Undervolting the Z11VN (P9500)

Just to share my results...

I have got the 9x multiplier down to 1.0250V now and this seems to have taken my max core temps from around 65°C to around 55°C. I've no idea yet of what this means for battery life, but I can confirm that the fan is far less noisy on full load with RMClock running with this setting...

As a (completely subjective) example, I set up some lengthy divx encoding and found the fan noise to be far less intrusive with the lower voltage applied.

I haven't bothered testing the intermediate multipliers yet, as my Z tends to spend most of its time either on the lowest or highest multipliers anyway.

If you decide to give it a go, please share your results with us.

Here's the guide which is pretty straightforward to follow:

http://forum.notebookreview.com/showthread.php?t=235824

 

It would be awesome if you could test the battery life, that is actually the only reason for me to go throw this on my vaio. If there is some battery life gain, I will do it right away.

 

I am returning my Z today, I would have undervolted but too much hassle for now =P lol

 

Please do report on effects on battery life.

 

I notice you say you have the P9500. According to the guide, because RMClock only supports integer multipliers (9x as opposed to 9.5x), it results in both undervolt and underclocking. This is why I haven't undervolted yet.

A few questions for you. On my machine, I can't start HWMonitor without it instantly crashing, does this happen to you? Also, 65 degrees for a max temp seems too cool, doesn't it? I remember when I fired mine up it went into the upper 80s....

 

It's true that RMClock doesn't support half multipliers, so you are effectively underclocking from max. 9.5x to max. 9x. But that was a price I was willing to pay for battery life... (it's not as if the changes are permanent)
In my case, I generally don't max out the CPU anyway and when I do, the difference between 9x and 9.5x isn't the end of the earth. So it's a good compromise for enhanced battery life IMHO.

I haven't got a good measurement of battery life before vs. after and not sure how useful that would be because mostly I'm just browsing and doing low intensity tasks and you can't undervolt the lowest 6x multiplier. But I'm easily getting 6 hours per charge, which I'm very happy with. Settings to achieve that are about 30-50% brightness, wifi on, disable the modem, iLink and other stuff I'm not using, use quiet mode, etc.

Although I installed HWMonitor when I tried this in Vista, I haven't installed it in XP. Hence the temperatures I quoted were given to me by RMClock - they're probably wrong but I only wanted a quick relative check to see that the changes were making a difference.
I might installed HWMonitor again later and check my temps with and without the undervolt applied.

 

why are you returning it? because of the premium carbon fiber?

 

btw, does undervolting comprimise performance?

 

Yep, I dont want my 20 day no Q-asked return window to expire.. I LOVE the Z but I am CRAVING a Premium Carbon fiber richer blacker look - SO bad - lol

I am in no hurry for it which is why I am going through the hassle of returning this one.. I figure since I am shelling out 3k, I might as well get the premium carbon fiber and not regret it when people flash it around on here! =P

The 20 day return policy on Sonystyle orders is the only good thing about buying from them - if it passes, its next to impossible to teturn/replace the darn thing w/o breaking it lol (and avoiding 15% r/s fee = SUX for the high price of Z)

 

You can also try this software:

http://www.pbus-167.com/

 

oh.. so you still have another computer you can work with?

i wish i could wait, but my Z is going to be my only computer for now, and as we dont have any expected release date in the US.. plus, it will probably cost over 3k which i cant afford

 

The short answer is no, it doesn't.

The longer answer is that it might slightly compromise performance if you have a CPU which has a fractional max. multiplier, such as my Intel P9500 which has a 9.5x max. multiplier.
The RMclock software only recognises whole numbers for the multipliers - so 9.5x becomes 9x. This is not a huge difference and in my opinion, it's worth it to save power/battery life.

If your CPU has a max multiplier of 9x (or any other whole number) anyway, then undervolting it won't affect performance. Essentially you're just fine tuning the power to the chip to make it more efficient. All CPUs are different though, which is why some people can only shave off a small amount of voltage before they get stability problems and other people have much more luck. (This is also the reason that Intel plays safe and provides many chips with more power than they strictly need, using a voltage that works for all.)

 

oh.. i see.. yeah.. did you notice any gain in battery life?
i ordered my Z with the same processor and was wondering if i should do it or not..
in case we get a BSOD? how would we solve it?
via BIOS?

 

I haven't bothered measuring the increase in battery life, but it stands to reason that if the chip is using less power, getting less hot and the fan is running less/quieter... there will be a consequential improvement in battery life. To be honest I was happy with the battery life before, but wanted to see if this worked as I've never tried it before on my previous laptops.

If you don't set RMClock (or NHC or whatever you're using to undervolt) to run and startup and/or you don't set it to apply your lower voltage settings on startup, then a BSOD won't cause a problem - just reboot and your normal voltage will apply and the system will be as stable as before you started. Then just raise the lowest voltage you tried slightly and do another stability test. Eventually you'll find the lowest safe value you can use.

 

My results are below. Temps measured with RealTemp (HWMonitor crashed on load). All voltages are bumped up 2 steps from crashing point. I'm sure some of you will want me to test battery life but I'll just say up front I'm too lazy to do that.

All temps recorded at 100% FULL LOAD.
Ambient: 28°
9.0x: 1.1000 : 80°/85° --> 0.9875 : 68°/71°
8.0x: 0.???? : ??°/??° --> 0.9250 : 64°/66°
7.0x: 0.???? : ??°/??° --> 0.8875 : 58°/60°
6.0x: Did not measure, can't drop voltage further.

 

is that in celsius?
that seems to be awfully high initially

my Z would run about 50 degrees when browsing the internet with itunes and messenger open

 

Sorry, I edited to mention all temperatures were recorded under 100% full load. I didn't really see the point of recording idle temps. You can't drop any further than the super LFM mode anyway, so unless I'm missing something, how did you get a temp improvement there? I idle at 41°/42°.

I'd like to know why InfyMcGirk's temps are so low. What is your ambient? My temps are like 20° higher than yours and that seems crazy!

Edit: I just found that RealTemp was using 95° as TJMax and Intel's site shows 105° for the P9500! That would make all my temps 10° higher than what I reported. That seems really high to me. I'm confused. =/ http://processorfinder.intel.com/details.aspx?sSpec=SLB4E

Edit2: ***? Did Sony ninja change the processors available in the Z? I swear before it was T8100, T8300 & P9500. Now it's P8400, P8600 & P9500. I would have just gone with the P8400 had I known. The extra cache on the P9500 is nice, but the 25w vs 35w on the T's was another big selling point for me. **** you Sony!

 

?
the P9500 is 25w

 

Right, that's what I'm saying, the T's are 35w. Weren't the T's what was initially offered on the Z?

 

My ambient temp would have been only around 20°C, for one thing. Also, my quoted temps are just what RMClock told me, which is probably not accurate. I wasn't looking for accuracy but rather a quick indication that the changes I was making were making a difference in the right direction...

I think it's always been the P range in most of the world. The higher power 2.8GHz T chip was only available from Japan (or via import from Japan elsewhere).

Only in Japan...

 

Very nice results. 14° lower. That must make some difference with fan noise.

 

I have P8600 in my FW? Would the same undervolting threshold apply to that processor as well as the P9500 - 9 multiplier at 1.0250v?? Thanks.

Gary

 

only in Japan, and it was the 2.8GHz processor.

 

Maybe, maybe not. Even if you had a P9500, you might need to use different values. I'm afraid you need to follow the process in the guide to find the right values for your processor. Every chip is different...

 

No !! it wont Compromise Performance.Undervolt and Overclocking is Different

 

Except for the slight underclock you get if you have e.g. a P9500 and RMClock can only set it at 9x rather than 9.5x multiplier, as mentioned earlier.

 

just wondering did you set to 1.0250V all of the indexes?

 

yeah, i tried running HWmonitor and it crashed on me.. dont know why.. so i ran the test using coretemp and i didnt get any improvement

before undervolting: max -> 82
after undervolting: max -> 82

so is coretemp not reliable?

 

My first guess would be that you missed a step in the tutorial - maybe didn't activate the performance on demand profile? The reason I say this is that I saw a pretty definite drop in temps even after the first change was made.

Did you use Orthos to max out the CPU as per the guide?

It's worth having another go and making sure you follow all the steps in flipfire's guide. I can't believe you wouldn't see any drop in temps, even if it's only a slight difference.

Just as a rough test, try monitoring the temps reported in RMclock itself. They may not be accurate but should demonstrate a relative improvement if you're doing it right.

 

Nope, I couldn't be bothered testing the intermediate multipliers rigorously so I checked the box for RMclock to automatically fill them in for me.

 

i did set the main profile as profile on demnad
but im thinking that maybe the results werent recorded properly due to the coretemp.. or do you think it is a reliable tool?

 

do i have to open rmclock everytime i boot the system for the changes to take place?

 

yup theres an option to start it when windows starts

 

thx danibui, anyone willing to test for the intermediate multipliers?
or we could split the work.. each on take one of the multipliers and test and post here the results?

it would be nice to get over 5 hours of battery.. i am getting just over 4 right now with 6/8 brightness

 

i tried the lowest voltage on all the multipliers which was 1.035 and system locked up and restarted =(

 

on all?
thats weird, i managed to go all the way down to 1.0250v on the last two multipliers as infy posted..

but i didnt really bother trying to lower the 8/7/6 multipliers

 

Tick the 'Auto-Adjust Intermediate' box and it will adjust the 8 & 7 multiplier accordingly by calculating from your lowest and highest voltages.

Lower the multiplier #, the less voltage it needs.

 

Unfortunately each processor (even of the same model) is different, so although you could get a rough idea of what's possible from somebody else's results, you couldn't guarantee that the lowest stable voltage for one person would be stable for another.

It's a shame that we can't lower the voltage for 6x, because that's the multiplier my CPU spends most time in when I'm just browsing or whatever.

 

I assume we're using the same processor P8600 2.4GHz
You need to set CPU type sellection as mobile and u'll be able to toggle even lower settings. The lowest is 0.9250v for me by default is 6.0x and Super LFM 6.0x, which is weird? isn't super Lfm supposed be lower?

Anyway i did got stability at FID 8.0x at 0.9250v =/ so now my cpu only has 2steps since it's pointless to use the lower FIDs at 0.9250v

I tried going lower by enabling all the VIDs with a registry tweak buy the cpu wouldn't go anylower than 0.9250v

 

My understanding is that the bus speed itself is lowered in Super Low Frequency Mode, not the multiplier. So 6x means "6 times the standard bus speed" and SuperLFM 6x means "6 times a reduced bus speed". At least, that's what I think it means.

 

SuperLFM is a feature that dynamically cuts the front side bus (FSB) frequency in half on idle

eg. My normal 6x idle is 1.2ghz but with SLFM 6x = 600mhz idle

 

Thanks for the responses,

Yep that's what i understand too, i think it might save abit more power by cutting to half the fsb when u're really idling but i believe a lower 6.0x voltage would pair up better with super LFM..Rather than using the same 6.0x voltage as Full FSb. I've seen some screen shots of RMclock featuring other cpus having lower Super LFM voltages to full FSB voltages.

Sometimes it's not efficient to run on lower freq since it'll load up the cpu higher which results in more power drain. i see that on LFM it sometimes loads up to 20% which same task being only 1-2% on 8.0x..

 

the lowest i can get x9 multiplier is 0.9875
here are my stable settings

Type FID VID
Super LFM 6.0x 0.9250v
Normal 6.0x 0.9250v
Normal 7.0x 0.9500v
Normal 8.0x 0.9625v
Normal 9.0x 0.9875v

Lowest temp on idle I gotten is 38c it averages around 45c normal use so its around 10 degree cooler difference. Fans to seem quiter although still blows non-stop its louder than my TZ i used to own. Core temp shows diffrent temperture than rmclock I wonder which one is more accurate. Everyone should post their most stable settings.

 

weird, my voltage is lower for mine LFM 0.8750v

 

That's very interesting as I don't have an option to go lower than 0.9250V. Would you mind sharing screenshots of your settings, as I'd like to try going lower if possible?

What CPU do you have?

 

Infy, have you tried Crystal Mark? It seems like the voltages go much lower.

I am not sure it's working though. RM Clock says the voltages aren't actually lower.

 

You can technically overcome this by setting a target CPU load so when it hits that it will bump up to the next multiplier. I just wonder if you set the target load too aggressively low if causing your multiplier to shift frequently increases power drain.

 

P8600

Type FID VID
Super LFM 6.0x 0.9250v
Normal 6.0x 0.9250v - 44DEGS
Normal 7.0x 0.9250v - 46
Normal 8.0x 0.9250v - 48
Normal 9.0x 0.9750v -5X

All tested full load for 45mins<

What i meant was that since my 8.0x multiplier could run stable on the lowest voltage, why not simply set 8.0x as my lowest multiplier since the power drain should be the similar to setting it at 6.0x...

Running on lowest multiplier would load up the cpu by maybe 10% more doing the same task, while they(6.0 and 8.0) could both are set at the at the lowest 0.9250v.. then it makes sense to set it at 8.0 all the way?

Actually No.. althought they're on the same voltage, it seems like 8.0x still drains alittle more than the 6.0x. It ran hotter by 2Degs on full load. so i guess voltages aren't just the factor. To test my observatoion I set super LFM FID at the same voltage as my 9.0 Fid and it ran like 10Degs cooler =)

 

Dont forget voltage isnt everything. Amps is what draws the power. You get lower draw on a lower clockspeed.

voltage x amp = watts

 

i tried to undervolt..
strange thing is that the undervolt will not be recognised until i manually go to the vista current power plan and change the scheme from Balanced to RMClock Power Management

this is irritating since whenever vista boot up, it will by default select Balanced and not RMClock, and the settings saved for RMClock will be deleted when i reboot computer...

Is there anyway to solve it?