AW M15x Throttling Issue Investigation - Stock clocks and overclocked.

Well....I just don't understand why m17x's adapter doesn't fix the problem...there's no other reasonable explainations....

 

Well there's two other explanations that I can think of:

1. The motherboard doesn't have the capacity to furnish enough power to the GPU and CPU when they're under load so it throttles the GPU. The fix for this would be a motherboard revision and Dell would have to offer it to all existing M15x owners.

2. Bios design flaw.


I'm more inclined to believe it's reason #1 than #2 but I'd love to be wrong.

 

pm55, oh no!
Intel says pm55 mobile chipest will move to b3 stepping, is that gonna help?

 

Hey Joker, how do you read the result chart? What I mean is do you just read the core/memory speed and decide its throttling because the number are lower than what factory should be? sorry for the stupid question.

 

Hi. New M15x owner (well, technically the wife still owns it until the 25th) and I've been reading this thread for a few days.

Am I the only one that thinks this looks like a badly implemented "safety" feature that's throttling the GPU to prevent overheating? If the CPU and the GPU are both running at full whack then you'll obviously get a fair amount of heat generated, and perhaps instead of actually monitoring the internal temperature the BIOS, or whatever - I'm a software writer not a hardware expert , is throttling in a situation that's assumed will lead to overheating?

The question is, can the heatsinks and fans handle both processors running at full capacity for any length of time (in which case this "feature" could be disbaled) or is it actually necessary to have throttling there (but in a less abitary way than it is at present) in order to prevent dangerous and damaging overheating?

Someone, at some stage, must have made a decision to implement the throttling, so they must have had some reason...

 

Hey Joker, idea taken from another thread but wouldn't putting an ATI 5000 series in there To replace the 260 eleviate this problem since it consumes 15 less watts?

 

Now that sounds logical, if that's the case, it will just be up to dell to determine how long the machine can sustain the heat and power before throttling it down instead of throttling it down the very second everything's firewalled.

 

That assumes it's a power-related problem, which may or may not be the case.

Indeed. Maybe my cooling pad will continue to see active service after I ditch the wife's Inspiron 1520 and move on to a real machine. Still, really needs that Dell bloke to come back with some answers or an update.

 

No, this is not logical as the GPU and CPU's thermal management are not connected to each other. They will both hit their own respective heat ceiling at 100 percent load regardless of whether both are going full throttle at the same time or only one or the other is being utilized.

 

Well Joker, you did say it throttled less/later when you turned down the screen brightness. That would indicate there is something monitoring the total power draw. I suspect software is involved because power-restricted hardware tends to just heat up and/or not function.

Now the question is why did they limit it to 65W? If its thermal, a BIOS fix would nab it without any bad side effects. My fear is they cheaped out on a switcher someplace and it is going to blow if they draw too much from it.

Has anyone geeked out on the mobo to see what brand/model buck regulators they are using?

Some other ways to load the system would be occupying all 3 USB ports, turning on the webcam, spinning up the blueray/dvd drive, and using the wireless adapter. USB Ports max at about 2.0-2.5W each assuming you can load them down that far.

My wife just called to say mine arrived, and I'll try to report back soon on how my testing goes

 

Please don't take this the wrong way, it is a genuine question and not an attempt at provocation.

How do you know that the heat management of the GPU and CPU are always separate in the M15x? Could it not be that the load of both are monitored and together can trigger a particular behaviour if they're both being taxed to their limits? Additionally, it looks like the actual heat the CPU and GPU are producing is not a factor here, the throttling occurs regardless of the temperature (unless I'm very much mistaken).

Now this is all still theory. Really need that geezer from Dell to surface...

 

PPL, these things take time. Asking feedback from "the dell guy" everyday really doesn't help; they've already told us that they're following this thread, and will followup when they're ready.

Now, I'm not saying that ppl should stop running their own tests, or giving more feedback to this thread, but maybe stop asking "where's that guy from Dell" -- pretty sure, he's (they're) working on it.

 

Don't worry, I didn't take it the wrong way. I just know what the inside of an M15x looks like and the thermal compartments are separated. The CPU and GPU have their own dedicated heatsink assembly and fan, thus even if one is maxed out, it would have negligible effects on the other.

Therefore, heat is in fact, not a factor at all.

It may however, be tied into antiquated BIOS programing that automatically throttled the GPU past certain power draws, as in the past, the thermal components may very well have been tied together and this throttling would be a failsafe to increase the life of the components.

 

An interesting turn of events has taken place....

I just went online and noticed that the 180 dollar charge for restocking has been cancelled. Dell/Alienware looks to have stepped up to the plate.

Because of this single fact alone, I think I may hold on to the M15x and wait for a fix.

If this cancel of the restock remains in place, I may even call back my buddy and tell him that Dell/Alienware owned up and he should order....

Time will tell.

-D

 

Hopefully this is true, and props to you for willing to change your mind about the situation.

 

Yay Dfanucci!

 

Looks like mine is throttling as well.

Prime95 + Furmark pegs one of the CPUs, heats the GPU to about 76C, but starts ping-ponging the GPU clocks back at 50C.

Dimming the screen didn't help to fix it

 

im reading this thread and wondeirng wow all the pages on this in just a few days
im still confused when i get my M15x what should i look out for was this problem fixed by the 210 wat or is it the same as the 150w box

thanks

 

The 210 watt power supply did not fix the problem.

 

im trying to understand what thsi problem is exactly does this mean that the system is running slower then it should run ?
or what is this problem exactly im not really computer tech but i would like to understand what it is an all
whats going on when will i notice the problem what does this problem cause ?

 

Guys, we have a problem here.

Furmark + Prime95 yielded a 166W draw according to my Kill A Watt tester. Thats well over the 150W rating of the power adapter supplied!

When the GPU throttles I see the current draw decrease to about 120W, which seems more reasonable. This is with the 820QM processor, 4GB of ram and little else going on... so I'd hate to think what kind of current draw you'd get w/ Blueray and the Extreme CPU.

Someone needs to fire the EEs that designed this thing

 

Well i posted killwats for my asus (in the first page), and it also goes over the rated 120watt of my psu (up to 138W ), not sure if it matter, but they might be rated some number and have some room to spare, but maybe it also depends on the quality and design of the psu. Since joker already ran it with 210watt psu and didnt fix it, i think its not the psu but mobo or maybe software.

Btw joker have you ran WinXP to see if its independent from 7?

 

I'm fairly sure a PSU is capable of outputting more than its rated capability...

 

it outputs 150W.
it takes in 166W.
150/163 = 90% efficient, which is not bad. nothing wrong in the number you see.

some power supplies probably can output more than what they're rated for, but there's no guarantee it would be clean power.

 

A rating is a rating. Good engineering practice is to not only stay under the maximum rating of your components, but also offer a fair margin for variations in temperature, humidity and component construction.

I realize the bricks will probably hold up... maybe.... for a while. But who the hell really knows what components the Chinese used when they made it, eh?

Amazing how quick some of you are to say a PSU brick will operate beyond its maximum just fine. Look at how well that works in desktops

- True about the efficiency conversion, though its hard to know where it falls on an off the shelf brick like this. Keep in mind my system isn't fully loaded either, and I could easily add another 10-15W if I attached USB devices, used the Blueray drive or installed the faster CPU.

- Copying some files off a CD along with the aforementioned tests yielded 181W. How much does the M17x supply provide?

 

I don't have a kill a watt meter but I do know that even with the 210W (it's actually a 210/240W switchable so it shows up as 240W in the bios) power brick, the system throttles just as easily as it does with the 150W. So it's definitely not a power brick issue. The problem is with the motherboard and/or bios.

 

Basically the M15x is putting on the breaks only when pushed to it's maximum limit. At the moment this will not effect you but in the future when new games also push the limits the M15x will become obsolete before it's time because it can't reach maximum potential.

 

With turbo disabled in the bios, a bit more information is shed on the throttling issue. Notice that each time the 4 cores drop below 100% load, the GPU throttles with them. At peak CPU load (100%), the GPU stays at 100% until the CPU is throttled below 100% and with it the GPU also scales back it's clocks. Normally the GPU cycles far more frequently when Turbo mode is enabled but with it disabled, the rate of the GPU throttling is much slower.

Normally when Intel Burn Test is used by itself (not in conjunction with furmark), all 4 CPU cores stay pegged at 100% but when Furmark + Intel Burn Test are run together, the CPU clocks begin throttling together with the GPU clocks. This seems to further reinforce the idea that this throttling issue is power related. Click on the HD button to view the movie in 720p (in full screen) so you can see the clocks more clearly.

<param name="movie" value="http://www.youtube.com/v/0cWy8rieIJI&hl=en_US&fs=1&ap=%2526fmt%3D22"></param><param name="allowFullScreen" value="true"></param><param name="allowscriptaccess" value="always"></param><embed src="http://www.youtube.com/v/0cWy8rieIJI&hl=en_US&fs=1&ap=%2526fmt%3D22" type="application/x-shockwave-flash" allowscriptaccess="always" allowfullscreen="true" width='560' height="340"></embed></object>



Any thoughts?

 

Hm. This strongly turns me think it is a BIOS issue. I think it is too strange they're connected together. Beside of this *if* this would be a power issue I would expect rather freeze or lock (as usualy on the desktops) and not slowing down. Think twice: the power system is too weak to carry full load, and whole system (I mean Windows and all devices) all time 100% stable ? No way. At least HDD should encounter some problems with unstable (too low ? I don't think so) power.
Just my thoughts, nothing sure as far.

K.

 

This could be a BIOS issue related to power, in that the BIOS is trying to prevent the system (specifically the CPU and GPU together) from drawing more than what it thinks the PSU can supply. With turbo mode disabled the CPU draws less power (right?) which would explain why the throttling occurs less often - perhaps a fan speeds up to cool things down and this pushes it over the "edge" and throttling occurs to prevent too much current being drawn. With turbo mode on the maximum power consumption can be reached very quickly.

If the BIOS is trying to prevent too much consumption then that would explain why using a more powerul PSU didn't work - the power management instructions have no idea what the PSU is actually capable of supplying.

And where the hell is that guy from Dell with the answers?

 

2 tech guys I spoke with in a IT engineering program on campus also find it to believe it is a BIOS issue at this stage.

 

I have no doubt Dell is aware of the problem and that a bios can fix the problem, they probably just have to figure out

1) Timing, how long to let it go on before damage may occur.

2) Contact the OEM manufacturer to figure out exact hardware tolerance of heat and power.

3) Test it

4) Issue a verdict whether it be a recall, replacement of mobo, or issue a bios fix....

Just be patient, these kind of fix doesn't happen overnight. There are tons of background issues that has to be dealt with at each of the stage mentioned above. Rule #1 do not harm, they can't very well overnight something out that would start killing laptop, I'd rather they take their time and do it right

 

Well I just took a electronics course on circuitboards a month ago and what everyone especially Joker is describing I can say with almost certainty is a bios issue because the way the throttling is occuring can't be done from merely the motherboard. Electricity in a circuitboard doesn't work like that so it has to be a bios issue which is good because that can be fixed, even if they have to rewrite the entire bios. It shouldn't take long for a fix so long as someone is really working on it. Hell I'd do it myself if I had a better understand of programming bios, hopefully I'll learn more about that once I start on my BA.

 

I agree, but just for clarification, turbo mode only kicks in when the other cores are idle, and only 1 core is turbo'd; not all cores get turbo.

The case where turbo-mode is initially on, it takes less time for all cores to hit max usage, because 1 is already at/above 100%.

/me shakes fists

 

That's not correct. Turbo varies frequencies between Quad, Tri, Dual, and Single core usage.

 

So Panda, what are you going to do? keep the M15x and sell the G60 or return the AW and put your T9900 back in the G60?

 

Anyone check the ACPI code in the registry yet?
You might be able to disable the throttling.
I have helped people in the gateway forum disable CPU throttling, maybe your problem is similar?

 

Thanks for the clarification on my clarification =) however, how does turbo kick in under quad core usage? I understand the case of 3 cores using the headroom from the 4th core being idle.

If I underclock the i7, such that 4 cores run at 100%, but since it hasn't hit the temperature or power usage limits (because of the underclock), it can enable turbo on all 4 cores?

Sorry for OT, it'll be my last question about it.

 

I haven't seen that mentioned yet. I searched my registry for ACPI and found some gobbledygook, what do we need to do?

 

Follow the attached guide.

Upload your asl file for me. I'll take a look.
But I don't think I can stop GPU throttling by changing the ACPI code.
Still worth a try.

 

if you screw up on the registry thing...can't you really up the system? Or are you basically just disabling the turbo mode? Won't that then make the system run really slow on 1 core applications?

 

It has nothing to do with turbo mode.

If the code is there (which it might not be, depending on how the BIOS was programed), then it may be possible to stop thermal throttling.

It would be as easy as increasing the throttling trip point to 100C.

Yes, messing with the registry can cause hardware damage, but it's not likely.
What you will probably get (if it's done wrong) is a BSOD next time you restart the notebook. Then you would need to hold F8 and boot "last known good config".

I don't know if this will work, but it wont hurt to look at the code to see what's there.

 

From what I read, the guide is meant to change the fan settings. This would actually be a nice fix for Asus owners so I'll pass it on to them.

 

Keep me posted via PM. If the situation goes South again, I can jump in and fix it up.

*Full disclosure: I have huge bandwidth issues in a new role, so I'm not as responsive to NBR PM's as I used to be in my last role.

 

I couldn't agree more. I got an AW machine for this exact reason.

I wanted a laptop that would just work, and run everything I threw at it for this and subsequent years.

Why do the most expensive laptops on the market (more or less) not provide this?!

 

I get this exact same issue in World of Warcraft, irrespective of how I set the graphics - it just doesn't run consistently, which is very disappointing for such a 'powerful' and expensive laptop.

I wonder if it has anything to do with the throttling issue?

 

Does this mean that it's worth turning turbo mode off in the BIOS?

Will this gain better performance from the M15x?

How would this affect single core usage also?

 

5150Joker: I don't believe that the TMonitor program is 100% accurate when reporting these CPUs. During testing on the Core i7 Desktop chips, there seemed to be some problems with TMonitor 1.01.

CPU-Z is also written by the same company and has some major issues accurately reporting the multiplier on these new Core i7-720QM mobile chips. Any multiplier reported below 7 is physically impossible on a Core i7-720QM but I've seen multipliers being reported as low as 1X by CPU-Z. That's wrong.

The method Intel recommends to determine the multiplier is outlined in their November 2008 Intel Turbo White Paper. This is the method that the i7 Turbo program follows which can be downloaded here along with RealTemp.

http://www.sendspace.com/file/5100no

I'm not trying to bash the competition. I'm just recommending before you put too much trust in any tool, including the tools I've written, you need to investigate these tools further.

Here's some background info about how these Core i7 mobile CPUs work.

The i7-720QM has a default multiplier of 12 but when turbo is enabled and C3/C6 is also enabled, it will continuously cycle between 12 and 21 depending on the load and how many cores are in the active state. You need to enable C3/C6 in the bios. This allows cores that aren't doing anything to enter the inactive sleep state. The multiplier can be changing hundreds of times a second so reporting the average multiplier during a one second sampling period is the only correct way to report these new CPUs. That's what Intel recommends and that's exactly what i7 Turbo and RealTemp do.

The maximum multiplier when 4, 3, 2 or 1 core is active on a Core i7-720QM is 13, 13, 18 and 21. So when 4 cores are fully active you can still get a +1 turbo boost from the default 12 multiplier up to 13. When testing on a Dell laptop, some high stress programs might trigger turbo throttling where the multiplier will cycle between 13 and 12 when all 4 cores are fully loaded. This throttling is built into these CPUs and is based on core temperature and power consumption. Some programs create so much load and heat that you might not be able to get above 12 when fully loaded.

When idle with EIST enabled in the bios, there will be a setting in the Windows 7 or Vista Control Panel -> Power Options called Minimimum processor state. If you don't want your multiplier throttling back at idle then C1E needs to be disabled and the Minimum needs to be set to 100%. When testing at full load I'd also have the Minimum and Maximum set to 100% so they don't interfere with the multiplier. When you want your multiplier to throttle back at idle then it's a good idea to set the Minimum to a low number like 5%. It won't actually get down to 5% but it will allow the multiplier to get as low as physically possible which is 7.0X on a Core i7-720QM.

Edit: Hopefully you can add this download to your first post.

 

If I'm understanding this correctly ... we should be able to create the same type of throttling issue on a G51J using same setup outlined in the first post?

True?

/I was ready to pull the trigger on this laptop but will hold out to see how this plays out.

 

Yes, it was written to change fan settings, but you can do much more.