[Conference Call] Nvidia GeForce GTX 580M 78C° Throttling [SEE POST #191 and later]

Hello working on my first post here.
I bought my m17x r3 with GTX 580m on 30.01.2012 and I am happy with my buy because in the country where I live Alienware products are very rare so it is at least a good conversation starter and overall it suites for my needs.
I use the computer mostly for working - running CAD and power system simulation software on it and works fine but when I use it for gaming I have lags and thanks to this forum I know it caused by GPU throttling.
Right now the system is totally stock so I am battling with the 63 degree throttle.
So is the wiedzu69 vbios update guide best way but unofficial solution to fix it?
I am not so informed about this issue in this thread but what I understand is that only solution to fix the 78 degree throttling is to apply for new replacement heat sink and hope that temperature doesn’t hit 78 degrees?
I think that main culprit to throttling is bad paste job because the heating system in my apartment holds the ambient temperature 19-20 degrees right now in summer the max temp can reach 33 degrees for few weeks and the m17x r3 should not have problems keeping things cool.
Last question is about warranty do I void the warranty if I open up the computer to repaste the GPU?

 

I'll take that one. There are about 20 people on this thread smarter than me to respond to the rest.

No. You will not void your warranty if you open the case and repaste. You are free to do whatever you want with your system. Your money. Your system. Enjoy it. Tweak it. Play with it. Tear it down to your heart's content.

The only thing that will void your warranty is if you damage any of the components performing ANY activity. That would include not grounding yourself and shorting something (hard for us to prove), screwdriver slip that damages something, Furmark madness for an hour (also hard to prove), etc. Even in cases where you have damaged the system, we can bring it in to the depot, for a fee, repair it and reinstate your warranty.

If a tech ever tells you differently, feel free to point them to this post and shoot me a PM with their Rep ID, which they are required to provide upon request. I'll point them to the appropriate internal documentation.

 

Thank you for the answer Dell-bill
But the warranty is actually quite complicated issue in my case.
I live in EU more exact in Estonia and here i don't have direct support form DELL. The warranty goes through the authorized distributor. My M17x r3 came form UK through authorized distributor and has 1 year alienware stock warranty and 1 year distributor warranty because of the consumer protection law. So basically i have 3 possible ways to mitigate issues: A) DIY B) hope that distributor deals with it C) rig will be sent to UK which means 2 weeks without computer.

 

Yeah, I should probably have called out the caveat that this applies in most regions. Some places where our support infrastructure is shared or not fully developed have different policies. If you ever have some egregious problems, I'm here to help with whatever influence we can bring to bear on the re-seller (often very little).

 

Thanks

Good to know that i and other users who may be in the same situation have support in the warranty issue if somethings happens. I think the most beneficial is confirmation when something happens and it should be covered with warranty then i have basis to complain

 

I see Nvidia have announced the release of the 6xxM chips yesterday. Articles : GeForce 600M Notebooks: Powerful and Efficient - GeForce The main thing they seem to be 'selling' is it's far superior power management and utilisation: Quote

'The most important consideration for notebooks is battery life, and to get great battery life you need a GPU that has great performance per watt.
Compared to the 500M series, the 600M series is twice as efficient'.

But all it's clocks look the same (GTX675m)

If this is the case then with one of these the dell engineers will have much more 'wiggle' room for the temp throttling and the P1 power throttle should disappear, or at least they should run cooler. Not any help to us 580 owners but I think it goes towards the root cause of our problems and dells reticence to fix them (totally) with software.
I for one wouldn't mind paying cost for an upgrade when it becomes available for the m17 if it puts this whole episode behind me, but that would require a herculean effort of good-will on dells part not to mention other OEM's interests...

 

The 675m is identical to the 580m. It is a rebrand. Not a refresh.

 

So all that stuff about the 6xx being more power efficient is just marketing?

 

No the other cards such as the 660m, 650m and 640m are genuine Kepler cards and supposedly more efficient. When the 680m comes out, it will likely be a real Kepler card. Efficiency goes both ways. You can either have the same performance with much less power demand such as with the cards I mentioned earlier or use the same power and deliver more performance which is what I'm hoping the 680m will be.

 

BTW, has nothing to do with the topic but is in response to my earlier post about not being able to attend the conference call...

I got the job!!! Now I gotta begin looking for a comfortable case (non-backpack) for my new system. Anyway...

Bill, since the 675m will be nothing more than a rebrand and the 680 will be better with power efficiency; can we put a bug in your ear to maybe pass that bug along to the development team that we would like the make sure our BIOS can accommodate them?

Not only would it help with goodwill (making us happy) but would also make Dell money by selling them AND save them money by lowering the power requirements of our M17x's, thereby causing less stress. Less stress means less warranty exchanges...

Just a thought...

 

I will pass it along. It's that "performance per watt" bit that might mean no joy, though. As widezu kind of eluded to. It might draw the same wattage and perform better. Hence, more performance per watt. We'll have to see what happens. Either way, the idea will be floated to the decision makers.

 

I dont even get crdit for bringing up the option near the end of the call about a bios with more aggressive fan options. haha.

On a bit of a remote issue that sort of brought up in the call, I was wondering about the known issues with the drivers. I am having the issue now, which I didnt have before, that when I close my screenand leave it closed for a while that when I finally do get to opening it up again it doesnt turn back on. It is as if the screen just doesnt get triggered to turn on. I am also starting to get chimes from the system as if wheen I close and open the lid, that he system is detecting newhardware plugged in. (same sound when you plug in a portable hard drve, new mouse, etc)

I am thinking that with the new nvidia drivers and the hdmi-in issue that causes the i/o port to shut off, that I may be having a similar issue as far as the i/o port shutting off and not coming back on.

My monitor shows as PnP in the device manager, and I dont know if it should show tagged differently, but I doubt it.

I am going to roll back to the 285 drivers and see fi that helps, but since this may be even remotely related to a known issue, I wanted to see if this sounds like it may be a part of that or not.

Any thoughts?

 

Many driver iterations ago, I distinctly remember an option about a standby signal being sent to the monitor and it having 2 or 3 options. I am not sure where this option went or if it still exists but in another application, but I am willing to bet if we (or just I) can find that option again we can remedy the standby issue.

The HDMI-In I am VERY happy to see as a known issue and being worked on. I just tried my 360 on my laptop and it does the same as everyone else's. Its there for a second and then disappears. It really looks like what my 47" did when I had my desktop set for a non-standard resolution and tried to plug it into the TV.

 

Interesting to know. I figured it was some sort of standby option that wasnt waking up the card or the port or whatever, even though have my customer profile set to never shut anything off. Most recently I was napping while a game installed. GTA4 EFLC and it was taking a bit, and I was exhausted so I napped and when I woke up the monitor wouldnt turn back on.

However, I am now using the 285 drivers from the dell support site and the little chime is gone, as it should be, and we will see if the other issue resolved itself. I think I am just going to stick with 285 for a while.

 

One thing about the fan profile that Louis eluded to in the call, from what I understood, is that the control of the fan in the bios is via the core temp reading of the GPU which is what we see in HWINFO and GPU-Z. The other comment I understood was that marketing asks for as quiet a system as possible but that even at full throttle the fans would do zilch, nada, zip if the heat sink wasn't spot on.

So while the core temps may be changed via a fan profile the other components (like the frame buffer or VREG) may or may not be afffected at all unless the heat sink and paste job is damn near perfect.

What Dell is worried about is not just the core temp, in fact, from what I understood the core temp is the easiest to manage. That's what we see in GPU-Z or HWINFO. There is no reading for other components like VREG or frame buffer temp for the GPU, it's all based on core temp.

Based on the math, the algorithms Louis eluded to, the 78 deg number (x) is a result of what apparently the core temp would be mathematically if the other measured components (Y,Z, etc.) are within tolerances.

So here's an approximation of Dell's thinking that I got out of the call:

If VREG limit (Y) is temp (Z) and frame buffer limit (X) is temp (Z1) then core temp limit is (78 deg).

So even if the fan profile is redone it may not mean the Frame Buffer, VREG or other GPU components that Dell is concerned with are being cooled efficiently and therefore while we may think the fan control would work it all depends on the math and the overall thermals of the card WHICH WE CAN'T SEE EFFECTIVELY WITH SOFTWARE! That's why Louis kept mentioning his "special" tool for measuring this very set of conditions.

In other words from what Louis said there is one and only one real fix for this issue and that is the correct (and correctly made) heat sink, proper installation with an extremely precise paste job and all while operating within ambient temp specifications. Rememember Dell tests at 28 deg ambient, running Furmark all day long and there are no issues and I for one wholly believe Louis on this one. The problem is the 580m makes a load of heat and when coupled with the poor factory quality of the heat sinks plus the damn near comical application of the paste, and, Voila! Throttle!

The fan control issue will likely drop our temps according to GPU-Z (and therefore appease a good many of us) but if the heat sink isn't right and we are getting a "false" reading of overall GPU temps for all the components then the result might be smoking alienwares flying out of windows simultaneously all over the world! Talk about UFO sightings!

After being privledged enough to be able to listen in on this call I am extremely confident that neither Bill, Chris or especially Louis will let this happen. If its not safe then it won't get done.

I truly think these guys are on our side and are absolutely trying to help us deal with the shoddy workmanship coming out of the factory. For what its worth I'll be asking for a new heat sink, reapplying myself after checking the quality of the sink itself and then we'll see where we stand.

Oh, and one more thing.

After all this, what do you guys think about starting a thread asking Dell that contributors to this forum be given special customized name plates that instead of reading "Built FOR" read "Built BY"...Especially Yoda and Wide, you two want one for Christmas?

 

Bold is what I was waiting for, no fix means no 580m for me right now, does not make sense to buy a expensive machine with a high class GPU if you cant utilize its power.

Just a couple of Q's below, might be answered did not read the whole thread.

-Specs and rules as in temp. and fan noise? Dont think fan noise is a problem since the 6990m can get to 90+ (?) degrees and the fans are at full blast, that would probably be the same with the 580m at 90+ (?) degrees.
-What are the specs from Nvidia regarding the temperature?
-A 580m with a ambient temperature of 35 degrees with the card stressed as far as possible. What are the GPU temps? Does anybody have that? (on stock machine, no fixes)
-Read that there is a member here with a 15 inch laptop with a Dell 580m that doenst throttle. What kind of laptop is he using and what are his temps?
-On other high performace laptops with 580m's such as Clevo, when does the card start to throttle on those machine's? 78 degrees or much higher?

 

I'm a little confused. Is the 675m identical (as it's specs seem) or is it same spec, lower power? That's a real 'cheat' if it is not any better power consumption. I see they are careful to talk about 'more power per watt'. Strange there is no 680 in the release but I've seen you mention it, maybe holding off for some marketing reason?

 

All that needs is an option in the bios "fan profile" -> (1)Best Accoustic (default) -> (2) Best Cooling

Just like some HDD's that have the option for high performance or high speed.. and the user can chose which suits them best.

Matt

 

Hmm, if the stock profile was to do what the hwinfo mod does I would not have seen the 78 throttle even though I am told that I have a poor pasting/warped heatsink problem so that is worrying. I've dome a bunch of testing, stock clocks, mild overclock, propped and not, as a reference point. Once I get my new heatsink I'll do the same and post the results.

 

The MSI GT783 17" and GT585R 15" have a GTX580M and a hotkey for TurboFan . It is loud but cooling is great

 

That would be a cool solution. I've used MSI kit on some desktop builds and they do always put in plenty of headroom for the enthusiast.

 

Vytis, your explanation of the issue a few posts back is very clear, thank you for this!

Does anyone (especially a Dell rep) know in what way a "new heatsink" differs from the heatsink in the new R3 I'm about to receive? In other words, should I ask Dell for a replacement after I receive my system because I somehow would get a better heatsink?

 

There are not different versions of heatsinks. It's a matter of whether the correct back plate is installed and whether the thermal pads are correctly placed from the factory. Chances are your new R3 may not be affected.

Dell/Alienware admits there are quality control issues and the factory that builds them for Alienware is not always on their best game. If everything is assembled correctly from the factory, your R3 will cool the 580M as good as any R3 is capable of operating with respect to the heat sink and thermal pads.

As an example of how things are sometimes not done correctly from the factory, my M18x had the incorrect back plate installed on the secondary GPU. The one it was supposed to have includes a thermal plate and pad for both banks of memory. The one installed from the factory only had that for one memory bank. After closer inspection, (yesterday night, in fact,) I found the back plates were installed on the wrong GPU (GPU0 back plate installed on GPU1 and vice versa).

 

Many tnx Mr Fox. How will I know if a correct backplate has been installed (apart from getting throttled). Or in other words, can I tell by for instance visual inspection whether my backplate is faulty? If yes, do you perhaps have a picture of a good and/or faulty part or know were I can find one?

 

I believe the backplate for GPU1 on the M18x is the same backplate used for the M17x R3 580M. Just a moment and I will post a photo for you, and other R3 owners can verify if that is accurate.

Edit: ok - here are some photos taken last night. I included a very large image so you can read the part number labels.

Spoiler

The back plate in the bag on the left side I believe is the correct part for the R3.


Take note of the yellow plastic insulator under the heat sink screw stands. These (or some other type of insulator) must be in place to prevent the back plate from grounding out against components on the PCB. If not, you will run into issues like this.



Below, I believe, is NOT the correct back plate for the M17x R3, but I'll let a savvy R3 owner confirm that.

 

Again many thanks for the help Mr Fox. Much appreciated!

Apart from the insulator material missing and the backplate being fysically abused (bended etc), is there anything else I should look out for?

 

Not that I am aware of, but I don't own an R3 so maybe there is something to look for that some of our R3 experts can mention.

If the insulator is missing and your back plate is grounded out on the PCB, you'll know it when you turn on the laptop. You'll have all sorts of issues. Look at the link I posted above that photo and you can see what I mean.

 

No it's the other way around. The L-shaped one is an M18x exclusive. The R3 has the single one and a large thermal pad goes on the other ram chips.

 

The narrow but very thick strip of thermal pad material (approximately 2mm) thick should be on the edge of the heat sink. One user found this piece to be fallen off and it was smashed on top of the other flat .5mm thermal pads instead of stuck on where it goes. It is not very sticky because it is so long and narrow. This piece is critical since it helps cool the voltage section of the card. Make sure it is stuck in the proper place and stays exactly there as you go back together with the assembly. There are pictures in the early pages of this thread. The voltage section of this card will get very hot if you don't get all of those thermal pads exactly where they go. Furthermore you have no way of monitoring this portion of the card since it does not expose its thermal sensors to the usual monitoring tools available to us.

 

Thanks for clearing that up for us, bro.

So, the attached is what the back plate should look like for the R3 (so there is no confusion).

 

Jody, you are referring to the pads where the red arrows are pointing, correct? (The heat sink is not for the R3, but I think those are the pads you are referring to.)

 

Yes, Exactly.

 

I think we are talking about the same thing. I went back earlier in this thread and robbed someone else's photo and marked it. It's all messed up in this photo and Louis was very clear that this is a critical section of the card to keep cool.

 

Thanks for that. Coming from a 6990M, I wasn't aware of having that area padded... even though I was getting no more than 65C no matter how hard I pushed it, I just opened my machine up and padded that area up. All good.

 

Thought you guys might find this interesting:

These are all thermal images of the 580m while running 3D Mark Vantage @ 26 deg ambient with modded fan profiles. According to GPU-Z max GPU temp was 67 deg.

Image #1: Temps at idle with fan very low, notice graph at bottom, these temps are cool even thought the colors may look intense:



Image #2: 10 seconds into bench, 750 core clock, fans just kicked in, GPU core at 67 which was the highest I saw in the log.



Image #3: At 95% GPU load, fans on max via custom profile notice that GPU core in center top is cool and temp here showed from 64-66 deg, yet look at bottom left corner



Image #4: Closeup of bottom left corner, notice temps, highest I saw here was 123 deg C!



Image #5: Credit to fellow forum member Kaltmond, I added the annotation, boxes around problem areas, box around MXM pins was also showing hotter than rest so I highlighted it:



Hottest I saw on the card are the mosfet chokes the arrow is pointing to. If you're interested in reading about these then check this out:

[Tutorial] Graphics Cards Voltage Regulator Modules (VRM) Explained - 3D Tech News and Pixel Hacking - Geeks3D.com

I also did some research on the MXM slots and I will post that later when I get a chance to collect my thoughts on it.

 

Really interesting. Guess I'm gonna try and get those area repadded

 

Awesome stuff, this really demonstrates what they said on the call, that core temp is not the concern here. What they are worried about is power draw and the related components, because while the core might be cool, those other components can be scorching.

Those temps are really just insane on the edges of the card.

 

Do you know what I see? The hottest bits are the bits that Dell have painted black as well as the labels and other additional stuff. The copper bits are actually the coolest. Go figure.

 

So ive been playing gta 4, testing out the new 301 nvidia drivers posted before. working pretty well. Without the fan profile through hwi I got throttle part way in, but with the fan profile I get around 71C

Wish i was lucky like some to max out core temps at 67C lol. However, with the new driver and the fan profile and the vbios, it seems like its working well. Just, as many people voiced opinion on, not fun havign to keep pulling up a custom fan profile each time to play games.


My only issue I have been having is with screen coming back on. When i close to screen and then later reopen it, there are rare cases it doesnt come back on, and I have to restart the system. This happens maybe 3 times in the last few days. Just lucky I didnt lose any unsaved work or anything.

Worried it might be a problem with the card itself, seems secure enough and works fine when on, save for the random freeze or two here or there in games), so kinda just testing now.

As far as setup for the new nvidia 301 drivers, just wanted to show that they seem to be working just fine. I know some people may have had trouble installing them, but these are nice, and the integrated powermizer disable feature helps. Sitting around the 40C mark with the custom fan profile.

 

This is exactly what I was hoping to see. Now that we know the area of the card that gets hot and has no heatsink; how do we mod the cooling system to accommodate it?

Perhaps get some ceramic based thermal paste (non-conductive) and add a thin layer, then add some copper shims to act as a bridge between that area and the copper pipes that are used for cooling now?

 

This information is both interesting and very very very useful. However, the card in picture is actually a Clevo card. A Dell 580M looks a tad different.

I padded the mosfet thingies, using the same thick sticky pad I used for the very long piece. Plus I put a tiny bit of PK-1 on the inductor thingies to make sure they were making contact with the thermal pad. Pretty sure I am covered now.

On stock fans and laptop slightly elevated at the back, my temps rise to 65C, then fans go full on the stock profile and it drops to 64C. This card is cool tbh, my Radeon went 69C max... both with a gaming overclock and ambient temps at 23C. Just shows that once the padding and paste job is done right, the beast shines.

 

Glad you noticed that the card I used for demonstration of where the hot spots are was a Clevo card. I did this for a reason but regardless here is a picture of a Dell card, can you tell any difference in the cards?



VS Clevo



Or this one which is a stock 480m:




Needless to say the thermal pics are of a 580m Dell OEM card from the factory and not of a Clevo card in a M17xR3. I only used the Clevo picture for demonstration purposes.

@Widezu, I also noticed that the hot spots are where the card is painted black, thought that was an interesting coincidence as well.

 

Just wanted to jump in with a public thank you to Dell-Bill and his colleagues. Just got a call from Alienware support, europe (India ), telling me they had received the request to replace my card and heatsink. usual caveat of 'if available' but engineer booked for tomorrow. Happy bunny - I'll be taking pics for anyone interested along with some comparative data.

 

*sigh* like ive said many times before, replacements do nothing. Let us know how it turns out but the 78C throttle will never go away. Just make sure you perform the power throttle fix in my sig if you run into issues

 

Just a question. I've been reading that an updated heatsink design is availabe. I recieved my system ealier in this month, does it have the updated design or do i have to call and get it?

Regards,
Howard

 

Yeah, I have read that but the dell engineer on the conference call was certain that i have a bad paste (and I found 2 of the heatsink screws loose) so I'll go through it and see - hopefully I'll never get to that temp as dell are saying - on stock clocks.
I already use your power throttle fix, the only change I make is to drop P1 by 5 just so I can see it cut in - thanks

 

@Mocky. Remember you will have to perform the fix again for a new card. Well if the paste job is bad then you will likely see an improvement

@hc2995, please post a pic of your heatsink so we can determine

 

Sure will once i get home. Do you need a ppicture of the x plate as well or just the heatsink itself?

 

hc, you should post photos of both, just in case the back plate is not the correct part.

 

Will do as soon as i get home tonight.