The final solution: A guide to fixing the M17x R3 580m thermal throttling

Hey everyone, I have been using a method that I believe avoids and disables the 77c thermal throttle on the Nvidia 580m GPU. This is a method suggested by another forum member a long time ago, but it is buried in another thread, and he doesn't go into very much detail. I actually could not find his post at this time, if/when I do I will update to credit him for the general idea.

Basically, he suggested that HWINFO64 was able to override the Compal EC. The EC controls the fans and power of the M17x R3, it is the source of the throttling problem. The fans are not aggressive enough, and after 77c the EC throttles the card. If we override it then we can enable a more aggressive fan profile, my testing also suggests that overriding the EC actually removes the 77c throttle completely.

------------
*Standard Warning - Note from NBR Moderation Team*

Third party BIOS Modifications, VBIOS Modifications and other system level tweaks (including fan control) can potentially cause damage to your system. Be aware of this before blindly flashing/installing.

If you brick your system or cause damage, chances are very high that your warranty will be useless. Know what you are doing, you are responsible for your actions.
------------

Step 1: If you haven't already, go use Widezu's guide and flash to the modified 580m vBios to remove the 63c power throttle. You want the normal one, NOT the .92 volt bios. You can find that guide here: http://forum.notebookreview.com/ali...throtting-fix-no-need-modify-system-bios.html

Step 2: Go download HWINFO64, I recommend the Self-installing EXE since you will need to use this program whenever you want to play extremely demanding games. You can download the newest version here: HWiNFO64 Download

Step 3: Install HWINFO64

Step 4: Now open HWINFO64, when it pops up make sure to check sensor only as shown, then go ahead and hit run.


Step 5: You should now be at the main screen for HWINFO64, the first thing you should do is go down the list and for each category of sensors, right click and disable monitoring. This is kind of optional, but I found that it make the program more reliable over long term use to disable everything I didn't care about. The only thing we want enabled is the Compal EC.


Step 6: Click the little fan icon at the bottom.


Step 7: The fan control window will pop up, system auto is what you click to reset to EC control. To setup our custom profile hit the Custom Auto button. The fan control look-up table will pop up. Go ahead and select the Compal EC in the dropdown box. Then go down the list and set it as shown in the image below. You are free to experiment here, but I have done about 20 hours of gameplay testing to get to these settings. Basically, anything over 4000 RPM will be max fan, thus 3900 is the highest medium fan setting, other than that I just do a nice slow increase. This is a good aggressive profile, but in my experience is not much louder or more annoying than the default EC.


Step 8: After you are done with the look-up table you can click OK, you should see the green active, hit the minimize all button and you are done.


General use: When you are done gaming and want to go back to the default fan control (which I recommend for normal use, since it is able to run each fan independently, where-as our profile runs both fans as a linked set) just open up HWINFO64 from your taskbar, click the System Auto button and quit.

When you want to game again all your settings should be saved, simply open HWINFO64, click the fan icon, and you should have the green Active under custom auto, hit minimize all and go to it. It literally takes 10 seconds to start-up.

Now, why do this? The more aggressive fan profile will keep your system cool under even the most demanding circumstances (at least in my experience). The few times my system ever gets up to 73c with this more aggressive profile the "jet engine" fan mode cools it down, I have never seen my machine gain temps with the fans at max. Even better, this seems to override the EC, so even if you go over 77c you shouldn't throttle. During my testing I tried setting the max fan to 3900 RPM rather than 5000. 3900 is the fastest fan setting before it kicks into "jet engine" mode, and generally it seemed to keep my system cool enough. I went about playing The Witcher 2 for a few hours, all worked well, no throttle. When I got done playing the max temperature shown for my GPU was 82c, the GPU had gone well over the 77c throttle, but nothing had happened, the throttle seemed to be gone. I added the 73c 5000 RPM entry after that, because I didn't want my system getting that hot, but it might technically be optional if the occasional jet engine bothers you.

Basically, this "safe" software fix has resolved the 580m throttling issue for me. It would be nice if Dell would step up and fix this with a bios update, but if they choose not to there is an easy solution. Plus, you really don't need to use this for every game, so far the only game I have needed to use this for was The Witcher 2, every other game (and I have played quite a few on my M17x) has stayed in the mid 60's using the standard EC even with my 20% overclock.

 

Mod applied. Much thanks for your work. If I spot the other post in my travels I will give you a heads up on the other guy's name. I tried to Rep you but I have already done so and it won't let me do it again.

 

+rep awesome little guide

 

+rep for a Great guide dude, hope this really fixes it for the time being.

 

I sent a message to Dell Support regarding this respectfully requesting that they increase the aggressiveness of the thermal profile for the fans and raise the throttle ceiling.

 

This is an awesome temp-fix until Dell decides to fix the throttle. The funny things about all of this is, we've done all the research for them. Just implement the change already, lol.

 

yoda, that's a great effort! +

 

Great guide. Definitely going to do this once my rig gets here.

*Finally able to give you rep. Hard to do it on a phone.

 

Nice guide... good work. For M17x R3 owners with GTX 580M this will be very helpful to avoid throttling issues while waiting for Dell/Alienware to deliver a solution that requires no user tweaks. +1 rep

Also useful for anyone else that wants to know how to use HWiNFO effectively with their Alienware laptop.

 

Just a side note, if you do this method via HWinfo, both fans on the m17x will be boosted which is a shame

 

Yeah that is the single downside. The fan profile will always control both fans so if your CPU heats up without the GPU doing anything, then the fans won't be kicking in.

If only there was something that controls both fans separately.

 

Yea, this is why I suggest to only use this for intense gaming, during general usage or even most games the default profile is better. It is only the very most demanding games where this is even needed. When you are playing intense games it doesn't matter that the fans are linked, the GPU will always be the hotter part so using it for fan speeds works fine. Plus, for whatever reason the CPU fans actually run slower than the GPU fans, so if I set the GPU fans at 3900 my CPU fans might be running at 3000 or even lower (even though they are technically a linked set). This helps to minimize the added noise of having both fans running while still keeping the CPU nice a cool.

 

I wonder if HWInfo can or could be updated to monitor and adjust fans by2 separate temperatures? As is increase fan speed to ____RPM when GPU and/or CPU hits __ degrees. Optimally the separate fans control would be best, but that would take some pretty system specific details for it to work on different laptops.

 

I'll have a word with the developer. I've asked him before but he told me he hadn't found a way yet.

 

+1 Thanks for some good information. If I theoretically ever receive my machine I will probably use it.

 

It does exactly the same thing with the M18x. Only, instead of two fans, it's three. That's OK for running benchmarks for a few minutes, but having three fans howling at top speed can get very loud. I would use the device that gets the hottest the fastest for fan control.

 

Looks promising

 

According to the HWINFO guys at the moment it just isn't possible with our system and EC. As you can see the options exist, they are just greyed out for us because the HWINFO devs have not found a way to fully interact with our EC.

This is what a lot of my testing was about, at least on my system 3900 RPM is the lowest "medium" fan setting, which isn't very loud (as in not really louder than the normal EC). You go to 4000 RPM or higher and the noise level probably doubles. Thus, my profile tries to remain below 3900 RPM as much as possible, and max speed is reserved for those occasional emergencies when the system gets too hot. You could probably raise the 5000 RPM temp up higher if you wanted, I mainly have it as an emergency catch-all, if the system is still gaining heat at 3900 RPM I want the fan to kick up before it gets too hot.

 

A forumer posted this solution a while ago, and told it on youtube. Well it's about time. After some experimenting I discovered this set up kinda messed up my system as I need to force restart in order for the fan to go back to normal. Also the gpu still heats up to 90+ and bsod occurs.

This is why I set LPC/EC support off(as according to hwinfo dev, m17x user should..or something), and for the fan speed, just adjust according to the given fixed value of 0, 2400, 4000( if i remember). I can't decide where is best to put 4k, so I just put at 70+ for 4k speed. This kept my gpu at 74 degrees.

 

Yea, he is the guy that gave me the idea, but at least all the info I saw that he provided was not very in-depth. I was actually having trouble with the program freezing when using the older version the other member suggested, and before I disabled all the other monitoring. You can try my guide if you want, I have never had it freeze since I set it up like this, and it is a lot more pleasant to run the fans between 3500 and 3900 RPM rather than in 4000+ RPM "jet engine" mode all the time.

I find that my machine almost never has to kick up to 4000+ RPM, so most of the time it is nice and quiet with this setup.

 

My laptop is currently being replaced. I will try it once i get it. Really hope we finally have a solution for the 77C throttle.
Thanks a lot *fingers crossed*

 

For some reason my GPU isn't being recognized. In step 7, I have no option under the reading drop down box. It also doesn't show up when I run 3DMark6. I'm able to see it just about everywhere else.

 

Which HwInfo version are you running, Sky?

 

I'm using the newest version of HWiNFO64. I even tried the 32bit version with now help. It'll recognize the GPU if I have a game running or 3DMark06. The only way I can get my GPU to be recognized is to apply the modded A08 BIOS. I've done that and have had no issues. Now I just don't have Optimus. Anyone know a fix for this besides using the modded BIOS?

 

You are selecting the Compal EC as shown in the guide?

I think it might be normal for it to disappear when Optimus shuts off the GPU, but that is why I chose the EC. On my machine the EC always shows up.

I guess my solution would just be to alt-tab out of a game when you need to run HWINFO64. Once you get it setup it only takes a few seconds to start, and it should work. Plus, as I mentioned, this fan profile is actually worse than the default for normal use, so you only want it when gaming anyway.

What I normally do is just start playing a game, often I don't throttle. If for whatever reason (higher ambient temps, less air flow, etc) I hit the 78C throttle then I alt-tab out and start HWINFO64 with this fan profile which solves the issue. In the past few weeks I have only had to do that once while playing BF3, so at least for me it is pretty uncommon. However, I had to use it every time when playing the Witcher 2, so it does depend on the games your playing.

 

One thing I wanted to do to my M11x, though not possible with it but looks like it might be with the M17xR3, is find a way to setup a hardware profile so that if I am on battery; the IGP is used and I can select optimus. If I am on AC power, it automatically disabled the IGP and just lets the 580m handle all the work.

Even if I have to reboot once, it would be perfectly acceptable. Two days a week I need amazing battery life (minimum 4 hours) and the rest I am fine with not even having a battery.

Doubt it is possible, but...

 

I followed every step of your guide. After the program loads, there is nothing under the Compal EC except for CPU. If I run a program that uses the 580m, then it would appear under the Compal EC. Since I was having issues with 3DMark not using the 580m, I just went in and installed the modded A08 BIOS. If I can find a fix for it so I can still utilize Optimus, I'll probably switch back.

 

@SkylineLvr, I also have no reading for GPU initially under Compal EC. What you need to do is to click on the small fan icon and select Custom Auto and them select it from the drop down.

 

There is nothing in the drop down box. Like I said, I followed the guide perfectly. It works great now that I have the modded A08 BIOS. It has something to do with Optimus not switching when I need it to. I've tried white listing programs with no success.

 

I used this guide to run Battlefield 3. It's working for several day.
Then suddenly bluescreen popup when playing game so now i'm not using it.

I found a good solution for myself.
I use "CoolFeet" to raise my laptop a little. I used a little one (about 1cm)
Now after i played 2-3 hour my high temp didn't exceed 69C lol
Before that my GTX580m throttle stop several time during BF3 play (78C limit)

 

I'm sorry, but I ain't exactly positive I understand the guide correctly.

"Step 1: If you haven't already, go use Widezu's guide and flash to the modified 580m vBios to remove the 63c power throttle. You want the normal one, NOT the .92 volt bios. You can find that guide here: [Guide] 580m GTX throtting fix. No need to modify system Bios."

Why would we need to flash the custom bios? Isn't the bios itself a solution to the 580m power throttle? Is there something I'm missing here?

 

There are 2 throttles. One at 63C and one at 78C. Flashing the bios removes the one at 63C. The rest of my guide removes the one at 78C. If you fully follow this guide the 580m (and it works for 675m as well) will have no throttles at all.

 

Thank you very much for a fast response, Yoda!

You're right, I missed that 63C throttling part.

A question then - isn't OC'ing both P0 and P1 to the same values in Nvidia Inspector a way to omit the 63C throttling? Or does it throttle anyway, and the custom BIOS is still required? My graphs do not show lower clocks above 63C (then again, even at 725/1650, my 580M rarely goes up to 64-65C [stock cooling, The Witcher 2/Battlefield 3, I seriously have no idea how's that even possible]).

Oh, and throttling or not - I am a tad paranoid when it comes to safe-guards; what is the maximum 'safe' temp for this particular GPU? Would the system shut down if it reached a dangerous threshold, with 78C throttling disabled?

Thanks in advance for taking your time to help me.

 

The main problem with P1 is that it lowers voltage with the stock bios. Thus, you can OC both states, but P1 is likely to crash since it has the lower voltage. I would expect at 725 that if you ever do go to P1 your system will crash, since that is almost certainly not stable at the lower voltage.

My guide doesn't so much disable the 78C throttle as it avoids it and makes it a little more forgiving. If your machine is still heating up even with my guide it will shut itself down in the mid 80'sC. The card is rated to 100C.

If you are not having problems with 78C then I would just flash the bios to avoid the 63C throttle (and any possible damage if P1 ever does engage with it OCed to 725, undervolting is surprisingly dangerous). Honestly, I run a constant OC very similar to yours and I almost never get close to 78C or need the procedure outlined here. You should only bother with this guide if you hit the 78C throttle (and it will be obvious if this happens).

 

Thank you A LOT!

I didn't know about the undervolting that occurs.
You got me scared right there for a sec.

Off to flashing I go!

Truly appreciated (both now and your guide).

Cheers!

Oh, sorry - one more question if you don't mind.

If I flash the custom bios, then I understand that P1 is actively gone altogether, correct?

In that case, what happens when I switch to "balanced" power profile in Windows?
Does it 'underclock' the card as usual? I sometimes prefer to play my less-demanding games at either 'balanced' or 'power saver'; less noise. Does it still apply after the flashing?

 

Nice title there brah.

 

OK, I'm getting some weird readings after flashing the custom vBios...

Is this how voltage is supposed to be? Weird spikes between 870 and 920 (even tho installed the regular .rom).

 

Was the screen during gaming? It should hold at .870 on both P0 and P1, it should not go higher during gaming.

You can check by opening up inspector, go to overclocking, and voltage is at the bottom. Insure that both P0 and P1 are set to .870 (you can't change voltage in inspector, but you can check it). If either is set to .92 you flashed the wrong bios. You want to flash the one that is just 580m, not 580m92v. If inspector shows both as .87 then that is just a known driver glitch that has been happening for a while. You can try updating your drivers from Nvidia to see if it fixes it.

 

It was outside of gaming, but I had origin running in the background (which sometimes causes the GPU jump to higher power profiles).

I closed it and played some Dragons Age: Origins - it didn't jump voltage anymore.
The voltage is capped at .87 on every profile in Nvidia Inspector.

I am positive I flashed 580m.rom, not the v92 version.

Newest Verde beta drivers from Nvidia - 306.02

 

It is just the known glitch then, it really isn't a concern. I think the last WHQL drivers didn't have it, but I use the newest beta drivers and don't worry about it. As long as the profiles are correct then it should not cause any problems. It is some kind of interaction with the boost feature added for 6XX cards and the fact that the 580m is the same as a 675m. It has been happening to people for a while now and I haven't heard of anything negative, it never does it under load so it isn't dangerous.

 

Not trying to hijack the thread to a different topic, but has this really been proven? Personally, I really want to believe this is true, as the overvolting is the only thing thats keeping me from migrating over to the 3XX drivers from 296.10. Am happy that this statement is coming from you, as you have been right about things on way more than one occasion

 

Best thing to do is just try it, the jumps to .92 when not under load are harmless. You just don't want it to do that under load. In some of the earlier 3XX drivers it would go to .92 even when gaming but I haven't seen or heard of that behavior on the newer drivers. Keep in mind they are up to 306, so there have been a bunch of revisions since the versions that had the serious .92 problems.

You can prove it to yourself by getting GPU-Z or another monitoring tool, game for a while, and see if there are any spikes during gaming. Worst that happens is you see that it does still spike and downgrade back. A tiny bit of .92v won't kill you. I don't recommend it for full time use, as it has been shown to kill cards after a few weeks months of full time use, but a quick test of the drivers won't hurt anything.

 

Sounds good, will try it out... So would I be right to say that the extra voltage (during idle) itself is fine, but if you make the card work while applying extra voltage is when we have a problem?

 

Exactly, the problem is that at .92v and running under load the card is drawing WAY over 100 watts, which is the max supported by the MXM slot. Once the MXM slot gets over spec then there are no guarantees on the quality of electricity (it may be uneven, have spikes, etc), not to mention the card is really cooking. It all adds up to an eventual failure.

At idle is really makes no difference, the card is not drawing much power, and the voltage itself isn't really harmful, it is the extra power draw and heat from the voltage that kills stuff.

 

Aah, I see it now... Nice use of simple words to explain a sly concept Thank you!

 

My m17x r3 is shutting down after 40 minutes of play, what will? Only happened after applying this FIX

my m17x is shutting down .. randomly

 

My does that as well, solved by adjusting the fan setting to 5000 rpm as soon as the temp hit 60.
The drawback is that the computer become so noisy while gaming.

 

I followed the steps you posted and my game is really good but suddenly my laptop keeps on shutting down.When i un-install HWInfo my laptop works fine.Anybody encounter the same prob?

 

The program will cause your fan controller to get stuck or confused and it'll stick to a low speed or shut off the fan causing a thermal shutdown. I have managed to fix it by simply unplugging the power from the laptop and plugging it back in but this is not a great solution to the problem.

 

Anybody have a fix for this issue?The HWInfo Shutting down my laptop?i don't mind the noise while gaming i used headphones anyways