be "editable")..

Advantages: No externally visible modding. No hot radiating surfaces on your lap. Unaffected by other hot components.

What do you mean by unaffected by other hot components? Couldn't read from context..

I guess the reason the thermal designers care mostly about CPU (and sometimes GPU) is that, while its idle temperature may be lower than other components due to aggressive power management, its 100%-use temperature is by far the highest.

True, though not totally in my case -- the CPU is the hottest, but the power supply thmal sensor especially as well as the Northbridge thmal sensor also get pretty cooking. And the problem is that when the CPU is done it cools down in 2 seconds thanks to the heatsink, while the fan keeps going at full blast while the other two cool on their own with a sloow time constant - because the fan is trying to cool but has no impact.

Milos

[Thinkerer]

The RAM slot was closer on the bottom, so I was thinking it would be easiest to go there (e.g. machine a copper ram slot cover with ribs - dreaming of course, not happening any time soon), but not sure if that would actually work well enough since there's no airflow except passive.

FWIW, the hot region under the T43's touchpad seems to rely on passive cooling through the bottom too. There's something looking suspiciously like a thermal pad visible through one of the holes in the bottom, and blocking that area (e.g., by placing it on your lap) raises temperatures be 3-4 degrees pretty quickly.

Re: crappy fan control above, this is why I'm very interested in mg's thread on the EC and your suggestion to look into whether the fan is directly PWM controlled from the EC (and that it might be "editable")..

We know it *can* be done in software control for some older models. But it was removed in later model's firmware. See the ibm_acpi patches posted to the linux-thinkpad list recently, they encode this knowledge.

Advantages: No externally visible modding. No hot radiating surfaces on your lap. Unaffected by other hot components.

What do you mean by unaffected by other hot components? Couldn't read from context..

There was this other thread here about binding the chipset to the CPU or GPU heatsink. That's great for idle, but when the CPU heats up under 100% CPU it will carry the chipset with it, perhaps out of comfortable range (the CPU is rated higher than the chipset, and better isolated from your lap). This effect is very visible with the CPU and GPU on the T43, which share a common fan radiator -- once things stabilize they're never more than a few degrees apart, even if one is idle and the other going full throttle.

True, though not totally in my case -- the CPU is the hottest, but the power supply thmal sensor especially as well as the Northbridge thmal sensor also get pretty cooking. And the problem is that when the CPU is done it cools down in 2 seconds thanks to the heatsink, while the fan keeps going at full blast while the other two cool on their own with a sloow time constant - because the fan is trying to cool but has no impact.

Evidently, the ThinkPad's engineers cared very little about minimizing fan action. The goal was apparently to prevent hardware failures and safety hazards. A constant moderately high temperature, and continuously on fan, weren't a concern.