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

Thanks for the suggestion. I will vary the load to get the voltage swings. I have done some of that since I am in Preformance on Demand mode but more extensive testing would be prudent.

 

Hi! I bought an Acer 5024wlmi for my girlfriend, and she gave it to me for 2 weeks to make fresh install and to thoroughly test it. Of course I had to underclock it

Currently I am testing it on 4 x 0.875V and it looks stable for now, on 0.850V it just hangs:



Regards, VidKo

 

0.875v works for me for a while but it would randomly reboot until I up it to 0.9v. It is solid at 0.9v. I have HP Special Edition L2000 with ML-32.

 

abugi, yep, I had random lockups too at 0.875, at 0.900 is rock solid. At 1.8GHz rock solid at 1.150V

Regards, VidKo

 

Stable?

I'm running my ML-37: [email protected] right now ... using Prime95 to stress test. Things actually seme to be OK. Haven't noticed any instability. Anyone successfully running on less power? Are most able to run at .900, or do most run higher? This is obviously much better than running at .975

Edit: This is down from [email protected] for me

 

A number of chips are able to do fine at 0.9v but probably most of us are at 0.925v. I'm actually surprized the ML-37 goes as good as 0.9v! I had the impression that ML-32 (or 34) and lower could undervolt better at 4x, that is 0.9v. Maybe we should set up a poll to find out

 

First I have to say THANK YOU, for this thread, I was just googling as I had bought a new Acer TravelMate 4402, I love it because of the price and the features, the x700 ati video card is the major selling point, because with the price of this laptop there are almost no other laptops that have a dedicated video card (though its only 64mb) its great, upgraded the laptop with 1gb ram adding of a total of 1.5gb

As I was reading the newegg comments about it, some ppl reported that it would run hot, and as I found out its not hot for me, but the cpu did peak out at above 75C which worried me a bit, idle 50-55C (the room temp is quite hot here, in AZ)

The cpu is turion ML-30, which too bad its not the MT, but after reading the main thread I couldn’t believe that I could change the voltages, I had suspected it, and thought that the values might be too high (1.450) as reported by RightMark, but other applications round up, showing 1.5V

Right now im testing it as low as 1.05V (system BSOD’s at 1.025v) this is amazing as the full load with prime is at about 60-63C and fan only turns on at about this range, the lowest voltage as many others have said seems to be .9V anything lower BSOD’s even if I find that the stable voltage are a bit higher then these tests, im still very happy with such results now lets hope I can find linux software that will let me do the same

 

so far i had tested 4h and 30min prime95 so its pretty good, at most though to be on safe side the voltage will be at probably 1.075, if any instability issues show up with 1.050v.

CPU turion ML-30 (LanCaster sh-e5)

 

Just to add to the list, I have a Compaq M2105us (bought on Black Friday for $400) that I upgraded with a Turion MT-40. After 48+ hours of flawless Prime95 testing, I've settled on the following for the time being (using Performance on Demand):

4.0x (800MHz) : .850V
5.0x (1.0GHz) : .900V
6.0x (1.2GHz) : .950V
7.0x (1.4GHz) : 1.000V
8.0x (1.6GHz) : 1.050V
9.0x (1.8GHz) : 1.100V
10.0x (2.0MHz) : 1.150V
11.0x (2.2GHz) : 1.200V

I know the ML series isn't bad, but given what I've ended up with, I'm glad that I went ahead and bought the MT-40.

 

I think that is really good for ML-37 ( I was able to go 0.925v, but I want to have some buffer, so kept at 0.95).

 

Great results. Is it stable at .85? Did you upgrade easily the processor? So far you are the third person, I think, to have upgraded from Sempron/Ml to MT. Well done, I just wish someone had taken pictures of the process

 

It's absolutely stable at 4.0x and .850V. Indeed, while I ended up with a vertical line lockup at .800V, the CPU booted and entered Windows just fine at .825V. Because .825V was right at the line, though, I didn't test it and instead went with .850V. As I said, at .850V it has done about 24 hours of Prime95 flawlessly. .850V also has the benefit of giving a nice progression of voltage up to my current max (11.0x and 1.200V). Perhaps if I'm feeling greedy in the future, I might drop down to .825 and see how far I can go with it.

I've intended to write a review of my notebook that goes through the upgrade process a bit. Hopefully my newborn will give me the time to finish that up at some point. To be honest, though, even without pictures, it wasn't very tough to upgrade. The key is to print out a copy of the excellent service manual and carefully go through it's CPU replacement guide one step at time. The main issues I contended with were what to do with the thermal pad - I kept using it while applying a very thin layer of Arctic Silver - and contending with the loss of Powernow!, something that led me to CPU RightMark and this very thread.

In the end, along with 1GB of PC3200 and a 7200rpm HDD, I have the snappiest notebook I've ever personally used, and some sweet voltages to boot. Not bad considering that everything put together cost less than a grand.

 

Looks great. But I think you could do even better in intermediate steppings( as these are more like ML series at this time, esp for 8x to 11x). Ofcourse, I am greedy

But I was wondering, why did you use thermal pad along with Artic Silver.? Any reason? is it a problem to fit without pad?

 

I can get greedy myself, but considering that I've just gotten these voltages tested out and set this week, I'll probably let it go for a while. Besides, using the Performance on Demand setting, CPU RightMark does all of the intermediate settings on its own. Do you know of a way around that? Regardless, I'm not sure that the intermediate settings are of all that much importance. In my testing to date, it seems as though 99% of the time, its either on the minimum or maximum levels.

As for using the pad along with the Arctic Silver, going in I couldn't have imagined doing that. It certainly isn't the most efficient means of doing things. Once I got the notebook torn down to the CPU, however, a couple of things stood out. First, the stock pad was a bit thicker than I had anticipated, and I worried that not using it would require a thick coat of arctic silver, something I don't like doing. That pad likely maintains better contact over time than a thick layer of paste would.

Second, while I could have just stuck with the pad (which was quite adequate for the stock Sempron, after all), the size of the core on the Turion is larger and didn't quite fit the indentation on the pad that the Sempron had made. I was concerned that the old indentation, therefore, would lead to gaps between the Turion's core and the thermal pad. It was because of that that I decide to apply a very thin layer of arctic silver to prevent any such gaps.

All in all, it wasn't the ideal solution, but in practice it has worked out fairly well. Heat is clearly transfering OK, since my fan (once I got RightMark working, at any rate) runs no more often than it did with the Sempron.

 

In RMClock, you can uncheck automatic intermediate option in popup FID selector and select your own voltages. You could start with 0.025v less than RMClock auto setting and godown one by one. In RMClock 2.0 you can set one FID you want to test as PowerSaving or Maximal profile and select that profile, so processor would not jump around and you can run your Prime.

And use of intermediate steppings, I use those for Performance On Demand(POD) for Battery profile. Right now, for battery POD setting I have 4x,6x,7x ( My Proc is ML-37). So on battery I never exceed 7x max(at 1.025v). I choose this option, because when I run WMV HD videos, it typicall needs 6x(1200Mhz) for 720P and 7x for 1080P.

For the thermal pad, I think could have safely remove and applied thin layer of AS5. The gap should be taken by heatsink load mechanism without problems( assuming the thickness is less than 1mm). But again, keeping the pad does not hurt that much, may be max 2C difference( although many people claim using AS5 reduced several degrees, with proper installation it can only make around 2-3C difference max, and we know we are no where near top ranges( Max CPU temp for MT,ML is around 95c). So, I would not bother much about it.

I really appreciate you guys for daring to replace CPU and letting us know your success

 

I appreciate the tips concern RMClock. When I get the itch to tweak things a bit more, I'll start using your suggestions.

As for replacing the CPU, it was an audacious move considering the lack of certainty regarding the results and my limited budget, but there's no way that I could get the performance I wanted given the cash I had to spend unless I took a risk. It's paid off in spades.

One more note: I've worked as a computer tech and have plenty of experience experimenting with different thermal solutions for CPUs. As I said, I was planning to go with arctic silver alone, but the thickness of the existing thermal pad took me by suprise. I don't know if that's standard on HP laptops, but it was the thickest I've ever encountered... thus my reaction.

I'd be very interested to hear about other peoples' experiences with replacing CPUs on HP/Compaq laptops, especially on the budget models like mine.

 

I have a ML-37 in my HP L2000 (with 1 gb RAM if that makes a difference). I use RMClock's Performance on Demand. I had it set on 4.0x 1.000 to 10x 1.300 with intermediate steps 6.0, 8.0, and 9.0. It appeared to work fine for the last few days. However, after reading this forum, I attempted to change it to 4.0x at 0.975 and 10x at 1.275 (with all the intermediate steps adjusted accordingly). I had to go to a meeting after about 30 minutes under that setup. I didn't want to leave my computer unattended so I reverted it back to the 1.000-to-1.300 setting. When I came back 2 hours later, I got a blue screen shortly after I started using the computer again and it rebooted.

I then tried to find the error message from my Event Viewer and this appears to be the error: ER_KRNLCRASH_LOG. (This is my best guess as I don't know the exact time the error occured but there are no other system errors near that time frame.) Another issue I had was when I tried to do a System Restore, it would not complete. It kept telling me that System Restore cannot restore my computer to previous points (and I tried 3).

Now the questions for you helpful souls are:
1) was blue screen related to too little voltage?
2) why is it that I cannot do system restore?

Thanks a ton for your help!

 

First of all, system restore is a system resources hog, so I would recommend you to turn it off. It won't do any good, IMO but just take hard disk space and memory resources.
If you restored the voltage to 1v at 4x,then the BSOD should not be from it although BSOD may appear if you apply too little voltage. You'd better do the undervolting excercise - get Prime95 and test at each step (or the lowest and the highest only, that's what I did). I truely doubt that the ML-37 in your L2000 is only capable of 1.00v at 4x. I bet it will do fine at 0.95v and most probably even 0.925v.

 

ML-37 should be able to do 0.95v very easily( mine goes to 0.925v, but I kept some buffer and running at 0.95v) and 10x at 1.225v( running at 1.25v).

So, I do not think it is because of voltage settings, may be it is because combination of RMClock and some other thing.

Which version of RMClock you are using?
if it is 1.8 My first suspect would be, did you select Always On power scheme or not?

Then first thing I would do is reset bios to defaults, restart in safe mode, and disable auto start for RMClock.

Then you may try different version.

 

Thanks for the quick responses!

To answer your question, chinna_n, I did not have "Always On" when my laptop gave me the blue screen. The laptop was idle for about 2 hours. I left it alone when I was in a meeting. I came back to my office, shaked the mouse, started reading more about undervolting, then BSOD. Which completely baffles me.

I ran a quick test earlier with Prime95 and RMClock V.2. I started with x10 at 1.300v. Prime95 didn't like it when I lowered it to 1.250. (It says there are possible errors or some such.) So I guess I then decided to keep it at 1.300. I am testing the lower end now and then a long duration run of x10 at 1.300v just to make sure it's okay.

So does anyone have any idea of how this BSOD occured? I know this is a very open-ended question and it makes me sound like some whiny noob, but I haven't had one in years (probably since I started using Win2K before switching to XP) so this sort of unsettles me. I don't mess with my computer much and have only recently (2 or 3 days ago) started trying to play around with undervolting. I have not done anything else to the computer. Any ideas would be greatly appreciated.

 

As an addendum: I did a quick memory and HD check through HP's online "Advanced Diagnostic Tools." It says both are fine. I know those aren't the best tools available but it was a first step.

 

i keep wondering why i can't set the voltage below 0.8volts on my mt-34, does anyone know something about that? both 4x and 5x are running at 0.8v so i guess i could go a little below that value, but i don't know how to do that...

additionally, is there a way to add a 10x multiplier?

thx in advance,
Ð.

 

You must be making a mistake! I'm sure the Turion cannot work at 5x with 0.8v! Please give us some screenshots but really there's no way this to happen.
You cannot add 10th multiplier on Mt-34 (9x200mhz = 1,800mhz)

 

Well I just tried to undervolt my ML37 (V2000z) and I was VERY disappointed.

at 4x, any voltage below 0.975 caused an immediate crash.

at 10x any voltage under 1.35 caused an immediate crash as well.



at those levels, i'm guessing its not worth the time investment to even figure out it it will be stable over time.

its a good thing I'm pretty happy with the performance as it already is.

bleah.

the strange thing is, when I ran AMD Power Monitor, it said I was running at .900v BEFORE I ran RMClock. RMClock said I was a 1.000v. Either way, on my particular CPU, undervolting doesn't seem to be an option.

 

Hi guys,

I have just noticed this thread.

Would you please summarise for me your conlusions concerning the undervolting capability of ML processors (and MT, if you can). It could serve as an argument in PM vs Turion discussions. I do not take sides, I am just curious to know in order to give better advice to ppl. For example, is it normal to expect 10 or 20% undervoltability as a rule?

 

The same process worked GREAT for my v2555us. With the laptop on my lap, I immediately noticed a difference. My crotch was no longer being subjected to extreme temperatures.

I was able to get the following results:

AMD mobile sempron 3000+
9x to 1.150 in RM Cpu (1.050 in AMD power monitor)
4x to 1.075 in RM Cpu (.975 in AMD power monitor)
(auto adjust intermediate un-checked)

I also noticed a temperature drop from 151 to 142 while running prime95 between the "stock" and new settings.

I now idle around 120F. It seems like that is a bit hot, but maybe I am incorrect. Also, is the sudden jolt from 9x to 4x and back going to be bad for the processor in any way? Should I add a stop in between, like 7x? Glad to have my laptop running at sub-volcanic temperatures, and I hope I see at least a small improvement in battery life.

 

Yup i indeed made a mistake, for 5x it's running at 0.825v (they're all rmclock values), sorry about that... but still, does anyone here know why i can't go below 0.8v and if there is a way (be it through modding or whatnot) to add additional multipliers?!?

 

wait a minute...did I do it wrong???

I thought the voltage RMClock shows IS the voltage...

when I ran AMD Power Monitor, It was showing .900 at 4x to BEGIN with...before I even ran RMClock. When I loaded RMClock, it was showing 1.000v while AMD Power Monitor was showing .900. I thought there was something wrong with AMD PM, so I shut it down. Maybe I'm doing something wrong here.

bananastalktome says he had to set his to 1.075 to get .975 volts in power monitor? I'm confused.

when I played with my voltage settings, I was setting .975 in RMClock and crashing at .950 in RMClock. (at 4X)..

 

The CPU only allows you to go to 0.8v.

I suspect it's RMClock that's wrong (misreading by 0.1V) and AMD Power Monitor that's right, but I'm not sure.

 

This is set by AMD (like 0.700 V for PMs), or it is the limit just for your processor?

 

Can you give us a screenshot of RMClock in the sheet with the voltage values? Sorry, but it's really difficult for me to believe you have 0.825 at 5x (1ghz) and 0.8 for 4x (800mhz). If so you must have at least 3 hours battery time with the 6 cell battery but probably more than that (most of us have around 2h 45min with 0.925 or 0.900 v at 4x as these are the best values I've seen for Turions).

As for the multipliers - the FSB speed is 200mhz and overclocking the CPU will increase this not the number of the multipliers. I don't know of ways to mod this (ask in AMD's forum or probably here in the Modifications forum)

 

It's set by AMD. There might be a way to override it but I don't know how.

 

It appeared that there was a small difference with the set voltage in RMClock and the actual voltage registered with the AMD tool. Not a big deal. Start with a higher voltage and work your way down until one crashes, then move up 1 increment (or 2 to be safe). the list in RMClock only went down to 1.075 for me, so that was the lowest I could set the voltage.

 

ok, well, those are the right values, and both of them are 7+ hours prime stable (i didn't test them any longer, for every step from 4x to 9x "just" about 7hrs was all in all still enough testing time me thinks... ). and about battery time, well, it's a desktop system so i don't know about that, i just posted because i thought i'll find knowledgeable ppl here in these terms...
and to make it worse, no i can't give any screenshots, because i trashed my windows install the other day (the system was meant to run linux anyway, windows was just for easier voltage testing).

yeah i googled aroung a bit, and as it seems there's currently no way of unlocking multipliers for k8 cpu's (fx is completely unlocked though), overclocking is only possible through raising the FSB

cheers,
Ð.

 

HMM...so if I was able to set my RMClock at .975 that might actually be pretty good. AMD powermonitor reads my "normal" 4x voltage as 0.900v on a ML-37, not 1.000v as RMClock says.

 

I was able to set my 4x VID as 0.875V, and ran DVD player for over 2hr 30 mins in playing "Troy" on battery from 100% to 1%. I kept tracking on CPU load on Inter WinDVD, it used over 80% of CPU load shown on AMD dash board. I haven't finished PowerDVD playing movie yet, but it has at least 15% lighter load for sure.

I am not testing so called 'Light use' that only typing and web browsing, but I expect over 3 hours battery life.

 

WinDVD has known issues of keeping the CPU load unreasonably high. PowerDVD should take about 30% and last considerably longer on battery.
Good undervolting, what is your processor?

 

Mine is ML-37. Take a look of my configuration http://forum.notebookreview.com/showthread.php?t=43124

Yes, that was little unbelievable result. I am using BatteryMark 4.0.1 to get numerical test result. Now it is runing on Conditioning mode in 4x @0.875V. running 167 mins and finished @ 5% bettery left. I haven't tested 'Life' mode yet. I expect over 3 hours at least.

I turn off wifi card during the entire test. I heard some report that this V2000Z won't change much in term of battery life if you turn off the WIFI. It may deduct some thing but I don't think it would be that much compare to 105 mins in http://www.laptoplogic.com/resources/detail.php?id=17&page=14

Will post further result after I had it done.

 

How are you able to set it to .875 and so on. MY voltage only goes down to 1.075 on the drop down menu. Is this good (i.e. my processor saying it cant go loweror it will die) or a limitation in the newest version of RMClock. My 4x is currently running great at 1.075 (actually .975 according to AMD meter) and I think it may be able to go lower.

 

None of the above. It is a limitation of the newest revision Mobile Sempron. With 64bit enabled Mobile Sempron clocked as high as 2.2GHz coming AMD need something to separate it from their more expensive lineups. Otherwise the Sempron is just 'too good'. Intel is doing the same thing: They've got LV and ULV processors to sell at a premium, so you can't undervolt below the startup VID on the new Intel Core processors.

 

You've got a Sempron, that is why .

 

This is really a bad news. I hop Turion won't do that. AMD is marked for DIY, restrictions will push away players. Maybe hard wired trick still work like Athlon XP did? That's lot of work. But I only see Sempron have ULV mark, no any sign found in Turion yet.

 

But you can still undervolt at higher steppings. Infact you save more battery that way, than actually undervolting at lower stepping.

 

huh? I have my laptop set up to work at 4x with the lowest voltage when on battery (800mhz is more than enough for taking notes) but you are saying that if I set it up to 6x at the same (or slightly higher/stable) voltage it will both increase power (1.2?ghz) and save battery? How is this so?

 

Yes & No. If you use 6x at same VID, you will surely save lot more power compare to 'what is originally marked in 6x VID', but not 6x will cost more than 4x for sure. When frequency changed, the impedance will change and that's why the power will not be linear to VID when you change the FID. It only linear when you fix FID and change your VID.

 

No, I do not mean when you are running idle, but when you have load. For example if your processing is jumping around a lot while using, then it helps.

Lowest mulitplier with lowest voltage is better for least heat and least power consumption.

 

Hi, just a silly question.

I use RMClock 2.05 and the system power control will show it is managed by RMClock. But after a long time of not doing anything, the system will automatically when to standby state, and never come back. I have to push power button to shut it off, and power up to bring it back.

Do any of you know what's wrong?

 

Check your Power Options. You've probably got System standby set to trigger at some point. Also, make sure your Power Scheme is set to Always On since RMClock is managing that now.

 

That is the problem. No matter how hard I have tried, whenever I changed working mode in RMClock, the power scheme will always change from Always On to RMClock Manage. I even delete RMClock Manage option in power scheme from control panel. But it always comes back after I change scheme.

In this way, I will never change any working mode but stay in Performance on Demand. Or I have to change Power Scheme every time after I change in RMClock, that is really handy.

 

Oh, I didn't know RMClock works that way. (I still use CrystalCPUID.) RMClock Manage is probably fine?