How to Supercharge the M11x with Core i5/i7 UM CPUs

Nah, your just being an jerk. Ya i can follow step by step instructions anyone can. I did not want to know how to install it, what i wanted to know is what does this technically do to my computer, coz i am not doing something to my computer i do not know. Jeez man, is this all you do all day, complain about noobs asking questions?

 

Heh, I DO have quite a few processes running...have a look.and this is pruned down to get good battery life.
Running Processes in Task Manager
Can anyone help me out as to what might be the culprit here??

 

When you're running wPrime are you seeing a cpu-z process running?

 

Hello Unclewebb

Please find the logs for both versions for your reference in the link below. I have included a readme file that explain about the log files to which is which so please refer to that to know what you are looking at. My system specs you can find on my sig. Thanks for your hardwork! Rep to you!

TS Logs.zip

 

1024M: 1224 sec -> 980 sec (ThrottleStop 2.88) -> 855 (ThrottleStop 2.88.2)
Wow, just wow

Edit: I like my new processor rating on WEI

 

I was under the impression that the CPU itself handled Turbo Boost, under a closed loop checking of the thermal conditions. Didn't know that the thermal limits were editable, and the BIOS had a role.

Anyways, have you figured out when, or based on what criterias, the BIOS lowers the limits back to default? Is it periodic? I assume that when you detect that the limits are back to default, you simply reset them back to the ThrottleStop values? So if you optimize this, you will minimize the time the CPU is reverted back to lower multipliers (but can't fully avoid it)?

 

Is everyone only finding improvement results when they OC? My scores seem to not change even with TB on, unless I only OC.

Or does TB only show significant improvement with OCing only?

 

Yes, you need to OC to see improvement.

 

38 fps average -> 47 fps average in SC2 during the same replay with 2.88.2 TS. I think this is what every R2 was waiting for, the performance increase associated with the $$ spent, battery life decrease, and loss of VGA.

 

Last night I found a work around for a major Windows limitation. Version 2.88.2 is a significant improvement over the original and you can find a link to this new version in the first post. 2.88.2 should give much more consistent performance, even when fully loaded. At the moment, with v.2.88.2, the only way to disable it is to exit ThrottleStop so keep that in mind when testing.

I'm working with Cosmin. to get every last ounce of performance out of the M11x. In theory, if you are being limited by TDP/TDC, you might be able to improve your wPrime times by another 1 second compared to 2.88.2. Not much left to be gained but we're so close now that I figure we might as well go for it.

instantcold: The turbo TDP/TDC limits that ThrottleStop lets you adjust, controls when your CPU gives you turbo boost. The problem with the M11x is that the Core i processors this model is using are all UM, low power, processors. These have the turbo TDP value set far lower than a regular Core i5 or Core i7 Dual Core does. The UM sets turbo TDP to 10 watts while I think the regular Core i5 are set to about 30 watts. When your CPU starts consuming more than 10 watts, turbo boost will immediately get turned off so your CPU is running at the default multiplier. On a Core i5-520UM that is fully loaded, the multiplier will start rapidly cycling between the default which is 8 and the turbo multiplier which is 12. If you go into the bios and increase the BCLK to try and go faster, you are still limited to 10 watts so at full load, this turbo throttling will happen more often and your average multiplier will decrease. Not only will full load performance not improve any, it actually seems to decrease when you start overclocking the BCLK because of this wall you are hitting.

ThrottleStop lets you raise this turbo TDP power limitation so at full load you will end up getting a lot more turbo boost. You will finally be able to benefit from raising the BCLK. Dell markets the M11x as overclockable so this little tweak lets you fully take advantage of that feature.

kopicha: Thanks so much for the logs. 2.88.2 is looking very good.

 

Instead of warning that this may void the warranty or other stuff, might as well admit that it was a flaw in that design you cherish there.

Also, the battery life can be preserved, since ThrottleStop can be turned off and/or used with a profile that uses the default TDP/TDC.

Also, this thread should be sticky (even though the users are sticky-ing it), for it gives back the m11x the value that was marketed by Alienware, but taken away by poor implementation.

By the way, as we`re discussing overclocking, it is simply nasty that Alienware masked an inability of the i5 520UM to reach 166x12 frequencies constantly (at least with the default voltage that is) when the multiplier doesn`t ping-pong down to 8x. I`m happy with the 160bclk, still, if there can be done more, why not?

Although the ability to set the BCLK is more than enough, I reckon a job well done would have included the following settings in the BIOS: CPU voltage (because undervolting helps battery life), CPU multiplier, BCLK/DDR ratios, QPI voltage, DDR timings, etc.

Let me give an example: you have a CPU that defaults 133x8 MHz at 1V. I can bet one year's supply of cigs that it can do 133x8 at 0.85V, resulting in lower TDP. When you go 160x8 at 1V, TDP is X. 160x8 at 0.85V is Y. Clearly Y <<< X, so instead of putting there a limit on the TDP/TDC, simply implement a damn CPU Voltage setting in the BIOS. A lot easier imho.

But still, that is wishful thinking

 

1 second on wprime = about 100 point increase in 3dmark06 cpu score and about 2 fps average increase in SC2 from what I have seen between 2.88.1 and 2.88.2 which showed about a 1 second difference.

 

I also got 1 second decrease in WPrime between versions.

 

Aw, it's too bad then that TB only shows improvement based on OCing.

I don't want to have to always go through my boot options to turn OCing on or off since I use my laptop a lot for school, meaning I want to optimize my battery life.

Or does anyone know if OCing would use more battery life versus not OCing?
If it's a minor difference then I wouldn't mind, but I read some users just leave their OCing on. But I'm guessing theoretically it would use more battery life since it OCing uses more power.

 

While some of you will want to explore what Webb has outlined, please note that doing so may result in decreased battery time. Also, and this is just a guess after dealing with tech support over the years, support will not work with you in the event of any errors, BSODs, or general troubleshooting. The reason for this is simple - using TS modifies the performance of the system (BIOS) from the way it was originally designed.

This thread is a 'modification' thread. Be sure you note the warning that Webb adds to his app, 'use at your own risk'.

Just to clarify -

My 'warning' is directed for owners who may be new to these items. Don't be fooled, there are many who have no idea what TDP/TDC, BCLK, etc. mean and will not take the time to read/research before diving in. For those of you who are aware, feel free as you already understand the potential results (and consequences).

Any other comments concerning my message need to be PM'd.

Thanks.

 

You can use TS to also conserve battery life as well. You can have TS just disable when running on battery. See Cosmin's explanation below. Keep it OC'd in BIOS and just make use of lower multipliers when in power saving.

 

That's not my concern though. My concern is since TS only works when you OC through the BIOS, that means the computer will always be in OC until I turn it off again through the BIOS. But I don't want to leave it on OC since I use my laptop a lot for it's battery life.

So asking my question again, would leaving my computer on OC have a major impact on battery life?

 

Well, you can certainly use the power management features of Windows 7 and let the bclk clock overclocked. The multiplier will drop and the CPU will run at a lower frequency, saving battery life.

You can also create profiles within ThrottleStop with the default TDP/TDC that will enable you to control even more efficiently the battery life from the processor's point of view.

Also, imho the most power hungry monster is the screen, not the processor Turning down the brightness all the way gives me around 1 hour extra battery time, making it perfect for reading in bed.

 

That's good to hear. Would you think that leaving OCing on 166 throughout my M11x's lifetime will be really damaging? I searched through the forums and a lot of people seem to leave on their OC, but no one really mentioned at what speed.

 

I don't own an M11x so you will have to try this and let us know.

If you overclock the BCLK in the bios and don't want to reboot and reset that when you switch to battery power then you can also use ThrottleStop to lower the multiplier or to Disable Turbo boost entirely and this should increase your battery life because your CPU will be running slower than what it was originally designed to run at.

Do some testing with this.

If you are gaming on battery power and using ThrottleStop to increase the TDP/TDC and to make your laptop run faster then that will consume more power. ThrottleStop is very flexible though and doesn't force you to do that. The decision is yours. It's dead simple to create a second profile in ThrottleStop and then TS can switch profiles back and forth automatically based on whether you are on AC or battery power. You can still be the boss of your laptop and decide exactly how you want it to run.

Edit: Why would Dell include an overclocking feature on this model if they thought it would damage your computer? That doesn't make sense.

 

Leaving your system overclocked does not have any impact on battery life. That's been covered here in the forum a number times.

 

Well, in my opinion the overclock should be tested prior to making an assessment on what the maximum BCLK would be appropriate.

Let me elaborate on that. The o/c can be tested for stability with Prime95. But the moment that all the cores go into full load, the multiplier goes down to 8. Therefore, you can`t really assess the maximum processor frequency that is stable.

This is where ThrottleStop can help. By enabling it, you get more accurate information. For instance, an 165bclk can be stable when running without Throttlestop, because when you full load all the cores, the frequency is 165x8 = 1320MHz, which is rock stable. But, when you use TS to pump up the multiplier to 12x, giving a 1980MHz processor frequency, you might experience errors or bluescreens or crashes.

So my advice is use throttlestop while testing different values of the bclk with Prime95 SmallFFT or LargeFFT. When you`ve reached a value that is rock stable (I usually consider 8 hours of LargeFFT testing in Prime95 without errors to be extremely stable), you can keep that for the rest of the M11x's lifetime. During gaming or other activities, the temperatures reached with Prime95 will never be reached (unless you use it for rendering in 3DSMax that is).

Ofcourse, you should use power management options and tweaks to prolong the battery life, as well as TS profiles or whatever rocks your world

Fact is, maximum processor performance with TS should only be used when the computer is plugged in, since it can and will shorten your battery life if the processor is running full speed overclocked

 

Anyone really concerned about learning more about battery power consumption should use the Windows Performance Monitor function to monitor that. It will show you total power consumption of your laptop so you can draw a graph while playing your favorite game and see what ThrottleStop is doing to your battery life.

Here's an example form the Core 2 world which showed how efficient Intel's CPUs are no matter how you have them set up.

http://forum.notebookreview.com/6369641-post5060.html

Threads in the Core i design can enter the C6 sleep state at idle where the core voltage drops down to zero. It doesn't much matter what your BCLK is when the core is this asleep and this is constantly happening at idle in the Core i. Intel has done a great job with these CPUs and they tend to be very efficient at idle given the performance they can deliver.

 

That is exactly what I had to go through to get mine stable. I haven’t done an extended test yet but what testing I did do led me to move from my “normal” OC setting of 164 (would not boot at 166), to a setting of 160. 164 and 162 would BSOD at times during the full load test and I came to the same conclusion that you spoke of.

And by the way, awesome software, thanks Unclewebb!

 

I have never noticed any difference in battery life whether I OC or not. If there is a hit from overclocking, it is not very noticeable. YMMV

 

ive always oc @166 never seen a difference in battery and its always been rock stable.
with 2.88.2 i have done 5 or 6 runs all approx 1 sec faster. range from late 24 secs to low 25's. My 3dmar06 CPU score has gone up by approx 450 - 500 points. My GPU score has gone down though which is weird. i now get around 7800 oversall when i have had 8105 on a lower GPU oc.
I will go and try some CPU intesive games like BC2 and F1 2010. This is all very exiting and making me even happier with my purchase.
Uncle you are the best thing thats happend to this forum in a while.
cheers my friend

 

You may be going too high of an OC on your GPU! I have seen this exact same issue. What I did was monitor the clocks with MSI afterburner and I noticed when I OC too much on the GPU that the clocks default back down to a “safe” clock which is much lower than stock. It sounds like a "gimped" mode where it still operates but kills your performance. After that happens you have to reboot the computer to get the it working again, otherwise you are stuck at those lower clocks.

This threw me for a loop too until I read about it in another thread. Basically I verified it by watching the clocks and it all makes sense now.

 

ahhhhhhhhhh cool
will re boot and lower clocks and test again
thanks

 

You should also be able to use GPU-Z to log your GPU's performance while testing to make sure that it's not reducing your clocks.

 

After using TS, my CPU score jumped around 500 pts to 2200pts in 3dmark06, but my overall score is arond 6900 due to my GPU - I currently have this OC'd to 500/(whatever the linked value is)/820. What do you guys have your GPU's OC'd at?

 

I was clear up to 580/1450/880 before I figured out what was going on. I had no heat issues and no lockups so I thought all was well until I did a 3dmark and noticed the drop. Then verified the performance loss in a quick game test as well.

I dropped back down to stock and did some testing to get a new baseline and then started working on a new OC setting. My first OC was fairly conservative at 500/1200/825 if I remember right. I am still working on my absolute limit but so far I am back up to 535/1270/850 I think. I tried 550/1290/860 but the problem came back so I know I can’t go too much further.

The thing that bothers me is every single other time I’ve OC a GPU I would know when I was going too far because of A)heat issues that would start to show up and B)artifacts or lockups would occur. On the m11x I have neither, no heat issues nor any artifacts or lockups so it was hard to identify.

 

here is the link to my best 3dmark score using 543/1296/905

Imageshack - 13949955.png

I have used this with TS at 20/60 on 2.88.2 but my score overall is 7700 approx over 2 or 3 runs. Is it what is given in one hand (CPU) is taken from the other (GPU) for some reason?.

 

Why not crank up the TDP/TDC setting much higher to about 60/60 so this is not limiting you in any way.

I've heard some GPUs are using self correcting memory. When you start pushing them too hard and the memory starts getting corrupted, it will correct those errors but will slow down your performance in the process. Something like that might be going on. Your computer will still be stable but performance will decrease just before you get to the limit.

 

We probably should not waste any more posts in this thread since it has little to do with Throttlestop.

Probably it is (GPU limiting but covered up by CPU increases), however I can’t see your image here at work, but from my best score so far here is what I can say. This is not 100% accurate as it’s just from memory. My 8100 score was just over 2100 CPU, and about 3500 in both graphics scores.

From your scores and those clocks I would think you are exhibiting the same issue I was, otherwise your score would be much higher. I think stock GPU clocks with TS enabled fully (with CPU OC at 160) I was getting 7600, so you’re about where I was stock GPU clocks. All of my testing is done with all 80+ processes running, all programs running, no special tweaks or anything to get a higher score, just plain ol “as I run my games”. I use the 1280x768 for 3dmark06 testing.

 

I agree probably wrong place to discuss 3d mark and GPU but as im not that tech just want to clarify if it is TS thats effecting the GPU. Quick last note I have done all the same settings on EVGA P as my highest score(without TS) but overall its now 7779 with TS.
I have Powermizer turned off etc. Not that 3dmark matters any its real gaming performance that matters so thats what I will now go and do.

 

Thanks guys for bringing this up. It's not that off topic at all. As far as I know, the TDP/TDC setting in ThrottleStop controls the entire CPU and Intel GPU. This stuff isn't publicly documented so anything you learn about it might be very useful.

 

I beleive we were both talking about the Nvidia GPU, not the Intel. That's why I mentioned it was off topic. If it was the Intel GPU our numbers would be closer to 1000 vs 8000

 

For R2's all the major components are cooled by an integrated heat pipe, unlike R1, where I believe there are two seperate coolers for each component. The GPU puts out more heat than the CPU, so when you're cranking up your GPU, it's going to affect how hot your CPU gets as they're both tied to the same cooling system.

I use FurMark (MSI Kombuster) to test all my desktop GPUs. If I get instability running it at full load, I back off my OC. I believe Dell specifically stated that using tools like FurMark can damage your GPU as it pushes it to levels unintended. That basically means to me that the cooling system can handle normal gaming but can't handle extreme maximum output, which is why I don't try to push my GPU OC too much.

I'd be curious to see those that have high GPU OC's (core not memory) if you ran FurMark and see what it did. I'm betting you would crash pretty quickly, especially while running a instance of LinX.

 

someone with TS and a good GPU OC with an i7 can top 9k no problem. I have an i5 only and look at that CPU score. that is as good as a 2.53ghz core2duo non ULV in my old laptop.

 

GDDR5 has EEC. GDDR3/4 don't.

However, it's still upto the memroy controller on the GPU.

For ATi, the memory checks itself for errors, and requests the data to be sent again.
For nVidia, the memory checks, and corrects the errors, without requesting the data to be sent again in it's entirety (as ATi's older controller does).

It's just different mindsets. For one, nVidia controllers are not as easy to OC to high clock speeds (from what I see). For another, ATi controllers are already at rediculously fast clock speeds to boot.

 

WOW with the new .3 i've been testing for Kevin SC2 on a 3v3 replay goes from 41 fps average (TS off) to 57 fps average (TS on)... i think Kevin has finally found a way to lock the multiplier to 12.00X not 8, not 11.5... not even 11.90... 12.00 LOCKED baby. L4D2 is showing about a 5-10 fps increase during mass zombies. Kevin is there a donate link you can provide us? I'd like to show some appreciation and love.

 

Absolutely would donate. Heck, I'll even cheer for the Oilers for a game or two.

 

holly monkey.. I just changed my FSB from 158 to 154 because after doing a stress test the computer finally crashed, it hasnt crashed at 154 yet. so I downloaded the new 2.88.3 and I got this:

27.056
26.645
26.556

that is something.. thanks dude, we should all get together and get this guy an M11X R2 for xmas

 

I wish I knew why my times are about 10 seconds slower than everyone elses Here is my log and screenshot. 45s down to 36s for me. CPU OC at 160mhz.

ThrottleStopLog.zip

 

I'll include the donate feature in the next release. The M11x is running like a beast with this new feature. I originally thought the Supercharge word in the thread title was over doing it but not no more.

Version 2.88.3 has a little secret for fine tuning the turbo boost setting. If you go into the ThrottleStop.INI configuration file, you can play around with this a little to fine tune that last ounce of performance out of your M11x.

M11x=20

is presently the default setting. The lower you go, the faster your CPU will go.

Maximum performance is achieved with a setting of about 8 in early testing. Only go as low as you need to go to maintain full turbo boost. Instead of the 8 multiplier that this CPU was running at when fully loaded and overclocked, now it is running at very close to 12 for a 50% increase in CPU performance. I think you will notice that in day to day use.

Thanks for all the help from the NBR testers that provided their log data. This really helped me nail this feature and what I learned about battling against Windows, should help me improve ThrottleStop for other users too.

dajohu: Do some more testing with 2.88.3 and use the above secret. What TDP/TDC setting are you using? Also turn off Chipset Clock Modulation and the regular Clock Modulation. This is not a problem for this laptop and might slow things down.

 

Ctrl + Alt + Delete open up task manager and close cpuZ it hogs 25% of the processor and it loads up automatically when you load up wprime

 

I tried that, but have never found anything like that running, plus the CPU shows 0% in task manager anyways.

Here is what I did find though, I started looking for anything that could be causing a problem and decided to disable my antivirus and afterburner. Ran again and then I get 26s. Aha! I thought i found something, however when I started them back up and tried again to try to repeat the problem, I'm still at 26s. Wonder what happened earlier? I bet I tried it 15+ times over about an hour period and everytime it was 44-46s without and 33-36s with TS.

*Edit*
After setting up throttlestop to autostart and rebooting a couple times I retested it and have not got it to be slow again. I really wish I knew what caused it to run so much slower. I will have to recheck on every day now so I know it's running full speed.

 

Once again i am blown away. I installed the .2 version and got another massive performance leap in 3dmark 06

.0 vs .2
http://img259.imageshack.us/img259/4404/30336600.png

I am going to download .3 and see if it get even better.
I'll be sure to donate when you include the button.

 

ctrl+shift+esc doesn't switch you out to the "whaddya want to do" screen, and opens the task manager directly

 

Really?!? I think my head just exploded, i never knew that thanks