NotebookForums.com › Forums › General Notebook Discussions › Notebook Guides › Simple Undervolting Tutorial - Pentium M Dothan
New Posts  All Forums:Forum Nav:

Simple Undervolting Tutorial - Pentium M Dothan - Page 13

post #241 of 1202
4/22/05 at 4:45pm
Quote:
Originally Posted by Picard
This is bizzare, whenever my browser's on notebooksforums.com, RMClock reports my CPU is on full time full speed BUT fangui reports I'm only doing 798MHz.
RMClock does a realtime calculation to computer the frequency. If you want the cpu frequency reported by RMClock, fangui, and speedswitchXP to match set "Realtime speed calculation" as the method for CPU reading in both fangui and speedswitchxp. These should always match now. The speed reported by windows may differ however- I recall there being a fix for this.
Reply
Reply
post #242 of 1202
4/22/05 at 8:16pm
EDIT: I just found your earlier post about DPI causing crashes. I think this is probably my problem.

Quote:
Originally Posted by mich43L
I found the problem. Appartently if you increase the DPI to 120 (from the default 96 in my case), RMClock GUI will crash:

http://forum.rightmark.org/topic.cgi?id=6:142

Nothing we can do but request for the bug to be fixed.
---
disregard my earlier post:

Quote:
Originally Posted by mich43L
How did it crash?
Yesterday, I did all the step-by-step testing for errors in Prime95 and found the preliminary settings for stress testing. All my testing has been done without "Run HLT command when the system is idle" checked. I tested the Min VID for 6 hours without a Prime95 error and without an RM Clock error.

Then I tried to do a longer test on the lowest Max VID that doesn't cause Prime95 errors and now RM Clock crashes within a few minutes.

I just tried it again and got the same error. Again, Prime95 continues running without any errors. This time, the error occured when I clicked on the Monitoring tab.

The error message says RM Clock "has encountered a problem and needs to close."

AppName: rmclock.exe AppVer: 1.5.0.0 ModName: mfc42.dll
ModVer: 6.2.4131.0 Offset: 0000c005

DrWatson Debugger also crashes after I close the RM Clock error box.

My question is, is the RM Clock crash without Prime95 errors just an RM Clock bug or is it a sign the voltage is too low?
Reply
Reply
post #243 of 1202
4/22/05 at 8:23pm
deleted
Reply
Reply
post #244 of 1202
4/22/05 at 8:26pm
Quote:
Originally Posted by mich43L
I think you get the same problem that I did and others who have a higher dpi than 96. I discussed this a few pages back.
It doesn't have to do with the system instability but a bug in the software.

Start on this post:
http://notebookforums.com/824638 and read down. There is a link there to download a new copy of the software with the fix.
Thanks, I just realized that after making the previous post. I started re-reading the thread and found your post about DPI causing problems.

Reply
Reply
post #245 of 1202
4/23/05 at 7:07am
Thanks to the OP and all the other helpful posters, I think I've found my stable settings.

Min VID: 0.732V - tested Prime95 6+ hours
Max VID: 1.084V - tested Prime95 7+ hours

Is that enough time to run the initial stress tests? I'm going to start running these voltages during normal use and see if I run into any problems.

I did try a Min VID of .716 again today and got the same instant BSOD, but .732V tested without errors.

With overclocking, people often recommend finding your max stable setting then turning it down one notch for normal use. Is it ok to undervolt at the max setting or is it safer to move it up one notch from the stable Prime95 settings?
Reply
Reply
post #246 of 1202
4/23/05 at 1:33pm
Quote:
With overclocking, people often recommend finding your max stable setting then turning it down one notch for normal use. Is it ok to undervolt at the max setting or is it safer to move it up one notch from the stable Prime95 settings?
It's generally good practice to move up a notch or two from your "crash voltage". You're talking 0.016v increments. The difference between 1.084v and 1.090v as far as power is concerned is almost nothing, but it causes the system to be more stable.
Reply
Reply
post #247 of 1202
4/24/05 at 12:02am
Well, I used Asdasl's method to calculate the additional battery life gained by undervolting.

With Prime95 running and the system at 2.0Gz, undervolting lowers power drain by 7-8W which works out to an extra 15 minutes battery life. Even more impressive is an almost instant 10c degree drop in CPU temp according to Mobilemeter.

With Prime95 running at 800MHz the difference was maybe 1W saved, too small to make an accurate call just by eyeballing Mobilemeter.

I ran all day with my stress tested settings without a problem, but I've bumped the min/max to 0.748/1.100V for an extra margin of safety and stability.
Reply
Reply
post #248 of 1202
4/24/05 at 12:32am
Quote:
Originally Posted by bloke
Well, I used Asdasl's method to calculate the additional battery life gained by undervolting.
Though I also recomended using the discharge rate to faster compute the battery life gain there is even a better way.

Just look what Windows reports for battery remaining life (hover over the battery icon in system tray or go to battery meter in control panel) before and after you run RMClock (you can also set the PST Peformance profile to "No management") and compute their difference. Systems with better BIOSes give a better estimation.
Reply
Reply
post #249 of 1202
4/24/05 at 1:51am
Great thread, I'm still testing, but with my 1.6 on my 6000d I've got the Maximal VID all the way down to 0.940. I'm holding steady at 38 degrees with Prime95 churning away.

One question, what is a good temp range for my hard drive? Right now it's staying the mid 40's to low 50's, I'm just curious where it should be.

Thanks.
Reply
Reply
post #250 of 1202
4/24/05 at 4:06am
For people wondering about performance while undervolted, I just did a few 3DMark05 runs with my CPU undervolted to 1.1V and GPU overclocked.

3DMark05:

5699 (Image Settings: High Performance)
5558 (Image Settings: Quality)

post with screenshots
Reply
Reply
post #251 of 1202
4/24/05 at 4:31am
wow that's pretty crazy. I haven't overclocked my quadro 1400, but even so I got 2666 on 3dmark05. Crazy to think they are essentially the same card. Yours obviously much higher clocked than mine, lol.

Tellerve
Reply
Reply
post #252 of 1202
4/24/05 at 3:35pm
Quote:
Originally Posted by mich43L
Just look what Windows reports for battery remaining life (hover over the battery icon in system tray or go to battery meter in control panel) before and after you run RMClock (you can also set the PST Peformance profile to "No management") and compute their difference. Systems with better BIOSes give a better estimation.
This is good for a rough estimate. Windows tends to be a little off I've found. I'm an engineer, so I like my numbers as exact as possible.
Reply
Reply
post #253 of 1202
4/24/05 at 3:38pm
Quote:
Originally Posted by Asdasl
I'm an engineer, so I like my numbers as exact as possible.
I'm a numerical analyst and prefer to approximate with a high tolerance
Reply
Reply
post #254 of 1202
4/24/05 at 3:48pm
Does anyone have the problem where it performs on Maximal even though you put it on "Automatic Management".
Reply
Reply
post #255 of 1202
4/24/05 at 3:52pm
Quote:
Originally Posted by Med62987
Does anyone have the problem where it performs on Maximal even though you put it on "Automatic Management".
Yes. It sometimes does that when I unplug the AC adapter and it runs on battery. Restarting the program doesn't help..only rebooting.
Reply
Reply
post #256 of 1202
4/24/05 at 4:57pm
Quote:
Originally Posted by mich43L
Just look what Windows reports for battery remaining life (hover over the battery icon in system tray or go to battery meter in control panel) before and after you run RMClock.
Well actually, I did take a few notes about the system tray remaining battery life during my tests, but I was too lazy to do any calculations...

2.0Ghz 1.084V rmclock auto: 1:39 remain at 93% , so @100% = 106 min

2.0GHz 1.308V: 1:19 at 85%, so @100% = 92 min

= 14 minutes battery life gained, which compares to my 15 minuted estimation with Mobilemeter...close enough

---
800mhz rmclock off: 1:51 remain 90%, so 123 min at 100%

= 31 minutes more than 2.0GHz@1.308V
= 17 minutes more than 2.0GHz@1.084V

I didn't make note of the 800MHz undervolted number, but I think the difference is less than 3%.
Reply
Reply
post #257 of 1202
4/25/05 at 1:59am
This is just a fantastic tutorial.

Using an Acer 4000, 1.7GHz w 1GB RAM.

Min VID: 0.7V
Max VID: 1.02V

Drop in Temp: 10 degrees Celsius
Increase in Batt: 20mins

Incredible improvement!

However, I've also noticed a slight increase in HDD temp (about 3-5 degrees). Has anyone experienced something similar?
Reply
Reply
post #258 of 1202
4/25/05 at 5:18am
erm has anyone encountered that the CPU load is always at 100% using task manager? im using version 1.5. i know many people have ansked this but there isnt any answer. At v1.5 the alternate cpu thing is missing and i've read that disabling the "run HLT command" thingy works but when u do the cpu will run at full speed.
Reply
Reply
post #259 of 1202
4/25/05 at 6:53am
Quote:
Originally Posted by qwerty007
erm has anyone encountered that the CPU load is always at 100% using task manager? im using version 1.5. i know many people have ansked this but there isnt any answer. At v1.5 the alternate cpu thing is missing and i've read that disabling the "run HLT command" thingy works but when u do the cpu will run at full speed.
It reads 100% because rmclock is making it read 100% by taking your excess cpu and running that dumb process you enabled with "Run HLT Command". If you turn that off, the task manager will report as normal.

The CPU running at full speed has to do with the power management settings in RMClock, afaik. You can set it to adaptive, minimal, maximal or none. Setting it to minimal should make it so that it doesnt run at full speed, but people with different notebooks seem to be reporing different things with rmclock.
Reply
Reply
post #260 of 1202
4/25/05 at 7:58am
If you set the profile to minimal, the CPU will always run at the "low" setting no matter the load.
When on automatic, VID and FID are selected in reguard to CPU load (determent from performance counters in the CPU itself). On (some) 855/915 systems USB root hub power saving functionality seems to cause those performance counters to report quite high load even when the system is more or less at idle. This is where HLT comes into play, forcing the CPU into C1 idle state when no programs are active. Task Manager doesn't know (and can't know, since it doesn't read CPU utilisation from the CPU itself but from the OS) that all RMClockHLT process is doing is forcing idle and therefore reports high usage.
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Notebook Guides
NotebookForums.com › Forums › General Notebook Discussions › Notebook Guides › Simple Undervolting Tutorial - Pentium M Dothan