IBM_ECW : Power management tool for T4x/R5x

[IbmPad]

Hello, here is a preliminary Linux version of the tool. Please refer to the manual included in the ZIP for instructions.

Please be aware that :
— the program in its current state is missing a lot of features present in the Windows version ;
— it has been tested only on one distro (Mandriva 2011 x86) and one PC (T42p) and may not work for you ;
— I know *very little* about Linux (in fact I don't like UNIX and its derivatives, so I never used Linux on a regular basis), so I may have made bad decisions on the design/programming — or at least some sounding odd for Linux users — or mistakes due to my ignorance…

[geka3250]

IbmPad, big thanks for this great program!
I have one suggestion to add "per core load threshold" because on dual core it looks like:
5%
10%
15%
...
45%
100%
Need to do less than 50% value to make full speed at full one core load for singlecore programs.
On singlecore CPUs it looks right:
5%
10%
15%
...
95%
100%
Thanks!

[IbmPad]

Minor update (v. 1.0.6.3) :
— Added the ability to read the CPU internal temperature sensor for non ThinkPad computers (as per user's request).
— Fixed some handle leaks and various issues.
— Small optimizations for lower memory & CPU usage.

Sorry geka3250 for not having replied earlier but I just saw your post. I will think about your suggestion but at first sight I believe it would add complexity for no real benefit — since the frequency/voltage can only be set for the entire CPU, not just one core (limitation of the CPU, not the program)…

[Adda]

There appears to be a serious bug in the new version, where IBM_ECW doesn't seem to save your settings, in stead is resets to potentially unworkable settings.
When I first used the new version, with my old config file, the settings where wrong (only hotkey settings as far as I know) so I went to change them to what I'm used to.
When saving the settings, IBM_ECW took a while to close, and even longer to restart, and the settings where not saved.

Tested on T43p and WinXP.

[IbmPad]

Oops, sorry for the inconvenience… I tried the new version on different computers and OS (from Win2000 to Windows 7 x64) before posting it and didn't notice new bugs… Please can you tell me your OS and, if you remember it, the version you were using before ?

[Adda]

The previous version would be v1.0.6.0

I can remove all hotkey settings and then save, once IBM_ECW comes up again, that setting is saved.
But when I make new hotkeys, and save those, the new hotkey settings are all wrong once IBM_ECW has restarted.
While IBM_ECW is restarting, the system is pretty much unresponsive.

[IbmPad]

Oh yes I didn't notice the hotkeys were messed up. I am going to fix that. Regarding the slow restarting, I cannot reproduce it but will investigate it also. I removed the download for the moment.

[Adda]

Maybe the slow restart is fixed once the hotkey system is restored.

I'm glad to see you are still working on this app, it is the best thing that has happened to ThinkPads in years.

[IbmPad]

OK the hotkeys are fixed (as usual it was a stupid absent-mindedness). Sorry you will have to set up your hotkeys again. But I don't know why the program is slow to (re)start. I just tried it again on XP, both service/non service and administrator/limited user and cannot replicate the problem. Since I did a lot of small modifications throughout the last months I don't know where to search… I will try to find the cause anyway.

[Adda]

No worries, the slow restart is a very minor problem, and doesn't seem to affect how fast the program launches at boot.

Edit: IBM_ECW restarts as fast as usual now, the slow restart must have been caused by the hotkey problem, all is good!

[IbmPad]

Great! Thank you for your time and tell my if you discover other bugs!

[Adda]

It will be my pleasure!

[Adda]

IBM_ECW seems to work fine on an R60, but as usual, I'm too daft to figure out how to set the base I/O address, I have tried different values, but get no meaningful readings.
The default (0x2000) shows no reading at all.

CPU control works, I got a Core Duo 1.66GHz.
Fan control works.
HDD control works.
PowerPlay profiles work on the X1300.

Edit: there is one minor problem with PowerPlay, the lowest setting (on the slider in IBM_ECW) loads Max Battery as it should, the medium setting loads Max Performance, while the highest setting loads Balanced.
Not a problem, just a little strange that's all.
I use the PowerAPI mode.

Edit: there is a problem with CPU control, there the CPU will switch between min and max speed, when it is configured to run at max speed only.

I'm using Win 7 Pro x86

Edit2: ok, I have multipliers ranging from x6 to x10, if I set the CPU to run at max speed, when the x10 or x9 multiplier is set as the highest, then the CPU will switch between min and max speed.
If I limit the max speed to x8, then the problem goes away, and the CPU will remain at x8 when set to run at max speed.

[Adda]

I have found another bug I hadn't noticed, CPU threshold settings don't save properly, some of them seem to be reset while others do not.

Edit: ah ok, I can't change the threshold for the highest multiplier it always resets to 100, is the on purpose?

Edit2: geka3250 is right, there is a problem with single core loads on dual core CPU's, IBM_ECW sees a heavy single core load as a 50% load.

Edit3: aaah ok, it seems I have miunderstood threshold settings, the CPU seems to switch to the highest multiplier, once it exceeds thres of the second highest.
So if I set thres for x9 to 50, and x10 is at 100, the CPU will go to x10 at a 50% load.

[IbmPad]

Update (ver. 1.0.7.0) :

► the system timer resolution (that programs can raise using NtSetTimerResolution — or timeBeginPeriod which in turns calls NtSetTimerResolution ; drivers can also raise it by calling ExSetTimerResolution) is shown and can be modified using a hotkey (or AC/DC rule) as following :
— on Windows 2000 the resolution can only be raised ;
— on Windows XP & Vista the resolution can be raised or lowered (to lower it, the programs needs to download the NTOSKRNL symbols from Microsoft and thus requires the file SYMSRV_ECW(64).DLL to be copied in SYSTEM32);
— on Windows 7 & 8 the resolution can be raised, or lowered if no other driver raised it using ExSetTimerResolution (again to lower it you need to copy SYMSRV_ECW(64).DLL to your SYSTEM32 folder).

► the program executable and driver sizes & memory consumption have been reduced ;

► the program is now compatible with Windows NT 3.51 and NT 4.0 (though I doubt someone is still using one of these versions !) ;

► (as usual) a few bugs have been corrected, and (as usual ) new bugs may have appeared…

[Adda]

Seems to work, my guess is the "clock signal" is the speed at witch the CPU clock is synchronized with system load correct?
Does this mean IBM_ECW supports HPET? (QueryPerformanceFrequency = 14.31818 MHz, try WinTimerTester)

Edit: the new IBM_ECW feature is causing some very serious DPC Latency problems, whenever the CPU is changing frequency (like when you play flash video fx.) (test with DPC Latency Checker).

Edit2: there is more to it then that, I need to test more, it seems IBM_ECW only plays a part in this problem, it's still there w/o IBM_ECW, gotta try more tests, could be a flash player problem.

Edit3: false alarm, I just never knew flash was as bad as that, the new feature does affect DPC Latency though.
It would be nice if the clock signal setting was easily synchronized with HPET though, for best possible performance (unless of cause I'm talking gibberish here).

Edit4: I feel a bit foolish now, it's the wireless card that doing it of course sigh...

[IbmPad]

I probably did not express myself properly. I was referring to the interval between two clock interrupts. When this interval is low, the CPU is woken more often, so timers are more accurate (mainly useful for multimedia applications), but power consumption is increased. Thus by altering this value you can choose what fits you the best.

I forgot to add in my previous post that when the program is not run as a service, I/O operations are moved in a new thread so that the UI should stay responsive. I also just added Windows 95/98/ME support (v. 1.0.7.1).

[Adda]

The new "cock signal" feature doesn't seem to have any positive effect, on Windows7 32 and 64 bit at least.

[IbmPad]

Update (ver. 1.0.7.2) uploaded on June 20, but I forgot to post the update here :
— NVidia GPU clocks are now reported correctly (using NvAPI.dll) and can be modified without needing an external program;
— added the AccessIBM/ThinkPad key as a possible hotkey (eg. “TPKEY” or “Ctrl-Win-TPKEY”);
— increased the maximum number of hotkeys from 10 to 100;
— fixed a regression in the x64 driver: the file “acpiec.sys” (Win2003) or “acpi.sys” (Vista/7/8.1) could not be patched;
— changed the dialog font from “MS Sans Serif” to “MS Shell Dlg” for better compatibility with Windows PE.

Update (ver. 1.0.7.3) uploaded today :
— minor driver updates for Windows 10 (in-memory patches), not tested;
— better compatibility with limited user accounts on Windows 2000 and XP;
— removed dependency on Shell32.dll and WtsApi32.dll;
— switched to the Visual C++ 2015 compiler and Windows 10 SDK.

@Adda:
I think the “clock signal” setting is not intended for general use and is sometimes buggy, but it does the job for me…

[Adda]

Excellent update, it seems to work really well, the clock signal feature works much better now.

[Thrakath]

Hey, sorry wasn´t here for a long time! The newest version is running fine on my T61 and another T420s (just fancontrol). I now tried to install the newest on my father´s T61. It had some of the very preliminary first versions with test signing aso installed that did not run stable. I now just tried to install him the newest version, but it does not start up. Small window with: The service is not yet ready, you can wait for it...blabla! But it´s never coming up. Do i still need the test mode? Or driver signing? I tried to delete all the old files (ini in Windirectory) but did not find the old driver! What do i need to do to get it running?

[IbmPad]

Hi, “Test mode” is still required; I have been using it for over a year on my main computer (Win 7 x64) and it runs well, along with the option to remove the watermark on the desktop.

Could you try to remove the service (“sc delete ibm_ecw”) and kill any remaining process (“taskkill/f /im ibm_ecw64.exe”) and run it again?

[axur-delmeria]

I'm trying to access the IBM_ECW homepage from a newly built X601 Frankenpad, but Bitdefender Free Antivirus is blocking it due to having detected malware in the site. Any ideas why this is happening?

[IbmPad]

This is most probably a false positive; I just checked the files and pages and they don't seem to have been tampered with. For reference, the current IBM_ECW.ZIP file's SHA1 is 7813B9B00F7C5B2868AD072339DA8D5C6A7CF3B6.

PS: it seems this is not new; back in 2014 the IBM_ECW program (provided this was the original file obviously) was detected as malware by some antiviruses (BitDefender being one of them) and more recently the website itself! I'm not surprised; I don't have an antivirus myself but some of my own programs are blocked when I plug my flash drive on other people's computers…

[flyingfishfinger]

Hi, I just tried building your Linux version. It fails with the following errors:

make -C /lib/modules/4.2.0-27-generic/build M=/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV modules
make[1]: Entering directory '/usr/src/linux-headers-4.2.0-27-generic'
CC [M] /home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.o
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c: In function ‘GetCPULoad’:
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:79:29: error: ‘struct kernel_stat’ has no member named ‘cpustat’
BusyTime += kstat_cpu(i).cpustat.user +
^
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:80:29: error: ‘struct kernel_stat’ has no member named ‘cpustat’
kstat_cpu(i).cpustat.nice +
^
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:81:15: error: ‘struct kernel_stat’ has no member named ‘cpustat’
kstat_cpu(i).cpustat.system;
^
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:82:29: error: ‘struct kernel_stat’ has no member named ‘cpustat’
IdleTime += kstat_cpu(i).cpustat.idle +
^
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:83:29: error: ‘struct kernel_stat’ has no member named ‘cpustat’
kstat_cpu(i).cpustat.iowait;
^
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c: In function ‘init_module’:
/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.c:287:9: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
ret = (int)device_create(DevClass, NULL, Dev, NULL, DRVNAME);
^
scripts/Makefile.build:264: recipe for target '/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.o' failed
make[2]: *** [/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV/IBM_ECLDRV.o] Error 1
Makefile:1398: recipe for target '_module_/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV' failed
make[1]: *** [_module_/home/rafael/Desktop/IBM_ECL/IBM_ECLDRV] Error 2
make[1]: Leaving directory '/usr/src/linux-headers-4.2.0-27-generic'
Makefile:4: recipe for target 'all' failed
make: *** [all] Error 2

Any ideas? Thanks for working on this great tool!

R

[IbmPad]

Hello,

It seems the structure I was using (defined in linux/kernel_stat.h) has changed since Linux 2.6. At first sight you can try to replace the offending code block by :

Code: Select all

for_each_present_cpu(i) {
  BusyTime += kcpustat_cpu(i).cpustat[CPUTIME_USER] +
              kcpustat_cpu(i).cpustat[CPUTIME_NICE] +
              kcpustat_cpu(i).cpustat[CPUTIME_SYSTEM];
  IdleTime += kcpustat_cpu(i).cpustat[CPUTIME_IDLE] +
              kcpustat_cpu(i).cpustat[CPUTIME_IOWAIT];
}

Anyway my knowledge of Linux is extremely limited; I wrote this quick-and-dirty Linux version when I had to use Linux for school. Now I don't have Linux installed anymore so I cannot test this. Furthermore this program is very limited and there are probably much better tools for Linux. I don't even know if it will be working on other distributions than the one I used—I think it was Mandriva 2011—since I didn't really bother to find out which interfaces/paths/services/etc. were specific to the distribution and which ones were common to all Linux systems, or all UNIX-like systems… But anyway if I can be useful to you I would be happy to help you.

[flyingfishfinger]

Oh, your tool is the best I've found (at least on Windows). For Unix, there's linux-PHC plus a TP-fan control. There doesn't seem to be a different tool that lets people control the CPU voltage, multipliers, fans and read temperature all from the same panel!
I'd be happy to do some testing, so I'll try out your fix and get back to you. Note that I'm not a programmer myself, but I'd be happy to try and report the symptoms in cases of failure.
Thanks for writing great stuff!

R

[flyingfishfinger]

Changing the code as you mention makes it compile and run, that's a good first step. However, it's not 100% there.
The GUI client requires libXm.so.2 it tells me, which is not found on Ubuntu (anymore?) and thus doesn't work.
The command-line client kind of works, but does not give me any readings at all (all blanks, "---" and zeores). However, I AM able to control the fan and Thinklight with it if the daemon is running.
I can't tell if changing the CPU settings has any effect, since I'm not aware of a good tool that can monitor voltages and actual current frequency. Certainly the program does not tell me anything changed, since everything is blank.
So, almost working but maybe still needing some tweaks. Please let me know if you have any updates or settings that I should apply or install.
Thanks!

Rafael

[Adda]

I think that IBM_ECW may cause some degree of instability on my R60 with Win7.
On very rare occasions it crashes giving me a "hardware malfunction" it's probably also the reason why it sometimes freezes when changing from DC to AC power.

Other then that it really does wonders for my R60, the T7200 2GHz CPU only needs 1025mV at max speed to work.

[Thrakath]

Someone solved the sound stuttering issues?
By uninstalling IBM_ECW i get much less, but still not perfect. Got a broadcom and the NDIS 6.20 causing most of the the DPC stuttering and are severely worse with IBM_ECW on...would still like to use this amazing software.

Edit: Was the clock signal under Misc. Resetting it to 0.0 made stuttering disappear.