M1530 - ACPI temps Mains vs on Batt

Hi All,

Here's an odd one I haven't found much info on...

When gaming, I noticed that the difference between my ACPI temps on mains VS on Battery are a bit of a concern...

For the first time last last night, it shut off on me which I guessed was the lovely overheating. Anyways I know my CPU and GPU temps are great and run very cool so it wasn't to do with GPU temps etc

The only difference when the machine shut off was I was gaming on Battery - So I ran some tests and found that there is a major difference between the ACPI temps when gaming on batt than when on mains.

Temps when gaming:

On Mains of ACPI (THM) = 71c max
On Battery of ACPI (9 cell batt) = 100c + !!!!!


For info on other temps on load:
CPU max = 60c
GPU max = 72c

It's nothing to do with positioning of laptop etc as I ran the test in the same position. I'm running windows XP Pro too.

So, what causes such a massive diff in temps of the ACPI chipset comparing mains to on battery ? This is like an instant change when you pull out the mains (under load), so I've narrowed it down to batt vs mains !

Any info appreciated !

Andrew

 

Correct me if im wrong, but doesnt the ACPI control the power of the laptop, meaning its the one that manages battery consumption, and also AC distribution.... so my guess is that the ACPI has a tougher job when operating on battery than when its operating on AC, hence the higher temps.

but if it shut off... then there IS a problem, how much time did it run before shutting down? because the battery can get drained in a matter of minutes when gaming, and you wont even notice the time that has come by...

 

Thanks for the info ! Good point - if the ACPI is working harder on batt then a possible reason for the sudden increase (if that is true on the way it's working).

It does however make the machine pretty unusable running on batt when gaming, now knowing this (I've yet to test other games).

my intial thought is maybe XP is the issue - on checking out the drivers related to various ACPI component in Device Manager, they are relatively old Microsoft drivers (whether this is the case with Vista) so maybe the coding or whatever for such drivers just isn't efficent / suitable under XP.....?

I'm going to do a clean install of Windows 7 just to see the results with the game in question, on mains vs on batt and eliminate the problem with Win XP.

And yeah it wasn't due to a dead batt ! Was the first thing I thought of actually but had lots of juice left - made it do it again today running on batt, chipset temp reached 105c ! where as the GPU temps and CPU hardly break 70c - pop it back on mains and the max the chipset ACPI temp then only gets to is 70c ish - annoyingly weird !

I'd be interested to hear of any others running win XP on their M1530, if there is any noticable change in temps of their chipset acpi temps when under load on mains vs on batt...or if it's just me !

Cheers

Andrew

 

I've spent the last couple of hours trying to reproduce this on my own system and have not been able to.
We have the same system, however I'm running Vista. I'm not overclocking either.
It sounds like it is something to do with XP's ACPI driver, but it's also entirely possible it's just a fluke and something weird your system does.

 

Thanks for testing, interesting...

Well, initial reports on Windows 7 reveals exactly the behaviour. Have installed a fresh copy of Win 7, with just the Graphics & Chipset Driver instaled. I'm running the latest A12 bios too - don't think theres any other drivers that manage ACPI to go on ?

Heres how I tested. I used suggested test program from previous threads relating to general temperature surveys -Rthdribl

testing Method:

-ensure on mains
-Run HardwareMonitor and monitor idle temps
-Start up rthdribl
-leave for 5mins for temps to settle under load
-remove power cord for a short time
-watch ACPI temps for sudden increase jump !

I notice an instant 20c jump in the ACPI temps (from 70c -> 90c +) from when removing the power cord to run on batt - also the cpu fan increases further due to temp increase

maybe it's just me ! very odd behaviour tho, and defo limits the use of the machine on the road hmmm

Andrew

 

A 20 degree jump is huge. Following your testing procedure, my ACPI temp jumps up about 4 degrees when the power cord is unplugged, but settles back down after 8-10 seconds.

How old is your machine? If it's going to effect your use of the machine you should consider calling Dell about it. From experience, it doesn't take much to convince them your machine is in need of a motherboard/heatsink replacement.

 

Mine doesn't really move much at all. Usually only a few degrees. Something doesn't seem right.

 

Thanks for the further input.. ! Something does seem odd here.

I think my only last option is to slam on a copy of Vista and at least test "as shipped" - if it does do the same it's going to be the only way of getting something out of Dell.

Mines only 5 months old - I got it around launch time of the XPS16, as the prices dropped here which made it affordable to me. Aware of the previous card & temp issues I was happy mine was running really cool. Until now ! (which of course could have been present to start with)

Anyways, off install Vista (ick! ;-)) . Report back later !

 

The ACPI/THM_ Thermal Zone ist an ACPI software interface to read temperatures. It is described in section 11 in the ACPI spec:
http://www.acpi.info/DOWNLOADS/ACPIspec30b.pdf

The operating system can read the temperatures of this software interface and choose among different cooling options (active and passive). But the operating system doesn't actually have a clue about what this temperature number means nor where exactly it is measured.

The Dell system bios of the M1530 provides the implementation to this generic ACPI software interface. I haven't seen any reliable information about how exactly this reported ACPI temperature is derived and which sensors are involved. Almost for certain the chipset temp (PM965) will be involved. However it is more likely a temperature function such as:

MAX(chipset temp; cpu temp; memory temp; hdd temp; battery temp)

It makes sense that the ACPI temperature is a combined temperature of several sensors if the cooling solution is only coarse grained (1 fan, TM1/TM2, shutdown). This is a common pattern. E.g. even the DTS Sensor of each core that you can monitor with CoreTemp is actually comprised of multiple sensors:

http://www.intel.com/technology/itj...d_Thermal_Management/p03_power_management.htm

One way to deduct what the ACPI temp actually reports is to disassamble parts of the dell diagnostic software and the bios. This is actually the option the author of i8kfangui has chosen. See FAQ #15
http://www.diefer.de/i8kfan/faq.html#99

Although the source code of i8kfanGUI is available ( click), I think nobody ever improved it or adapted it to the needs of the M1330/M1530 (you would have to continue to disassemble dell binary code). All I've seen are ports to other operating systems.

I think the demand of a software to control the fans for the M1330 (it works to some degree with i8kfanGUI) and M1530 (reading temps works to some degree, but controlling the fan speed doesn't) is higher than the demand to know exactly what ACPI temp is displaying. Still nobody has put some serious effort it. Neither have I read inside information from dell developers nor have I seen people providing some insight with other approaches (e.g. cooling spray).

What does that have to do with you battery + 20°C temp? Lithium-Ionen batteries work optimal in the 10°C - 25°C range. At 90°C most of them cease operation. A security circuit inside the battery will most likely shut down the power supply and with it the notebook before it will explode exactly like this:
http://www.youtube.com/watch?v=WeWq6rWzChw
In cases were this security circuit had flaws, the battery packs got recalled (Apple, Sony, Dell, HP, ...).

But what happens before it reaches 90°C? Temperature monitoring is mandatory for any Lithium Ion battery during charging and for security reasons - see above. This information is most likely used to throttle performance and therefore power consumption in order to keep the lithium battery within a operating temperature range. Either the Dell SMBIOS reads the temperature by means of the SMBus and incorporates it into the reported ACPI temp. In this case you probably indeed reached 90°C within the battery. Reading battery temps isn't part of the ACPI Spec - though that doesn't mean Dell does do it. If the Dell SMBIOS doesn't evaluate the battery temps it could still achieve performance throtteling by monitoring the battery discharge rate and flag a high temperature such as 90°C as soon as the discharge current is too high. In this case the high ACPI temp would trigger performance degration as a side effect.

What battery discharge rate does Vista performance monitoring actually report when running rthdribl and orthos at the same time? You can check in Vistas performance and health monitoring / + / Battery Status / Discharge Rate during DC operation? I wouldn't be surprised if it's ~ 60 Watts. I didn't design the Dell battery, but it's likely they are not made of these high drains (lithium ion batteries in general aren't particular well for this purspose. Hybrid powered cars incorporate the air conditioning unit to cool their batteries if desired).

Running rthribl and orthos at the same time during DC operation might very well be insane and a safe way to physically destroy your battery - at least reduce its capacity.

I'm not an expert on M1530 either, but does the Dell nVidia VBIOS (Video BIOS) have a reduced core/shader/memory GPU clock when running on DC? I know the GPU falls back to reduced performance once it hits certain temp thresholds. But does it run at reduced speed on DC operation, which is bypassed by certain drivers and VBIOS combinations? nVidia and Dell both encourage to install the latest system BIOS with their latest drivers:
http://www.nvidia.com/object/recommended_software_updates.html

I would expect that you don't see a +20°C temp jump during lighter loads - e.g. running just orthos without rthribl. And if that's the case, then you definitly triggered some security mechanism that has been put in place to protect your hardware and yourself.

I would encourage to always use the most recent BIOS (A12 for the M1530 at the moment). Even if other people report less fan noise with older revisions. Those people simply don't have the technical documents and insights nVidia and the Dell BIOS people have. The only exception is if you don't care about breaking hardware and at all times have Dell premium support to instantly repair/exchange your hardware.

 

Good read ! So I understand reported temps or control will have nothing to do with the operating system ? Purely down to the control & coding of bios / chipset . Which, just after installing Vista has made no difference anyways.

I agree running any real stress test under battery enviroment has an impact and such only try to replicate this on DC for a very short time. And certainly warey now of running anything on batt !

I have however noted that CPU stress does not cause the jump in ACPI reported temps, but stress more so on the GPU (running HDRIBL). The GPU does not appear to downclock when switching between DC & AC, and I have never ever noticed a downclock in any games etc as the reported CPU and GPU temps are pretty cool in whatever scenario.

And also as I understand and read the ACPI THM_ value in HardwareMonitor IS the chipset temp.

Will just have to limit how intensive what Im using is on batt, although being able to downclock GPU on batt would certainly help. Still trying to get my head around why the sudden jump in reported acpi temps is instant (like flicking a switch) from jumping between AC & DC. But at least I know what is causing the issue (GPU Stress) , and if such protection is built in then no hardware replacement will solve

thanks again for the detailed information

Andrew

 

Unfortunately it's not that easy. Old days APM worked that way, ACPI doesn't. For thermal management ACPI and the operating system work hand in hand. Demand based and context based management can only be triggered by the operating system, while most of the emergency functions regarding thermal management work independent of the OS and are triggered by ACPI and SMBIOS. That's important: Consider a crashed OS due to some faulty driver. The laptop is supposed to NOT self destruct due to overheating in this crashed OS state.

--

Did you conduct your DC test with your GPU being clocked at 700/1400/925 or stock 475/1400/950? If possible post the battery discharge current while conducting tests in DC mode.

I have posted this several times myself in this forum and I did post that actually before registering here. It gets repeated over and over again, but that doesn't make it correct. The assumption that ACPI THM_ is the chipset temp is the best approximation we have at least. Though recently I doubt it's true.

Dowclocking the GPU only helps in a linear fashion regarding power consumption. Lowering its voltage would lower power quadratic and lowering both would help even more. Disabling shaders while just browsing the web in aero would help even more. But GPUs are dammed stupid regarding power management compared to CPUs.

It's still interesting whether other observe the same behavior.

 

Understand perfectly thanks . love to know how these things all work !

Yes all tests run using stock settings for the moment - even thought gpu o/c settings run at well below the GPU temp thresholds, I don't want to put any additional load when testing further at this stage.

I have used a prog to check Battery discharge values, under the following environment :

Stock CPU settings (no volt mods via rightmark)
Stock GPU settings
Screen brightness @ 75%

-ACPI Temp under AC conditions running Rthdribl 5 mins = 65c
-ACPI Temp when removing power cord (instantanteous jump !!!) = 93c (but then thottles back to 83c when fan kicks in faster)
-ACPI temp after replacing powercord (instantanteous temp reduction) = back to 63c
-ACPI temp under no stress on AC = 44c

Draw on battery on DC running Rthdribl in above conditions = 48Wh

 

In your situation undervolting is very useful: While undervolting basically does nothing in noload/light load situations, it has a tremendous impact on power consumption on full load. A 35W TDP CPU easily turns into a 25W TDP component.

Even though everything points to the GPU so far: If you have valided undervolting settings, just use them. It reduces CPU power consumption and battery drain in Orthos by a great amount. Would be interesting to see whether the instantanteous temp jump changes (or its delta).

48Wh is a measurement of capacity. I suspect it's 48W (Watts). I would have expected even a higher drain - doesn't look too bad. Although those lithium ion batteries don't like these high drains (48W = 12V * 4 Amps).

 

I did have undervolting set previously, however with eliminating an operating system fault I have no set back up yet . An run stock for the moment to eliminate different factors.

I will do this later to prove what the jump is understand heavy GPU load, and to reduce the drain on batt, but I suspect little difference as it seems full load on the CPU has little affect on this sudden jump in ACPI temp.

Thanks again