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

It probably new RMClock version write something in register table to get hold of it. But I cannot find a way to remove it. But manually Hibernate function works fine, it can in and out.

Anyone have similar problem?

 

so how much time extra battery time do you get? how much battery life u get from your computer?

 

My 6 cell 4400 mAh full charged battery will last 3 hrs 11 mins for light usage(BatteryMark 2004 test result).

 

you're lucky you got the 44000mah battery... a lot of us have the crippled 40000 version

 

That is right. He is lucky and we are loosing exactly that 20Min battery life.( which makes us to cross 3 hrs).

 

I think HP fixed this problem recently. I expect new order should have 4400 mAh. I had checked some online battery reseller, they sell 4400 mAh only, no 4000 mAh for V2000Z. The 4000 one should be a temporary replacement at begining of the product line. For a company, managing 2 capacity for same chase doesn't make any sense. So I would think that is a mistake, and may had been fixed already.

Well, I maybe wrong, or something......

 

I have a Gateway MX6425.

I have the following settings with RM CPU Utility

On Battery:
State 0 FID: 4.0X VID: 0.875V
State 1 FID: 8.0X VID: 1.150V
State 2 FID: 9.0X VID 1.250V

On AC:
"Maximal Perfomance" profile with no settings changed, so default (since it doesn't matter to me what the processor does when it is plugged in)

I have yet to test battery life, but it looks really promising. I could get 3 hours before, and that was without speedstep installed (wonder why Gateway didn't install it with the PC )

Anyone have any suggestions/comments? I think the 0.875V is good, but I got tired of testing all the VIDs so are the State 1 and 2s alright? I never optimzed them...

 

I have a L2105 w/ ML-30 1gb ram, brightview 14", 6 cell.

I undervolt to 1.2v w/ 1600mhz stock.
Drops temps nicely and great battery life.

I do 1.1v at 4x (800mhz)
I like to be gentle so I can really tell where I lose stability.
I haven't dropped it to 4x too much and done cpu intensive apps (I try and be gentle on the battery when i am on it)

No life numbers though.

 

you can do much better at 4x than 1.1v. Set it to 0.95 and you can be sure it will be stable. The stock voltage is 1.0v...

 

Can anyone reply to my post above please? I am curious....

 

so how much battery life you got ?

 

I used to only get 2 1/3, now I get 4! This is with the screen fully dimmed, wireless off and just typing notes or surfing the 'net at school on ethernet.

 

0.875 is very good. Few people managed to get that low with stable performance. What kind of battery do you have? 6-cell? How many amper hours (it's written on the battery, probably 4.4ahr)?
I get a bit better value on my L2000 for 9x, 1.2v, so maybe you should be able to do better at the maximal step (yet I only get 0.925 for 4x),which will reduce heat while playing games or doing CPU intensive work.
Furthermore, using Maximal Performance while on AC power will increase the temperature of the laptop, which is not recommended at all. The CPU gears up and down so fast that you shouldn't feel any difference if you set it to Power on Demand (RMclock 2.05).

 

Oh, ok, for the 9X I got a VID of 1.1V. I also set it to do that on ac power also.

Thanks!!

 

Ahh so the stock battery saving setting is 800mhz / 1.0v.
Well I run it at 1.0v and 800mhz in class and yeah it worked fine.
1.2v on the 1.6ghz 1mb cache
I like the increased battery life.
It competes with Intel better.
I will try .925v on 800mhz.

 

Hi,

I did some undervolting and here are my results:

Profile: Performance on Demand
-AC Power & Battery
4.0x 0.925 Temp(48*C)
6.0x 1.025
9.0x 1.150 Temp(53*C)

Other tests:
9.0x 1.500 (default) Temp(67*C)
9.0x 1.075 OK Temp(51*C)
9.0x 1.050 FAILED
4.0x 1.000 (default) Temp(48*C)
4.0x 0.875 OK Temp(47*C)

From my tests, I settled on the settings above because I want the extra buffer and also, I'm mostly near an outlet so battery power is not a major concern. I'm already happy that the load temps dropped from 67*C to 53*C.

-Thanks

 

For absolute stability with my ML-32, I need to use:

0.95v @ 800MHz
1.225v @ 1.8GHz

 

If you are stable at 0.875, then you should set it to 0.900 at most. It helps keep the laptop cool while browsing.

 

At 0.875, it was stable for maybe 15minutes but I haven't tested longer than that. Also, I didn't see much a temperature difference between 0.875 and 0.925. Maybe after a month of using the settings I decided on, I'll try lowering it and run it for another month and see if it's stable or not.

 

hm, for testing use Prime95 not just some period of time. You may not experience a problem unless the CPU is pushed to the max for some time.

 

Agreed, use the CPU torture test. In the testing I've done, it'll show an error quickly, and will show instability that windows doesn't see -- in a couple tests I saw Prime95 report an unexpected/invalid computational result which didn't occur at a higher voltage.

I’m no expert, but I figure if I can run Prime95 for eight hours with no errors, it’s as stable as I’ll ever need it to be. So far I’ve verified 4.0x @ 0.875volts and 9.0x at 1.100volts. From the other reports, I figure this is pretty decent.

I’m still working on verifying the intermediate steps. ML-32 cpu (V2401CL).

-- Lew

 

I did use Prime95 but I figure that was only testing the CPU at full speed(9.0x). Can Prime95 test the CPU at low speed(4.0x)?

 

Can Prime95 test the CPU at low speed(4.0x)?

Sure, and at any speed in between (5x, 6x, 7x, etc.)

Set up only a single speed in an RMclock profile and use that. Prime95 will max the cpu at the clock rate you locked the laptop to with RMclock. If you have multiple steps, then it’ll ramp up of course, but if your profile has only 4.0x then the cpu won’t go beyond that clock rate.

That’s how you can thoroughly test each clock rate step. And once you’ve figured out what your system is stable at for each of the steps (4x, 5x, 6x, 7x, 8x, 9x, etc.) you can build a Performance-On-Demand profile with the minimum stable voltage for each step, thereby maximizing your battery efficiency.

-- Lew

 

Oh, by the way, I've been out of the loop on laptops for a long time. What are people using to benchmark their battery life and/or to run conditioning cycles?

Thanks.

-- Lew

 

I use a stopwatch

 

One issue with under-volt is that instability creeps during transition state. The notebook might be stable at particular multiplier and voltage settings but it becomes unstable when you mix them in "Performance on Demand" mode.

I am testing the following setting which was designed to minimize voltage jumps. I havn't had a crash yet in 12 hours but I need more testing to be sure.

4x @0.9V
5x @1.0V
8x @1.1V
9x @1.175V

 

Abugi, RMClock 2.05 was made to prevent this instability you are talking about. I haven't had problems in transitions but some people really have reported unexpected lockups. \
Btw, I haven't tested my system for so long. To tell you the truth I have only tested in the lowest and the highest step for a longer period of time.

 

I actually use the laptop and report actual usage timings. So it is most realistic time in typical usage( wireless on 100%, most browsing, brightness at 2nd step from bottom). My battery is 4000mAH( not 4400mAH) and I am still getting 2 hrs 50min after 6 months. How do I read timings?, the event viewer in windows gives you when exactly laptop is started and when it went to stopped/hibernate. So, no wild guesses( or oops, I forgot to note down time).

 

Cool. Now to find the time to actually use the laptop for more than a few minutes at a time. Pesky life interfering with the computer & websurfing time!

 

I thought I'd post my numbers for my ML-37. I started tweaking it last night.

10x @ 1.225V gave me errors in Prime95 after 5 minutes
10x @ 1.250V ran prime95 for 8 hours with no errors

4x @ 0.825V crashed immediately (screen corrupted)
4x @ 0.875V ran prime95 for 10+ minutes with no errors
I am running prime95 4x @ 0.850V right now at home... we'll see how it fares when I get home this evening.

 

I just got my Compaq V2000z yesterday (ML-37). I'm backing it up and then will start testing and post my numbers.

 

Hi, all.
Been watching this thread with great interest. Looking forward to your results, hitobahadur. Thanks for your contributions chinna_n, vassil_98, pompoko, all of you!

I'm about to enter my order for my very own L2000, and thanks to you I'll know just how to prepare it. I just can't decide on one thing: whether to order the 2.0 or the 1.8. The ML-34 or the ML-37. Sure, the ML-37 gives you more speed (I have the 50 bucks), but the undervolting results aren't as good...are they? You can't set the numbers as low, but is the end result what I want? I'm not so concerned about battery life. I just want the thing to run not-so-hot and to perform well.

So here's the question: does the boost in performance from the extra .2 GHz justify the possibility of maybe running hotter and maybe having to replace fans in the future? Or does that danger even exist?

Thanks much for your opinions/judgements/discusions/whatever you can offer on this.

-T

 

Yes. I faster processor is almost always better. The boost is definitely worth it. Heat difference will be so small it will be virtually unnoticable.

 

I think if $50 does not matter much, then ML-37 is way to go. The reason, ML-37 can always run at 1800Mhz at lower voltage( or atleast same) than ML-34 at 1800Mhz. And you also get benifit because the rated wattage at 1800Mhz for ML-37 is only 29watts vs 34watts for ML-34 even without undervolting. You start at lower base line of power at the same speed for ML-37 and also you have power when you really need.

 

ML-37 is the better choice. I wanted to get it but unfortunately I didn't have time to wait for it to be build and shipped. ML-34 is more of a budget CPU now (that's what I have)
Unfortunately, every chip undervolts differently. I've seen ML-37 turions that do fine at 0.875 and ones that won't go below 0.95, so it really the draw of the luck. MT series on the other hand are standardized, which is their greatest advantage.

 

Thanks for the fast (and clear!) responses. You've given me what I need, and I'm going to go for the ML-37. So glad I asked.
I will order it this week, planning to post my results here in a couple of weeks. Hope I get one of the more pliable ones.

Thanks again,
T

 

Ok, all, here's what I got:

- HP Special Edition L2000 CTO NB
- Microsoft(R) Windows(R) XP Professional with SP2
- AMD Turion(TM) 64 ML-37 (2.0GHz/1MB L2 Cache)
- 14.0 WXGA BrightView Widescreen (1280x768)
- ATI RADEON(R) XPRESS 200M w/productivity ports
- 512MB DDR SDRAM (1x512MB)
- (I added 1GB Patriot PC2700 DDR)
- 80 GB 5400 RPM Hard Drive
- DVD+/-RW/R CD-RW Combo w/Double Layer
- 54g(TM) Integ. Broadcom 802.11b/g WLAN & Bluetooth
- 12 Cell Lithium Ion Battery

I was pleased to find that the AMD processor driver was already at the current version found on AMD's site (1.2.2), not the old 1.1 that users had trouble with (see postings after the L2000 review last year), causing the fan to run a lot, etc. So, the heat issues weren't really super bad to begin with. My 2-year-old hp zv5034us 2.66 GHz P4 was worse. Using AMD's Dashboard Demo to monitor and RMClock to manage, I set these parameters:

4x - 0.950v
6x - 1.000v
8x - 1.175v
10x - 1.275v

I started with 4x at 9.25 and 10x at 1.200, but got a blue screen and/or lots of errors with Prime95. Been running stable at the above numbers, including encoding an XviD movie (Monty Python, episode 24 if you want to know) and just a few other taxing operations. Under normal browsing, etc, conditions it generally stays comfortably at 4x (about 798 MHz). As for the 6x (1.20 GHz) and 8x (1.60 GHz) settings, it seems to use the 8x more often. I could almost surely fine tune the intermediate settings more, but I'll have to figure this out as I go since this is my first adventure in undervolting.

There have been no reasons to complain about heat. It does get hotter underneath, but I only notice if I pick it up - it rests on an oversize Edward Hopper book if it's on my lap. There's none of the wrist-numbing heat under the left palm I've read about. The fan rarely comes on, and then just breifly. I'm not cheifly concerned about battery life, but I've been quite pleased with the longevity when I have gone without AC - and did notice a deceleration in battery consumption after configuring RMClock.

Overall, I like the machine. It's quick. It's wide. The LIVESTRONG letters aren't as obnoxious as they seem in the pictures. I didn't think I'd mind, but the hd really could be faster. The keyboard feels so nice. At least the wireless button is still blue. Ok, enough with the mini-review.

-T

 

Yes, this is the main problem with all laptops. If you are used to a fast desktop HDD, a laptop would offer no match unfortunately. Even the 7200 HDDs in laptops are slower than you would expect.
L2000 is a nice machine

 

Those low stepping values look very similiar to mine, so I am guessing,
you could run 10x at 1.25V stable and 8x at 1.075.

Here are my current steppings

4x 0.95v
6x 1.00v
8x 1.075v
9x 1.175v
10x 1.25v

Anyways, it is good to know you like it. BTW you could turn off many unnecessary services in windows XP to reduce unnecessary CPU usage to reduce heat.

 

Now, this time I know I am kind of crazy.

I am trying to test undervolt performance on my new Athlon 64 3500+ desktop now. I am able to do it like this

5x 1.175v
6x 1.200v
7x 1.250v
8x 1.300v
10x 1.350v
11x 1.400v

I am not so aggresive this time, just initial setup VID like this without Prime95 test yet.

But do anyone know where to find those TPD figures for Athlon 64 instead of Turion? What would the TPD under 1G? 1.2G? 1.4G? 1.6G? I just know at 2.2GHz, it will cost 67w. But I don't know how much I can save by stepping down the FID.

Why I do this? Mainly because the heat, and save my electric bill.

Just being greedy.

 

Undervolting started on desktop Athlon 64's. That's what the folks at silentpcreview.com were doing (great site btw, you should check it out). I found it worked just as well on AMD notebooks and popularized it here.

I'd guess that you can do full speed at 1.2V core (assuming a decent cooling system), which would drop your TDP to around Turion ML levels. Your CPU won't go below 1.1V core but lowering the clock speed lowers power consumption too, just not as dramatically as lowering the core voltage. Do the usual stability tests to make sure of course.

 

That's what I am talking about. Thanks.

I get 5x in 1.125v now, and test well so far. It might not save much because my fan is not a smart fan like mobo, it will not stop.

 

OK, not much exciting. 5x @ 1.10v is the lowest I can go.

I check the TPD from AMD datasheet. 5x is 27.2W for 3500+.

Anyway, done with my 3500+, just make it cooler.

 

I managed to undervolt my Athlon63 3500+ in ZV6000 to 0.875 at 5x!

 

The Athlons underclock pretty nice as well. I have been running for most part of the whole year and half with these settings...
(DTR Athlon64 3200+)
4x @ 0.850V
6x @ 1.175V
10x @ 1.300V

Pretty good undervolt considering the stock settings...
4x @ 1.1V
6x @ 1.3V
10x @ 1.5V.

 

Do you mean Athlon 64 mobile version? RMClock won't give me any VID under 1.1V. How can you make it? Or My motherboard lock the lowest VID. I use MSI 482M4.

 

Hi guys. Genuinely sorry about asking the following newb question
what softwares do you use to undervolt cpu?

 

Thanks for the reminders, chinna_n. I have turned off a host of unused services and added a 9x setting. Always more tweaking to do, eh?

Speaking of which... Now that I have some things locked down here, the next project for me is to wipe out the default install and put Windows XP x64 on it. Found all the important drivers except Broadcom wireless (802.11b/g and bluetooth). I may need to move this to another thread, but... Has anyone else had any experiences, good or bad, with putting 64-bit Windows on the L2000?

-T

 

RMClock or CrystalCPUID.

 

Hm, I think it's a regular desktop Athlon64. I don't know how many watts this is but 0.875v for 5x (1GHz) is pretty good ( the processor does not support 4x step)