Undervolting the AMD Turion 64 ML-34 processor (L2000)

What are the setting you have in RM Clock for Battery Profile (minimal and Power Saving level).?

What I observed is CPU clock would not increase CPU speed with artificial loads like Prime95/battery eater pro with the above settings. So proc runs at minimal speed though it is taking 100% CPU.

Another thing is, in your case voltage difference between Min and Max is not that much, so your power consumption difference is only about 10 watts from 800Mhz to 1600Mhz.

 

pompoko, how does this affect your cpu temps? does your fan turn on significantly less doing simple processes (web browsing, etc)?

 

The profile in my RMclock

Battery Profile : Automatic management
Battery mode : Power saving LV4

and I set the windowXP's power management to Home/Office desk so it won't interfere with the RMclock's CPU stepping management. (no double stepping just in case)

The CPU clock did run to 1.6Ghz if the applications requested for hard load. I checked with
1. Mobile Meter
2 AMD's Power now! dashboard
3.The RMclock itself

They all reported that my CPU is 100% load and is running at 1600Mhz so I think it's no mistake.

The AMD Powernow! dashboard also show that my Battery savings guage is more than 50%.

I'll include the pictures about before running Battery eater and after running the Battery eater.

And about the CPU temp. It's around 50 C. but the fan didn't turned on so often when running at full speed like before underclocking.

 

If you want to generate lowest heat while running on battery, select Battery Profile: minimal. This way CPU speed do not jump around too much.

 

i got the same cpu that you have but i couldn`t get same numbers. Using your number i had BSOD all the time i had to raise them about 0.050. what exact kind of cpu you have? what serial number or something. or maybe you know something about different models of ml-37...

thanks

 

I do not know the processor serial number. Each processor is unique, some do verygood at UNDERVOLT, some hardly support any UNDERVOLT. This is just luck. In another thread one guys was able to run his processor at 0.90v @4x and 1.05v@8x that too for ML-28. Which are I think extremely good( from default 1.45v to 1.05v!!).

Even if we know the Proc serial number, there is nothing much we can do. It is not like desktop where you can buy MB and processor separately.

I am sorry I could not help much. I too got crashes when I tried 0.90v@4x and 1.20v @10x, where as blksnake was able to do verygood.
Sorry.

 

That happens to be me?

So far no crash no reboot. It's just as stable as the normal voltage (with no performance loss.). I can still finish loading DOTA map at the same time as before underclock.

Now, for me high load or minimal load is kinda the same
because...
high load = 1.05V
minimal load = 0.9V
difference is only 0.15V which is a small number IMO

What I need to know now is that 4 hours battery life (full load nonstop) from a 12 cells LION battery is good? I expected for it to be like 5 hours+ (greedy ) anyone has info about this?

 

At full load a 6 cell will not last more than 1:10 min to 1:20min min. So for 12 cell it should not be more than 2:20-2:40min is what one should expect.

And you are getting 4 hours and not satisfied!!?

 

lol I must be too greedy

so my guess is that if I do normal works it should last like 4:30-5 hours (maybe) because as you all know.... my high and low voltage are so close. and HP's spec that a 12 cells will last at the maximum of 7 hours (in an ideal test environment) for a V2000. So in real life it should be 5:30-6 hours at best.

In conclusion. I guess a Turion64 cannot beat a Centrino at battery life. no matter how hard we try?

*** I love Turion because centrino's logo looks cheap (it looks more like an airline company's logo more than a technology logo). Funny reason eh?

BTW Chinna_n u live in Seatle? I'm in Bellevue but I go to work in Seattle every day. maybe we walked pass each other every day? lol

 

Here's what I could get of Turion ML 34 1,8 ghz 1mb l2 cache, l2005.
4x - 0.925 (white and blue lines at 0.9)
9x - 1.225 (at 1,2 Prime95 torture test failed after a while; the notebook did not show any other signs of instability!?)

I created an intermediate step 6x, which RMClock automatically calculated at 1.050. I'm wondering if I should try removing the automatic calculation and try to lower this one? Further, should I create one more intermediate step, say 8x? I'm not sure what the benefit of intermediate stepping is as I cannot see the processor going above 800mhz (I've put the AC Profile at Minimal, as explained in the "V2000z fan problem" thread . Chinna_n, did you say that RMClock seems to distinguish between testing and continuous load?

 

It's been a while since I checked this. Sorry about that.
About RMClock seems to distinguish between testing and continuous load, I am doubt that now. Probably I was confused with two things in RM Clock settings(Management Settings and Performance Profile).

Performance Profiles let you choose between Minimal ( Locked to Min speed), Maximal ( Locked to Max speed), and Auto ( varies speed).

Management Profiles, PowerSaving levels and Performance Levels changes the Threashold CPU usages for stepping up and Stepping down CPU speed(So only comes into picture when Performance Profile Auto is selected). Currently these are hard coded values in RM Clock based on Management Profile. PowerSaving Level 4 try to keep to CPU more optimized for battery.

This RM Clock settings has been confirmed by RM Clock Author/Developer Dmitri Besedin. I requested him for the provision to change/set Thresholds for CPU stepup/stepdown CPU usage and threshold times. He said he will try to provide custom settings for next release.

When using RM Clock make sure you select "Always On" power scheme. Home/Office Desk can get you into hangups. Windows controls processor speed with Home/Office desk also, which causes conflicts with RM Clock.

BTW, you could create 4x,6x,8x,9x multipliers in your case.

 

Alright so I'm thinking about going through with trying to undervolt my V2414. But, before I go ahead with this, I noticed something. If you single-click on the battery in the system tray it brings up what looks like 7 Power Configurations (Home/Office, Portable/Laptop, Presentation, Always On, Minimal Power Consumption, Max Battery, and Win DVD). Now when I have this on Max Battery only, it drops the processor speed down to 800 MHz. This is evident when you look at the General Tab under System Properties and this seems to be the only setting in which this is done. This is the only noticeable difference so I am not sure what else it does to try and reduce battery consumption. I guess my question would be is what are the differences and advantages to using each method (undervolting and changing the configuration to Max Battery) to try to increase battery usage?

 

By undervolting you give the cpu less volts for the same amount of work. That is, you can run at 800mhz at 1v (or 0.975) with the original Windows stepdown configuration or at 0.925v with RMclock. This means that you do not give up performance /both cases 800 mhz/ but gain battery life /0.05v or more savings/. The clock speed is not under primary consideration here; its the voltage. By undervolting you discover the minimum amount of power the cpu needs to run at the given step and at that amount it generates less heat.
I don't know what's your processor but it seems that it could be celeron /lazy to search for the model/. You cannot undervolt a celeron.
Advantages - the fan would not kick off that often. For some basic stuff, it should not kick off at all. This depends on the processor, however. Further, the cooler the notebook runs, the longer it will live.
Disadvantages - hm, I'm not particularly fond of having too much processes running in the background and the RM utility would be one more. Other than this, I don't see any. If you find the stable operating mode, you forget about the undervolting and simply use the laptop.
Ask if you have questions about undevolting; you can find all the steps and all the progams in this thread, however.

 

Thanks. I got it now. My notebook has a Turion ML-28 so when I get some time I'm going to do this real soon. From looking around though it seems that maybe undervolting the ML-28 may not produce substantial results since it is already a slower operating processor but it's worth a try. I know it's a low processor but it fits my needs well as a second/portable computing option. This being because all my money is going towards putting together a Pentium D system. But that's for another thread....

 

Better read this first. HP has a fairly decent Athlon 64 X2 desktop if you don't want to build your own.

 

I would be careful buying a Pentium desktop. My cost-benefit analysis totally rejects it

 

Actually, ML-28 may undervolt pretty good. Atleast that is my observation sofar. Like ML-37 may only work at full speed at 1.25V volts, but most ML-28/30 can comfortably work at 1.2v at full speed(1.6Ghz). You get most benifit out of this, by reducing lot of heat as well as improving battery life.

Best of luck with your undervoltings, and let us know your results.

 

actually ML-28 can go as low as 1.050V at full speed without any problem and make the battery life increse to 4-5 hours on a 12 cells battery

any of you know what's the battery life of a V2000 (centrino) with 12 cells battery like? I just wanna know if our Turion64 can somehow compete with centrino with battery life or not.

 

I fortunately got a really good price on an Intel Pentium D. I ended up paying around $230 for the Pentium D 820 (only option), 945PSN MB, XP Pro and some adobe elements software and that included the shipping. I couldn't pass up the price and for that amount I might as well build my own. Other wise I would most likely be buying a pre-built system. Although I haven't built a computer yet so this has been something I've been wanting to do for awhile. I just received my video card in the mail today. ATI X700 for only $99. I just need to get some memory, a sata hd or two, and a case/psu and i should be up and running and with not that big of a dent in the pocket either. Fortunately I have some spare CD/DVD roms laying around. Any recommendations for getting good prices on HDD and DDR2 memory. I've been online to a couple well known places (newegg, etc) but I'm still looking for some better prices. Sorry for getting off topic.

 

I just got done undervolting the laptop the other day. Using Prime95 and RMClock 1.8 the results said that I could get 0.950(4x) and 1.050(8x). But when I fired it back up I kept having problems with Firefox crashing so I took them each up a notch but still had the same problem. Right now I have the 4x at 0.975 and 8x at 1.150. Definitely noticing the difference. The battery meter doesn't tick down every few seconds, my lap doesn't catch on fire, and the fan hardly ever kicks on now. Thanks a lot guys for the help.

 

I'm really happy you managed it and like it. Yet, try going to 0.925v /I think even 0.9v for ml-28!/, set the powerscheme in RM to Minimal and test with Prime95. Mine /ml-34/ cannot handle 0.9v but is perfect at 0.925. Since most of the time you are using the cpu at minimum clock /800 mhz/ the greatest savings are here!!! I run mine at Minimal all the time and switch to Automatic only when playing games (windows power scheme always on!)

 

yeah so i tried getting it down to 0.950 running at minimal on the battery. worked great for about 30 min after i tested it using prime95 and then the sucker just bsod'ed on me. hmm...well we tried. i'd rather have it happen while i was posting on here rather than typing up that report due later on today.

 

ML-28 can do 0.90V for 4X (800Mhz)

mine is 0.90V at 4x and 1.050V at 8x

no problems so far it feels like a normal CPU but with much greater battery life (set the windows power scheme to Always On or it will freeze from time to time)

 

Tried it again real quick with the Always On scheme. Still didn't work. It locked up at 0.925. It seemed to be working alright during the Prime95 test but I ran a few programs to get the HDD going and that's when it seemed to cut out. Not sure why this is because I though the HDD was powered indepenedently of the CPU but I could be wrong. But i guess my final settings are 0.950 at 4x and 1.100 at 8x.

 

Not all CPUs are same. So everyone UNVOLTING figures may vary. In general ML-28/30 UNDERVOLTS better and that is why you got good voltages( 1.15v instead of 1.45V at 1600MHz).

BTW I think pompoko is extremely lucky with his processor. Not many people got those voltages( no else I saw).

 

this made me (somehow) feel good about my lappy :hp:

:centrino: <=== can someone tell the admin to make something like this but with Turion's logo instead of this centrino thingie

 

Thank you to all of you. I learned a lot aout undervolting from this thread.
My ML-28 results are 0.925v at 4x and 1.125 at 8x. Has been rock stable for 3 days now.
Temp went down a lot. It hardly ever reaches 58C now. Battery life seems longer but have not actually measured / timed it.

 

Just wanted to share my undervolting results.

I have the v2000z, ML-32, 1280 GB of RAM, 60GB HDD.
Using RMClock. I got:
4x - 0.900V
9x - 1.150V

I also have 6x and 8x intermediate steps with voltages selected automatically. Been running stable for more than a week now.

 

I think verygood results, even 9x voltage is good.

 

Has anyone tried tweaking the memory timmings???
Actually, Im using lower timming than default ones.

For example, I disabled 2t Timming, lowering Refresh rate Tref, reduced Read preamble, and changed other timmings... wish resulted several seconds less in SuperPi.

 

finally I've decided to undervolt! so far so good

 

What are your results, klas? I see you are considering MSI? Is it for the MT processor?

 

finally figured out how to use it lol =) im at 43C right now

4x .9
8x 1.125
9x 1.15
10x 1.3

no clue how it is lol just changed it, seems fine i dunno

 

check this thread

http://forum.notebookreview.com/showthread.php?t=33962&highlight=RMClock+settings

anyways good to know that understood.

I was wondering if it is stable at 0.900v at 4x?

And if you can do 9x at 1.15, most likely you could do 1.25V at 10x.

And if possible add 6x stepping with 1.0v, it helps a lot.
I have 4x, 6x, 8x, 9x, and 10x.

 

Well I'm currently discharging because my wear is at like 28% after 3 days, i don't know how to really test it, temp dropped again 42C =) thats like my cpu temp for my desktop lol

 

Soul814, you have a really good chip if it's stable at 0.9v! I wonder if there's any chance for the memory to affect the results. I know it sounds stupid and crazy but so does the positive memory effect for SuperPi. Is it possible that the CPU would need less power if the information that reaches it comes in different bits/chunks?
I'm probably talking nonsense but it just came to me.

 

well it seemed alright on .9 i changed it to .925 for safety, i've been using my notebook on these settings for about 2 hrs now
4x .925V
6x 1.025V
8x 1.125V
10x 1.275V

 

Did you try it with Prime 95 for a few minutes? That's the real test

 

prime 95 nope i'll dl that now =)

-- what u do is run tortune test right? any settings i have to set? I ran it a few mins and it seemed stable, i did like 3 mins, temp went up to 57C

 

you're gonna have to run it for at least 4-5 hours... some say more

 

You need to run Prime95 for typically more than 12 hours( 24 hrs typical), but 6 hours is minimum, because it is proven many times errors do occur before 6 hours.

After 6 hours it is high unlikely we get an error, but people do run it to make sure( I do 12-24hr).

I would suggest 6 hrs atleast.

And for testing CPU, select Maximum Power/Heat option in Prime95. also try to run it on AC and without Cooler, so we will also know if temp related issues are there.

 

HP Special Edition - ML32

1.15v @ 9x
0.875v @ 4x

The notebook did crash as soon as I lowered to 0.85v @ 4x. I was hoping to get down to the 0.8v range for max batt life.

 

WoW! those are very good results. But I would suggest bumping the voltage by .025v from the lowest stable voltage( it is recommended as sometime it causes hangs during transition).

So in your case 0.90v at 4x and 1.175v at 9x.

BTW it is good idea to have few intermediate steppings,( 6x,8x etc)

 

This is really amazing. Are you sure it is stable? Mine is not a very good undervolter but still the best I can get at 9x is 1.2. Quite a bit of a difference from 1.15

 

I have been running Prim95 torture test at both max and min clock speeds. Never crashed once during the testing or normal use. I also have those two check boxes checked:

1) CPU Low Power Enabled.
2) Northbridge Low Power Enabled.

I think I am just very lucky that the lastest AMD Turion were made even better for under-volt. (Purchased 5 days ago)Maybe AMD is ready move the voltage down across the board again. From the postings, it looks like most ML chips are qualified to be MT anyway. AMD should have skipped ML from the very beginning.

 

hm, I don't have the second thing checked...did you check it yourself or it came as default? This is how my RMclock looks like (see pic)

I think it's more fun with undervolting The MT series would be kind of boring

 

I checked "Northbridge Low Power Enabled". It is not on by default. However, I don't think it has anything to do with the under-volt result.

 

im stable at 9.0 @ 4x, i should try knocking it down a bit lower and testing that out

 

No, these hangs I am talking does not happen when you prime at that stepping, but rather during transition, like from 4x to 9x jumps or from 9x to 4x jumps, processor is unable to stabilize with border voltage.

Anyways, RMClock 2.0 has that feature built-in to provide stabilization voltage to proc.

So, which version you are using?

 

I absolutely agree with you
it is lot more fun to bring voltage from 1.45v to 1.15v. MT can hardly undervolt (the best for MT-34 is 1.1v from default 1.2v that too on desktop MB).

So, in the end undervolted MT series and Undervolted ML series both use same power. Better yet, uses even lesser at 4x than MT.