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

Hi all,
I wanted to share my experience so far undervolting the new AMD Turion™ ML-34 processor found in my HP "LIVESTRONG" L2000 notebook.

My undervolting interest comes from the desire to reduce the heat output of the processor, which of course is the leading cause of electronic failure.

First let me state that I'm no undervolting expert. In fact, this was my first attempt ever. So if I make any mistakes, please forgive me.

Processor specs:

• AMD Turion 64 Mobile Technology™
• Processor clock speed: 1.8GHz
• Cache memory: 1MB L2 128K L1
• AMD64 core
• AMD PowerNow!™ technology
• HyperTransport™ technology (1600Mhz System Bus)
• High-bandwidth, low-latency integrated memory controller (Up to 3.2 GB/s memory bandwidth)
• Enhanced Virus Protection with Windows® XP SP2

I found the frequency operating range of this processor to be between ~800MHz - 1.8GHz. The nominal voltage range of the processor is between 1.0v - 1.5v. The nominal operating temp reported is ~50.0ºc. At ~55.0ºc the CPU fan will briefly come on until the temperature stabilizes.

WARNING – You can damage your laptop while adjusting the CPU’s voltage and or/multipliers. Your warranty will not cover such damages. I do not recommend anyone try and I'm not responsible for any harm as a result of trying to replicate my experiement.

Tools used:

• AMD Power Monitor
• AMD Dashboard Demo v2.0
• Prime95
• CrystalCPUID

I followed Mr. Dan Zhang's guide to determine my processors’ undervolting limits. (Recommended reading! )

My results:

IDLE
Original voltage:
1.0v
Minimum undervolt setting:
0.9v
Original temp.:
50ºc
Undervolt temp.:
49ºc




PRIME95 TORTURE TEST
Original voltage:
1.450v
Minimum undervolt:
1.175v
Original temp:
66ºc
Undervolt temp.:
55ºc (Corrected temp.)



As you can see, the AMD Turion 64 ML-34 processor takes nicely to undervolting at full load. A 11ºc (Corrected temp.) drop in operating temperature is impressive. I did not conduct exhaustive testing to determine the long term stability of running the processor 0.275v under spec but I did not encounter any problems.

Note – I was not able to reduce the idle voltage of the processor (1.0v) as reported by the AMD Power Monitor. The laptop became unstable when setting the voltage below 0.90v. At 0.90v the AMD Power Monitor still read 1.0v and the CPU temperature barely dipped compared to the stock setting.

Note – I noticed that the “Max Battery” Power Scheme setting had no effect on the CPU throttling (PowerNow). The “Home/Office Desk” setting would remove the CPU throttle all together and CPU would always run at 100%.

"I live strong"
Blksnake :hp:

 

Very nice reporting.

Too bad the idle voltage wasn't lower. I read somewhere that there is a conditioning process to properly break in a new processor. Maybe that could help you get your idle voltage lower, and your idel temp down.

 

Nice report!

I'm an engineer in the computer chip industry (used to work for AMD on memory and logic chips) and would like to make a couple of comments.

Under voltage won't hurt your CPU, but of course it could cause a lockup, or bad data being written to the hard drive which could cause a crash and require a system restore. I'm not saying it's bad to undervolt, just trying to be specific about the risk.

Low voltage is good to reduce power (increase battery life), and keep the temperature down (cooler on your lap). Don't be too concerned about the life of the CPU, modern chips are routinely run at 125C ambient temp and are reliable (not saying it's OK to always run that hot). The main failure mode, metal electromigration (metal lines on chip open up causing loss of an electrical connection), has been greatly improved with the use of copper instead of aluminum and improved step coverage.

So far as cycling or conditioning the CPU goes, I don't see how that would make a difference. We routinely measure min/max voltages a chip will run at before and after burn-in, and it very rarely changes (if it does change, that's usually a problem with our measurement, or a defective chip).

The fact you couldn't undervolt much at idle is kind of disappointing. Maybe that's why HP isn't offering the lower power Turion MT chips? Maybe AMD just can't make enough of them?

 

It looks like the Turions are able to use less power at a given voltage than the Athlon 64's. MUCH less power. An A64 at 1.4V has a TDP rating of 62W. Granted, that's overstated, but the Turion stays within 35W at 1.45V. I had assumed the Turion would run 1.2V at most.

There have been some amazing power consumption reports of the new Venice desktop cores, so I'm not completely surprised here. AMD has been making great progress on power reduction.

My Mobile-class A64 3200+ can do 1GHz at 0.8V. Maybe the Turions have to be run at higher voltage? AMD did say they tweaked the Turion's transistors for lower power consumption.

I'll defer to an EE to explain what the heck this all means .

 

I just discovered that I misreported the undervolting temp. at full load. My original post has been corrected.

66ºc Full Load temp.
55ºc Undervolt full load temp (Corrected)
11ºc Drop in operating temp. (Corrected)

 

Those are some awesome results. Have you seen any stability issues so far? I'm getting the 1.6 GHZ version in a while, so I can't wait to try the same thing

 

I curious to know how that ATI 200M video card runs. Could someone run 3DMark03 on your L2000?

There is also a newer 3DMark05, but most comparisions I've seen have been with the '03 version.

If interested, someone might want to try out a program like RivaTuner to see if the 200M clock settings can be tweeked.

 

55C is when the fan kicks in, so basically as soon as the fan starts it stops the temperature increase cold. Nice.

 

While running the processor full load (Prime95 torture test) the fan would turn on and stay on during the entire test. I found the fan to be "noisy".

While undervolting the processor I noticed the fan slowed and became quieter.

I may perform more exhaustive tests in the future and monitor the fan speed as well before and after undervolting.

 

blksnake, those are some excellent results. In mine, I've got the ML-37 (2.0ghz) and couldn't pull quite the same numbers. 1.300v was the lowest I could run at full speed and 0.925 is all I could get at the low end (800mhz).

Temps at the high end maxed out around 71-75 degrees before undervolting hit ~65 degrees after. At the low end I pulled roughly the same numbers listed above. It is interesting to point out that at 1.8ghz I could run the same voltage as blksnake, 1.175v. I wonder if those new ML-40 chips are just AMD realizing their chips are coming out a lot better than expected and can run much faster.

For me, I'm just glad to see that I can lock the thing at 800mhz, kick my feet up on the couch, and the surf the web without the fan coming on or my legs getting 2nd degree burns. It is a little warm, but definitely not too hot for the lap.

 

Hi Renoguy!
3DMark is meant to test the biggest and baddest equipment (CPU, Video cards) to the limit. This laptop's GPU (ATI X200M) is not meant to compete in the same category as the Nvidia 6800GTs or ATI's X850XT... With that said, it can offer decent performance in games if the graphic options and resolutions are set to low. My 3DMark03 score was only 767.

ATI's PowerPlay option in the display properties will adjust the GPU's speed from 100MHz (Economy mode for longer battery life) to 300MHz (Performance mode) depending if the laptop is plugged-in or not. I don't have any interest in overclocking my GPU for a negligible performance gain.

 

Thanks for the test, blksnake. While the ATI X200M is much better than Intel graphics, I'd of expected better than a 3DMark03 score of 767.

I'd been debating between the L2000 and some Centrino Sonoma notebooks. I ended up finding a great deal on a Sony S460, and got it. It's very nice in some ways, but I think it gets much hotter on the bottom than the L2000, but cooler on the top. And it's fan runs a lot too, but I've made some adjustments to improve things.

Anyway, it has the nVidia Go6200 GPU. While it too shares memory like the X200M, unlike the X200M it has it's own dedicated 32mb memory built in. So I'm getting 3DMark03 scores around 2600, and can get over 3000 with some overclocking of the GPU. But I hear it's not all that great for games either.

Amazing how rapidly things change regarding games. My last notebook had dedicated ATI 9200 graphics with 64mb RAM and was supposed to be good for games, but it only scored about 1100 (which was good for the time).

If you are really into games, guess you need to buy a new laptop every 6-12 months!

 

I had a Sony S460 and it was definitely hot or hotter than the L2000.

I sent it back and got the same config in the L2000 at 1/3 the price.

The GPU is a little better, but not enough to make it gamer, etc.

The non-replacable hard drive on the Sony S460 was the killer for me.
I also love how the three USB ports on the L2000 are spread out around the machine, instead of squished together like on the Sony.

Weigh wise, the Sony is a little less, but it you take a S460, with an extended battery and the AC adapter, and compare weight of L2000 with extended battery and AC adapter, the overall weight nearly even. Also, you get full built-in card reader with the L2000, and much better speakers.

And that doesn't even touch the 64 bit processor which will be nice one the move to 64 happens over the next 2 years (on the software side).

 

blksnake, do you have crystal CPUID in your startup menu? and do you know if there is an easy way to make it automatically enable the multiplier/voltage settings without having to right click and select every time i boot up? thanks

 

Use the /CQ /HIDE options in the Startup folder shortcut.

 

I switched to RightMark CPU Clock Utility (RMClock) 1.6 to create and change my voltage states at different multipliers or P-State transitions (PST). I found it to be easier and more practical to use.

It can be found here:
http://www.softpedia.com/get/System/System-Info/RightMark-CPU-Clock-Utility.shtml
It has an option to " Run automatically at startup".
Good luck!

 

Since you have a faster processor to begin with (which I assume has the same default voltages as the 1.8GHz version), you would naturally not be able to undervolt it as much at the top setting.

 

I have the 1.6 GHZ processor and I was only able to get to 1.175v as well...I could do 1.150v, but that's borderline since one step below that caused my system to crash. The low end was also only .925v. I was hoping for lower, but I think that'll do.

 

i have an acer 5021 with the 1.6 ml-28.
at 4x, minimum voltage is 0.9
at 8x, minimum voltage is 1.075

i have it set at 0.925 and 1.1 just to have an extra margin.

 

Those are some awesome numbers...wish mine coulda done that

 

relatively awesome. i'm pretty happy with the 8x. but deep down i was hoping for the "one chip that rules them all," one that could under to almost p-m levels 0.7v at 4x would be increibly nice...

 

Hi All,

On my L2000 (2.0 GHz), I've tried to undervolt to 0.95v at 4x (800Mhz).

It does put a brake on fan cycling...however, I get system unstable
after some use. The screen goes blank with some interesting color
patterns (like stripes etc).

It has happened quite a few times with 0.975v and 0.925v as well.

Any suggestions?

Imran

 

Verify your undervolting settings with a program like Hot CPU Tester Pro. Some chips undervolt better than others.

 

Use a CPU stress test program while you undervolt. By doing so, you can tell if your voltage will cause any problems under usage.

Prime95 is one option.

And I'd recommend you follow along with laptoplogic's article on how to undervolt. cpu undervolting

:amd64:

 

Thanks,

I'll try the suggestions above.

Imran

 

It may be a stupid question:

Can anyone explain the three settings "automatic management,
minimal, maximal" under Performance/Power profiles?

If I want maximum battery life, cooler (fan free) operation,
should I select "minimal"? I'm confused.

Thanks

Imran

 

If you're running off battery, I'd select "minimal", which will keep it at the minimum multiplier/voltage you set. If running off of the AC adapter, automatic management should be fine. Also, it probably won't be fan free on minimal...but it will decrease the amount of time it cycles on. Hope this helps

 

Does it mean that on 'minimal' setting it will not go past
4x (800Mhz) no matter what's the CPU load?

Thanks for explanation.

This forum is awesome!!

Imran

 

No, it means it'll use the slowest speed necessary. It'll jump up to full speed as needed. They really ought to call it "normal".

 

Hi Brian,

You are right. THough it stick to lowest speed 98% of the team.
But I have seen it jump to 1.6Ghz occasionally.

Thanks for clarification.

Imran

 

I read on RightMark forum that it's better to have
"Always On" power mgt scheme setup with RM Clock
Utility. Someone reported occasional system freeze
fixed by this.

I've been trying this (Always On) with minimal mgt
in RMClock Utility. So far, no system freeze. I've my
fingers crossed.

Imran

 

Does anyone have any idea why, under XP Pro, when I start a game, it seems to cancel out the effect of crystalCPU thingy?

Also, I can't seem to get RightMark to work for me - it won't recognize the turion processor.

 

Make sure you have the Windows power management set to Always On or Home/desktop, that solved that problem for me.

Rightmark should work for you anyway, it doesn't recognize mine as a Turion either, but as long as the FID/VIDs show up, you should be fine.

Lastly a question for anyone using RightMark on a ML-34, I set PST 0 to 4x and 1.000 V, however, on the first (General) page of Rightmark it says Minimal 4x and 0.800 V. When I select Minimal for the Battery profile, will it go down to 0.8 (and lock up) or just go to 1.000 V? Or did you guys set up the batter profile in a different way?

Thanks!

 

So, I needed to get rid of the brightview screen on my L2000, and re-ordered the same machine, same specs, everything.

On the old one, I could undervolt the machine down to 1.175 at max CPU. On the new one, I only seem to be able to undervolt it at maximum to about 1.225 before I get the crash (It crashes at 1.200). Granted, this does bring the CPU temp to below 55 degrees, which is nice...

Should I quit while I'm ahead? If I'm running mostly word processing on battery, does it really matter how low I go?

More importantly, is there that much variation in the individual chips?

 

Yes, the amount you're able to undervolt can be different between chips.

Undervolting my Turion ML-34 processor only made a substantial difference under a full load. (See first post for more details)

Quit while you're still ahead

 

In reguards to the 3dmark earlier in the post, are you sure you were referring to 3dmark 03?? There is a 3d mark 2001 and 2005 but I havnt heard of 03. that 2700 # you were getting sounds like a 01 # as thats what I get with my 200M. 3dmark 05 would get a low # like 750 for the 200M.

 

I've been using Speedswitch XP for a while now with good success keeping the CPU down in the 800Mhz range on Max Battery. Yet battery life could still be better so I'm looking to try to use the RM clock utility instead.

I tried using RM clock without monkeying with any of the settings and found that its actually quite stable and seems to be doing a better job than speed switch at keeping things cool and battery life longer.

Whats more, I can have stages of performance unlike with Speedswitch. I was able to put in intermediat multipliers for 5, 6, or 7. Then set voltages for each one heance allowing for even more precise speed / voltage control.

Presently after using the program, the CPU fan starts up alot less when its set to conserve power when plugged in and the fan doesnt turn on at all when on battery. I managed to get the Idle Minimal at 4x down to .950v, any less and the sytem is unstable underload. You wouldnt think that reducing the idle voltage by .05 woud do anything, but I think the system is locked to that voltage and so needs the fan less making things quieter.

HP L2000, Turion ML-30, DVDRW-DL, 60 GB HD, 1GB RAM, BrightView, Wireless.

 

I got my V2000Z last night and was able make some progress on UNDER VOLTING with excellent help from this thread.

My Configuration is: ML-37 , 256x2 DDR333 Samsung chips, 40GB Fujitsu 4200RPM HDD.
I used RMClock to under volt at 4x and at 10x (Min and Max).

My results:
At 4X(800Mhz) 0.925 is the least I can go, but it stable at that, ran Prime 95 for an hour. Anything below that I am getting screen artifacts.

At 10x(2000Mhz) 1.225 is the least I can go stable. Ran Prime 95 for 2 hours. at 1.2 Prime gave some Warnings after few minutes.

I did not monitor temp at this time. But looks like even at 10x it is mid fiftees to high fiftees.

Anyways, during this test my room temps were around 68-70F.

But I have a question for all Undervolt gurus, Can I leave those lowest setting like that at 0.925 and 1.225 or do you suggest me to up voltage by 0.025 for safety like 0.95 and 1.25 ?

Thanks
Chinna

 

There is no safety issue, its stability. run it at the lowest voltage you can without crashing. you can only harm your cpu if you overclock it, which you cant do with the turion anyways, at least with most laptops.. and even then its fairly safe.

 

Thanks for the quick reply.

I mean, safety from crashes. Like when I tried 0.9 volts, I got vertical color stripes on my screen. But was working fine at 0.925 even after 1 hour of Prime95.

In this scenario processor temp remaining 50-55 C at current ambient temps. but assume there is a raise in ambient temp, that will automatically raise Processor temp too. In that case is there a chance of crashing at that border voltage?(Like you know for OC we do that go to the border line and up it little).

What I am not sure in this UNDERVOLT case is, Will the temp of the Processor affects the stability at the given voltage? ( This is my first UNDERVOLTing, though I did lot of OC).

Thanks

 

if it crashes, up the voltage slightly till it doesnt lock up...

 

After all these tests and undervoltage, can anyone say by how much the battery life increased in minutes? In another post, a guy said his L2000 with a 12 cell battery last 30 min longer than before. Does this mean that the 6 cell battery will make it to 3 hours or at least be close?

 

If you are just browsing or using word and cpu is running at lowest step it may not add more than 5-10 min I think.

But if your CPU is jumping around a lot going above lowest settings you may well able to add more than that.(First thing I did is removed/uninstalled every dam thing HP Installed on it )

Please see this thread for more details in how much you can save
http://forum.notebookreview.com/showthread.php?p=1152685#post1152685

The standard battery rated in around 53Whr if I remember correctly, and supposed to run for around 3 hours assuming battery is holding full charge.

Infact I am able to stretch it to 3 hours(Plus/Minus 10 min).
Mine is V2000z with ML-37, running at 0.95v at 800Mhz, 1.25v at 2000Mhz and RM CPU is set to Minimal and created 7 mulitpliers(4x to all way to 10x). Fan never came on, not even once in 3 hours!

Hope this helps and let us know if your UNDERVOLTing helps you increase battery life.

 

Here are some more numbers I am able to attain after doing some more UNDERVOLT experiments.

[email protected]
[email protected]
[email protected]
[email protected]
[email protected]

These are the stable voltages I got. I ran Prime95 atleast an hour against each voltages.( 6 hours at 4x and 6 hours at 10x).

Based on my calculations, at these voltages my CPU may be consuming around 16 or 17 watts at 8x(1600Mhz which is equivalent to ML-30 but at 16-17 watts ) at full load at that stepping.

In all these tests temp never raise more than 55 C, which is when Fan kicks in and stops at 50 C. ( Room Temp around 70F). Even at 10x with full load fan never need to run continuous.

While browsing and word editing, fan never came on.

Sigh! Relief, my previous toshiba with P4 2.66Ghz has fan keep coming every minute even when just browsing. Bottom used to get hot after some time.

 

Hi all,
I have recently bought an Acer 5021 with a turion 1.6gHz cpu and I can´t seem to be able to undervolt the cpu using RightMark cpu utility.

The fid and voltage fields are grey and I can´t change their values.

I have the latest amd drivers and I am logged in as an admin.

Does anyone know what could be the cause of this?

Thanks in advance,
Johan

 

I hope you already figured out how to do it. If not, go to manangement tab, check the use P-Satet Transitions(PST) option. Click on Add button , on the popup windows select the multiplier and voltage. By default it calculates voltages for all intermediate states.

In the management tab you can select Performance/power profiles for AC and Battery operation. Minimal means it will try to run CPU at lowest mulitplier, but increases the speed when required. At Maximal, it runs at full speed irrespective of load.

Anyways, best of luck with your undervolting, and let us know your results.

 

Mine is AMD Turion ML-28 (lowest one I know. But I'm poor )

I made
FID 4.0X VID 0.900V and FID 8.0X VID 1.050V
so far it works just fine. No reboot no crash.

I made 4.0X 0.875V and 8.0X 1.025V. it worked but the windows will restart for no reason in like about 2 hours of use.

So I tried to raise them 1 step higher and now it's very stable(with so much power saving ^^).

2 hours of 3D game (warcraft3 battle.net) 4 hours of web surfing and 2 hours of burning roms and watching Anime. That's total of 8 hours usage and no problem so far

test results
I used Battery Eater '05 Pro to test the battery life by making the CPU and GPU run at full speed (Open GL render)

Tested Notebook
Compaq Presario V2000z
CPU AMD Turion64 ML-28
RAM 1Gb (2x512) DDR PC2700
HDD 60Gb 5400RPM
12 cells LION battery
Wireless Lan Disabled

Before Underclock = 131 Minutes
After Underclock = 240 Minutes (That's 4 hours )

 

240 min. uhaaaa!now we're talkin'

 

WOW! These are the best UNDERVOLT figures I saw for Turion. And 4 hours of battery life!! WOW. I am feeling dizzy now!!

Congrats for excellent chip you have!!

 

And yes, I'm not lying. The result was real

1 thing to discuss here is that the (2 times) increased battery life is when the CPU is running at full speed.
I think the battery life will not be that much increase for idle/minimal speed because, as you can see

maximun VID reduce from 1.45 (or 1.5?) to 1.05. That's 0.4+V reduced.
minimum VID reduce it from 1.0 to 0.9 (only 0.1V reduced)

so I don't think it will be much diferent if we run the notebook at half speed all the time. The battery life should be a little bit longer but not much longer.