got mine yesterday delivered to work and it's pretty nice, then i took it home and the fan smacked me in the face. even on the quietest mode it's loud. the closest thing that most people have probably heard, and that i can compare it to is an 80mm fan @2000rpm behind a grill...

for the time being i can live with it since it's mostly for work and i can't hear it there, but i'm still on the lookout for something to slow it down. perhaps a future version of speedfan will support it, i sure hope so.

Have you tried the HP Thermal Wizard aka Cool Sense Wizard? Reviews I've read say you can slow down the fan using it.

"It's quiet in "quiet mode" The fan always runs, but its not bad."

If you did a clean install of windows you have to install the HP Support Assistant to get the wizard.

another poster said: "For those that want minimal fan noise they need to use quiet mode in HP's cool sense and also run in power saver mode in W7 which will not noticeably slow the machine for normal uses."



http://h10025.www1.hp.com/ewfrf/wc/docu ... =c02638953

Video review for those not familiar with this laptop http://www.youtube.com/watch?v=lsnSATGW8Wc

Just remember the reviewer didn't modify the fan settings so her comments about the fan aren't what you'd deal with if you took a couple of minutes to do something about it.

If you have the opportunity, please take some power measurements. I'm curious what the power draw is while plugged in:
- when charging
- when not charging and off/sleep/hibernate/idle/playing movies/etc...

I was thinking an E-350 based laptop (plus a USB tuner) might be the low power king for a HTPC.

thermal assistant aka coolsense is active and on 'quietest mode'.

from what i can tell fan speed is linked to temps, so limiting the cpu speed will work to an extent, but at a cost to the performance.

like i said speedfan is not yet compatible, i even went into the 'secret bios' and enabled thermal fan control but that did nothing.

i took it home again last night, at idle the fan is more like an 80mm @1500rpm but it doesn't take long for the little bugger to get warm and then we're back to ~2000rpm. it's pretty much intolerable.
if you have a quiet pc and a quiet environment then i wouldn't want to use one of these next to it.

no power measurements unfortunately because i don't have a meter.
i don't doubt that its more efficient than the desktop stuff though. perhaps a nettop would be more suited to your needs, the zotac zbox was reviewed a while ago and from memory it did about 11w idle.

on a positive note i got hardware acceleration working
using the amd fusion system monitor to watch the cpu and gpu use about 30% during 1080p playback was fun.

any other zacate laptop owners please chip in!

Does anybody know what kind of fan is used in this thing - and if it would be possible to replace it?

I'm waiting for this baby to come to Denmark - and sitting on my hands not to order it from the US and risk all the hassle if something is not right and having to ship back. But all this talk about it being noisy :-/.... me no like

/Casper

looks custom.

I came across this thread about decreasing the speed of the fan on the DM1 - it links to a few other threads as well.

http://forum.notebookreview.com/hp-comp ... noise.html

/Casper

I know this does not deal directly with the problems...

Are there any unnecessary processes running in the background that you can kill to reduce processor load?

Unfortunately, the fan is not controlled through ACPI or AMD's drivers on this laptop. There is a WMI call that can be used to alter the cooling profile (quiet, normal, cool) and I've been able to use this to reimplement HP Coolsense - unfortunately no fine-grained fan control yet.

Using the quiet profile, the fan activates at 50C and progressively gets faster until it hits max at ~65C (4 discrete fan steps, ranging from 2000 to 5600RPM, according to ACPI). Personally, I would modify this to cool passively until ~58C and hit max at 70C (never happens - even prime95 couldn't get this over 65C here).

As far as I can tell, the only 100% solution would be to locate the registers that report/control fan speed and manipulate those - basically reimplement SpeedFan (or wait for it to support this laptop).

Another alternative might be to somehow hack the temperature report to read less than it actually is. I did manage to implement this for the ACPI temperature, but unfortunately it didn't modify fan speed. This could mean one of two things: (a) the fan controller is reading the temperature before the modification takes place or (b) the controller is not relying on the ACPI temperature but one some else (e.g. GPU temperature).

If someone is feeling adventurous enough, it would be interesting to dismantle the cooling assembly and apply new thermal paste to the processor. If this produces a visible difference in temperatures, the fan would stay at passive cooling for longer (right now, I hover at 52C which is just 2C over the active-cooling threshold).

Undervolting would also help, but I don't know of any programs capable of doing that for E-350 yet. Note that AMD overdrive does not work on this GPU yet (if ever).

tl;dr I really hoped for ACPI support but unfortunately there just isn't any. Need alternative method. I really hate it when this happens, bad showing HP!

very informative post StApostol, i'm hoping for future speedfan support.

from my last try bios settings made no difference, i may try again next time i have it home. i also want to put my ssd in there which will help.

one interesting thing is that exhausted air is only ever slightly warm, even when gaming. this may suggest that a faster fan has limited effect on temperatures... the heatsink looks pretty dinky so i think it might have reached max heat dissipation even with the fan on low. just a theory.

I've been poking around with an awesome (and dangerous) little program, called rw-everything and I've been able to uncover some more info on the external controller (EC) section.

The fan has an on/off switch and there is a number of control bits that control the cooling profile (these are the ones used by HP Coolsense). There is a RPM display (read-only) a temperature display (read-only), a second temperature display (read/write) and several other bytes that need more testing to understand the use of.

Using this information, I should be able to write a simple app to continuously modify the r/w temperature display, forcing a different fan response. Unfortunately, this wouldn't be 100% reliable, so I'll keep on looking. Anyone else interested in this, download the rw-everything, click the EC button and use HP Coolsense to modify your cooling profile and see what bits change. With a little effort, we should be able to crack this.

the screenshot on that website scared the crap out of me...

Ok, after several more hours of testing I still wasn't able to control the fan directly. It looks as if the hardware has three preprogrammed fan response curves and allows you to choose one of those. I hoped that it would be possible to disable this and control the fan directly but even if this is possible, I wasn't able to discover which bits to toggle.

I did manage to force a thermal shutdown (oops!), kill wifi and bluetooth (reboot fixed), disconnect the battery and discover how this laptop understands whether it lies on a table or on a lap (it contains a pretty accurate 3d accelerometer, pretty much like an iPhone).

In addition to the info in my previous post, I've found something that looks like a fan voltage value - but this appears to be read-only. Will let you know when I have something new to share!

Hi all,

I'm also a dm1z owner (well, the dm1-3130sf, french version) and I'm interested in silencing it a bit, so I'll take any info on how to do so. Moreover, I'm a Linux user, which means HP Coolsense doesn't apply. So I will look for information if some is available.

So far, what I have seen is that the system idles, at around 10/11Watts (ACPI info through intel powertop, with Firefox / Evolution running). I have a plug meter, and will try that soon. Some are already hacking it on blogeee.net's forums, by reducing voltage to the fan. I am also a cpupowerd user on a older AMD system, to good effect, so I'd be interested in any knowledge on reducing voltage for CPU states.

You can achieve the same effect as coolsense by using a program such as rw-everything. I haven't been able to find something like that for linux yet - but something *should* be available (or possible to implement). See here, for instance: http://cateee.net/lkddb/web-lkddb/ACPI_EC_DEBUGFS.html

If you have that, you simply need to write 0x02 to the EC addresses 0x5A and 0x5B. This will switch the laptop fan to the 'quiet' profile, which is much more tolerable than the default 'optimized' profile. (0x01 is the 'cool' profile and '0x00' is the default/'optimized' profile).

I don't think there's voltage control for fusion yet, but that could also help (I don't know how much, though, since the CPU is already low-power). My Watt meter shows around 12-14W idle depending on what programs I have open, which is pretty reasonable given potential power brick efficiency.

Edit: I just realized that you said 3130sf, which is a slightly different model. In this case, be careful when writing to the EC (keep a backup of all 256 bytes first, so you can restore if something goes wrong!) First of all, look under /proc/acpi/fan/ for any fans - if any appear, you should be able to control them through lm-sensors and the fancontrol scripts.

If that doesn't help, try this: install iasl and decompile your dsdt file with that (it should be somewhere in /proc/acpi). Search for a table that looks like a fan speeds in hex (it will probably be like this: 0x00, 0x18, 0x24, 0x36, 0x40, 0x64, where 0x64 is hex for 100%). If such a table is available, bingo! You can modify your fan speeds directly (do keep the 0x64 speed intact, though, for safety). Also look for a regions called "THRM", "TMP", "TZ0" - these define the code that controls your temperatures, active and passive cooling zones.

If nothing works, the EC might provide a solution. The more information we can gather, the more likely we'll be able to solve this issue. Voltage mods would also work, but they are unsafe and would break our warranties - should only be a last measure solution.

Thanks for all the info. I believe ACPI has information on the profile used, however there is no fan control yet (i.e. /proc/acpi/fan is empty).

acpi -V gives me that, which may be of interest :

Battery 0: Unknown, 100%
Battery 0: design capacity 4394 mAh, last full capacity 4553 mAh = 100%
Adapter 0: on-line
Thermal 0: ok, 46.0 degrees C
Thermal 0: trip point 0 switches to mode critical at temperature 98.0 degrees C
Thermal 0: trip point 1 switches to mode hot at temperature 94.0 degrees C
Thermal 0: trip point 2 switches to mode passive at temperature 90.0 degrees C
Cooling 0: LCD 0 of 10
Cooling 1: Processor 0 of 3
Cooling 2: Processor 0 of 10

Since the EC values you list are (0, 1, 2), could they refer to the trip points of Thermal 0 above? And of the Cooling 1: (Processor 0 of 3) ?

I'll try decompiling the dsdt table next, this looks interesting.

Your output looks identical to mine (temperatures notwithstanding).

Try the dsdt route, as that's relatively simple and safe if it works. If not, you can reduce fan noise considerably using the EC trick - but I still don't know of a way to write that on Linux without recompiling a kernel (/proc/acpi/embedded_control/EC/info lists the r/w addresses but doesn't offer a way to use them).

My understanding is that 0, 1 and 2 refer to cooling profiles embedded in the BIOS or the fan controller. There's no apparent way to modify those, yet.

One other thing: on reboot, press F10 enter your BIOS screen and immediately press A to enable the advanced settings. There are a couple of settings pertaining to thermal control there - but I haven't seen any difference at all by toggling those.

I decompiled the dsdt and could not find much. A few tables starting at 0x00 and ending in 0x64, but not much in the way of finding out if this is a fan control.

However, I found this :


In what seems to be a long initialisation sequence.

I'll try to change the Zero to 0x02 on the second Store, maybe it will do the trick ?

I tried too the advanced settings in the BIOS, after I discovered that the A key is Q on a french keyboard... but changing settings in the BIOS didn't do much.

I will also try to document myself on cpupowerd and what information would be necessary to get to use it for undervolting the E350.

Great find! That could indeed do the trick (enable quiet coolsense profile).

I'll have to take a closer look at the dsdt, there's lots of interesting stuff in there.

(Please note if something goes wrong, just disconnect AC and battery and press the power button for ~10 seconds and it will restore to default settings. You'll need to remove the custom dsdt though. This is extremely unlikely to happen but it's something to keep in mind if something goes wrong and you can't boot - have had this happen on HP laptops before, for various random reasons).

Undervolting support would be great but I have absolutely no idea what that would entail...

Ok, i compiled a new dsdt, but I didn't manage to upload it, even after adding it to the initramfs. I'm sure there must be a way to write a script able to toggle those EC registers directly !

I spent some time looking throught the BIOS and Kernel developers guide for family 14h (Brazos) and apparently undervolting may be simpler than I expected. It could be done in the same way as with intel CPUs in the linux-phc project, since an E350 has support for hardware P-states, and a register holds the Vid associated with each P-state. I'd prefer a script, however, since keeping the kernel patched with linux-phc is a mess on the long term.

Ok, I'll look more into that.

I have a solution for activating coolsense with Linux. The acer_ec.pl script, available around the acer_acpi module effort, allows one to write into the EC registers.

http://aceracpi.googlecode.com/svn/trun ... acer_ec.pl

With it, switching on coolsense is simply this :

sudo ./acer_ec.pl := 0x5B 2

Next is to work on undervolting since now temperatures are rising

Replying to myself again, but this may be of interest for others.

Undervolting done under Linux on my dm1z, at least to a reasonable level (I'm happy with what I have achieved, now I'm using it to see if it stays stable). However, I lack the ability to really measure the effects, apart from the temperature under load.

Some of the difficulties:
* Undervolting is done by raising the ViD field for each P-state. However, computing voltage from ViD is listed in a AMD document under NDA . Each P-state has a Amp rating, but since I can't get a voltage, I can't estimate the power gain.
* As soon as the ViD change from the BIOS settings, powertop doesn't give me any power use estimates.
* My plug wattmeter isn't precise enough. And, anyway, it seems I'm measuring a lot more than the CPU power.

So, anybody has access to better power measurement tools ? An E350 motherboard, maybe ? Or does something like CPU-Z can list the voltage correctly ? I can provide the values I have reached for the relevant registers.

Wow, nice!

Could you please post how you did this? I'm interested but I have no idea how to proceed.

What values did you reach? Is there a measurable difference in load temps? I have a power meter so we could cross-reference our results (they won't be very accurate but better than nothing).

Yes, the result is good so far. With the silent profile and undervolting, it seems the main noise under use is the hard drive.

Undervolting is a bit technical, but not too complex. The E350 has hardware P-states, from P0 (the highest performance one) to P2 (lowest performance). P0 is 1.6GHz, P1 is 1280MHz, P2 is 800MHz. Each P-state is described by a MSR register with an Idd estimate in A, a frequency and a voltage.
P0 MSR is at address C0010064
P1 MSR is at address C0010065
P2 MSR is at address C0010066
If you use Linux Gnome cpu frequency applet, you can toggle by hand each cpu core to one of the P-state.
A MSR register is 64 bits. In a P-state MSR, bits 15 to 9 are the Vid (The highest the Vid, the lowest the voltage).
So, for example, P0 MSR for my dm1z is 0x8000012600002810
Starting from last, 10 is the frequency, and 0x28/2 = 0x14 is the Vid. Undervolting consist of raising the Vid by writing a new value in the MSR register.

I used the wrmsr and rdmsr commands to do that, and I started to increase the Vid.

# Now, P0
sudo wrmsr -p 0 0xC0010064 0x8000012600002A10
sudo wrmsr -p 1 0xC0010064 0x8000012600002A10
#Significant jump!
sudo wrmsr -p 0 0xC0010064 0x8000012600003010
sudo wrmsr -p 1 0xC0010064 0x8000012600003010
#Significant jump again! Lock up
sudo wrmsr -p 0 0xC0010064 0x8000012600003810
sudo wrmsr -p 1 0xC0010064 0x8000012600003810
# Smaller jump. OK
sudo wrmsr -p 0 0xC0010064 0x8000012600003210
sudo wrmsr -p 1 0xC0010064 0x8000012600003210
# Smaller jump. OK
sudo wrmsr -p 0 0xC0010064 0x8000012600003410
sudo wrmsr -p 1 0xC0010064 0x8000012600003410

And I stopped there for P0. I set P1 to the same value (since, by default, Vid is almost the same). I went a lot deeper with P2, never reaching a lock up point.

Each time, to set a new value, I switch with the frequency applet both cores to another state, write the MSR registers, switch the cores back to the P-state I'm working on, and check if the values have been correctly applied with:

sudo rdmsr -p 0 0xC0010071
sudo rdmsr -p 1 0xC0010071

In this register, the current Vid is in the same position as in the MSR, and the number just before (0, 1, or 2) is the current P-state. So it's easy to check that the changes have been applied.



Finally I reached the following :
#P0 ViD 0x1A (original 0x14)
#P1 ViD 0x1A (original 0x15)
#P2 ViD 0x35 (original 0x26)

P2 has a huge decrease. In P0 and P2 (the ones I tested with burnMMX) you see a decrease of a few degrees, more in P0; up to a point that for P2 it would hardly trigger the fan ramp up with two burnMMX running. At full load on P0, it would decrease system power by a few Watts.

Awesome, thanks!

Testing as we speak.

I've managed to reach the following VIDs:
P0: 0x38
P1: 0x3c
P2: 0x3a

with some headroom left. These are 15' prime95-stable (will do longer tests later on) and there is measurable improvement in load temperatures. Unfortunately, idle temperatures are markedly higher than on Win7: ~54C (linux 2.6.38) vs ~50C (Win7) with ~21C ambient and silent fan profile. It was recently reported that linux 2.6.38 may suffer from a power management bug that increases power consumption by 10-30% - this could explain the difference.

Another thing I noticed is that the system doesn't outright lock up when reducing voltage below recommend limits. The first indication that something is wrong is a general slowness to respond (mouse clicks, keyboard input, everything becomes slow). Prime95 does not report errors at this state, which is very curious. Decrease voltage anymore and the system will immediately hard lock. Increase voltage and it will suddenly become responsive again. I wonder if anyone else can reproduce this.

Does the programming manual list maximum VIDs, or do these change according to the CPU model?

Thoses ViD for P0 and P1 are impressive ! It must really lower the power usage and temperatures, to a point where you don't get much gain in power by going to a lower P-state. My lockups were true lockups, however.

In the BIOS and Kernel guide for family 14h, there is no limit for ViD. A register allow the BIOS to set a min and a max ViD, but it's not CPU dependent. Maybe there is something in the chip detailed data ? My estimations, and they are really just estimates, is that a ViD of 0x14 may be 1.100v. From the AMD guide and their examples, I deduced 0x24 is 1.000V, 0x2C is 1.000V too?, 0x34 is 0.892V, 0x3C is 0.798V. But why on earth this information is under NDA ?

Idle temperatures are lower on mine (48°C for over 22°C ambient), but I'm on 2.6.35 (ubuntu 10.10). I may test a 2.6.32 (LTS) kernel under 10.10 to see if the idle power rise is as seen by phoronix (an increase from 2.6.34 to 2.6.35, and then another increase from 2.6.37 to 2.6.38) is visible.

Have you tried the following method?

Patching DSDT Table

Personally, no. I tried a part of it (decompiling the dsdt, searching for a fan speed in it) without finding anything, but then I don't understand much of what is in a dsdt anyway.

Now, given that, with undervolting and the "quiet" profile, I need to have my ear right next to the left vent to hear the fan spinning at a very low speed, it seems there is little need for dsdt patching. Turning off the harddrive or upgrading to a SSD, maybe