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

Hi
i called dell regarding my 580m throttling issue , they said they may be willing to replace the gpu and the heatsink , they told me before that they would call me back to run some diagnostics . the thing is i did a clean install sometime back and deleted the recovery partitions ( i do have respawn discs) .....will this create a problem ???? let me know guys

 

Should not be a problem... As long as you havent damaged the system with a doing of your own, it should be fine...

 

Hi, No problem as previous poster says, Dell expect many AW users to work on their machines. Only if YOU do damage do they refuse warranty.

What country are you in? dell released a document (look back in this thread) about it being a known issue so replacing the heatsink is simply a matter of calling, explaining you have overheating, and they replace it.

If you want or think the card needs replacing maybe they now want some data. Mine started to artifact after constant overheating and following a heatsink replacement they then replaced both card and heatsink. Again no testing or diags.

Since it happens in games they cannot remotely see it happening but maybe they want you to install HWInfo and set it to log temps and clock speeds during a gaming session. This still only proves overheating, and a heatsink/repaste being needed.

What makes you think the card needs replacing?

 

i just spoke to dell and they said they will be replacing the heatsink only for now ........im not sure if it will fix the problem .....my throttling problems are severe like while playing battlefield 3 it can throttle in and under 20 sec ....im somehow managing using a cooler master u3 .........what do u think ? is the problem with heatsink or the card ??

 

Have you performed my fix just in case?

 

yup i did it long back......it throttles when it reaches 78 C

 

If you're hitting the 78C throttle that quick then you have a serious problem with either the airflow or the heatsink contact. The card should be OK, the throttle is to protect it from excessive heat. My card was running for an hour or so gaming before it throttled but was always close on the temp after about 15 mins or so.

You should check out your heat exchanger (the bit with all the cooling fins hanging off the heatsink). It could be totally clogged up? But if Dell are already coming to replace it then unless you can't wait...

Make sure you look at the old one though, if it is all clogged up then you know you need to clean it more often

Just another thought: have you been trying fan control with HWInfo? You can cause the fans to not work at all if it goes wrong.

 

The fans are clean , i did check them a few weeks back . i do use hwinfo sometimes but i have problems with it , it sometimes powers off the whole system , happens randomly

 

Yeah, I had some bad things happen with HWinfo although kicking the fans into top speed around 65C helped a lot. Since your 'peak' is in seconds then this might only help if you start them off at top speed. Since you know your fans are working OK it's prob not an issue but it does install a service so removing it will make sure it's not affecting fans in any way...

Lets see how a new heatsink/pads/paste go. If you've already repasted and it's the same then this will probably not help but at least dell will need to replace the card next

 

yeah hoping for the best

 

I would like to get into this thread and find out the resolve. I've had problems with this M17x R3 from the start... most with AW Customer Support (to the point I had to open a formal complaint). Finally got a new system 6 months later and BOOM, can't run a game that my 460m had maxed with no issues.

This laptop gets so hot it makes me actually sweat... when playing a game it throttles within 10 minutes (tf2 on low settings, trine 2 on low, bf3 on low, bc2 instantly throttles).

This has been a serious issue for me, so for the DELL reps responding, please let me know what the status of this issue is.

Entire System is Stock
Have never run gpu stress tests
M17xR3
580m (not sure which case vents I have)
Win7 Home Premium
i7 Extreme 2920xm
750gb in raid0
16gb Dual Ram


Thanks,
Jim

 

Hey Guys,

Sorry I've been out of the discussion for the past few weeks as I have just been slammed with work and travelling.

After quite a bit of time working with Dell over the 580m issue what we found in my case was that if you use the power throttle workarounds or disable Optimus using the "unlocked" bios' (as I did) it changes the characteristics of the built in power throttle and therefore you will hit the 78 deg throttle no matter what if you push the 580m with high GPU load. From my understanding the power throttle was put in place so that the GPU could not be pushed to these higher loads (producing a large amount of excessive heat) which became evident when Dell and I compared notes of GPU load in various games and applications. With the unlocked bios and power throttle workarounds my GPU was loaded significantly higher than the Dell unit using the factory bios, mine throttled, Dell's didn't.

The reason for this is the amount of heat produced by the 580m at very high GPU load is sufficient to cause premature failure of components and excessive skin temperature on the bottom of the case(see my previous post with the IR pictures).

I used to think this was Dell's problem that they had insufficient cooling but after I measured the current draw on the 580m at 95% load and the heat with a IR temperature gun I realized that what I was observing on my 580m at full load (95%) was actually beyond the MXM specifications listed on the MXM website. In other words the 580m was the issue as if the GPU met the thermal characteristics (at full load) as prescribed by the MXM specs then the 78 deg throttle would not be necessary.

Now there are workarounds like using the fans full blast, propping up the laptop or using a cooling pad but all of those have their own disadvantages.

As far as the heatsinks go if you are using the Dell bios and you're still hitting the 78 deg throttle then you need to repaste or call Dell for support. If you are using an "unlocked" bios or any of the power throttle workarounds then you will hit the 78 deg throttle at some point if you push the GPU to its maximum. For me the magic number was 95% GPU load.

Use GPU-Z to log your temps and GPU load. If you see high temps or throttle then make note of your GPU load and room temperature. If you're below 28 deg C room temp and 90% or below on GPU load you should not throttle. If you still do then most likely you'll need to repaste, have a bad heatsink or another issue and need to contact Dell support.

 

Hey Vytis, You back up how it happened for me and Dell's response. I'll call power throttle P1 and temp throttle P8.

P1 happened in most games. If I overrode it using the BIOS fix then temps gradually rose until P8. A repaste and replacement heatsink cured it for the most part unless I overclocked. The card was still giving me problems with artifacting - strangely in old games, so they replaced that. I never got a real chance to test it out since another (!) problem made me ask for a replacement, which they did

Now, the 675 (supposedly the same as the 580) is a different beast altogether in the new R4. I've only seen P1 in high stress games and then only if I OC above 700. It also seems to have a different algorithm since before P1 would stay for a few minutes after load had dropped, on this one it goes in and out very quickly.

Since I can't find a VBIOS for the 675 I've given up on getting any more out of this card. I can of course use NvidiaInspector to up the P1 state but as I can't override the voltage drop that accompanies it it's not viable to OC.

I'm now just waiting on the 680 to get to people here to see if it's worth saving up the pennies and move on from a clearly problematic card

Edit: The dell engineer on the call said that the GPU usage we are observing is not a true reflection of usage. It approximates in the most part but is actually frame buffer usage. He had utilites (that he said were confidential) that read actual utilization.

 

Micky, You're correct about the utilization being an approximation (and frame buffer is a big part of the number) but regardless the GPU is a sum of all the components and like we saw in the IR pictures the GPU core doesn't seem to produce as much heat as some of the other components.

The 95% load reference I worked from shows at that measured state in GPU-Z the card produces sufficient heat that it will hit the 78 deg throttle unless the power consumption is throttled prior to or cooling is increased. Even the 78 deg temperature for the throttle is a relative number based on Dell's algorithm and not based on any one component of the GPU but on the entire card.

My personal take on this is that the 580m core being a GF 114 core (same as a 560 desktop card) is capable of way more than what the mobile card configuration can handle as the 560 is listed as a 300 watt card and the 580m is a 100 watt card, that's a big difference.

As far as the 675m goes I am under the impression its the exact same hardware as the 580m but I also have read the 675m is less prone to throttle but not as capable of an overclocker as the 580m so it makes sense there are some bios/vbios changes responsible for the differences.

 

Thanks, isn't the 100 watt just related to the mmx specifications limit? Anyhow, I can bear out the 675 on OC'in being less. I just assumed a poorer chip quality in the one I got. Anything above 710 and I'll start to see small flashing 'lights' around the screen in some games

 

Am I reading this right...? We have to order a NEW GPU AND heatsink to resolve this issue...?

"For out of warranty customers, you will need to contact Dell Spare Parts for you region and purchase the part below -
RFYN9 N12E-GTX2 GeForce GTX580m Kit (Video card 3MF8R + Thermal Assemblies V1K2G)"

 

If your out of warranty, it appears you need to purchase the new parts. Shouldn't be an issue for you since your sig claims you like to spend a lot of money, lol. Sorry, had to do it.

I would email one of the reps one here and see if they could do anything to help you out. They are all very helpful, and may be able to pull some strings to help you out, if you're out of warranty.

 

I am not understanding how spending $600+ on a new GPU is a "solution" to the problem...? A new heatsink maybe, a vBIOS flash, maybe. A replacement GPU...? No. I sent an email off. We will see. And at one point, I had tons of money to chuck at whatever I wanted. Now, not so much. I was the first to test getting a 475M working with 3D to no avail, that was $900 in the trash.

 

Great. They called me back and said my laptop is out of warranty. I am wondering how can that be? I was given a FREE upgrade to 2 years because of the recall and everything...I log into my dell account and see that my 1/X/2013 has been changed to 1/X/2012 sometime in the last 6 months. I now get the joy of calling to find out WHAT HAPPENED TO MY YEAR UPGRADE

 

I paid out of pocket for a new heatsink and apparently I already had the redesigned one. Temps are EXACTLY the same as before. I hit 78c at around 140 seconds. Even when this was brand new I was unable to run FurMark for more than a couple minutes without throttling I am not sure how you are all getting away with it. I've tried EVERYTHING. The absolute ONLY difference I am doing is putting a thermal pad on the back row of RAM chips on the GPU.

 

Seems strange. In my M17x R3, it was literally impossible for me to hit 78C I could not get it to no matter how much stress. What thermal paste are you using and what amount?

 

I tried with Stock, Ceramique, IC Diamond...Same thing with all of them. If the card itself were faulty it would be having other issues so I am not sure where to place the blame. Even with stock on a brand new heatsink it shouldn't act like it did before. With fans at full blast it has gone up to and passed 80c

 

Also depends on how you apply the thermal paste I reckon and also your ambient temps. I mean I'm not really in a very cold place, I would say definitely over 24C ambient and even when I was in Thailand in 32C I wasn't able to hit 78C. Have you tried elevating the back?

 

I have a gaming laptop cooler and even with it on it doesn't seem to help. I have been applying thermal paste for a long time and am aware you are supposed to use a tiny amount just barely enough to layer the die. It's roughly 75-78f in the room I'm in probably close to your temps. I have tried elevating it as well.

 

This is very strange. I'm thinking that either you are not pasting/padding correctly, or there is actually a problem with the card.

When I went through all the overheating problems, and was on the conference call here, one thing that came up was that when I initially checked my R3 out the heatsink screws were not fully tightened. Also one of the pads was missing and another was not placed correctly. The AW engineer said that my heatsink and/or card could be 'warped' due to overheating when the heatsink was not correctly seated.

Now, you say you added extra pads to the memory. I thought all the memory chips already had pads? Since you got a new heatsink all the required pads should have been in place so there should be no issue there but if you have used a pad too thick then this will reduce the contact pressure and distort the card. (I assume you have used the pics in this thread of the pad layout to make sure they are all correct)

Maybe you could try to see how the heatsink and card are mating by cleaning it all up and applying a tiny dot of paste on each of the contact areas, and five tiny dots in a dice layout on the die. Screw it all together and then dissasemble carefully to see how it has spread? It might identify an area that is not making good contact for some reason...

 

I was wondering is it possible to use the hwinfo fan fix method while using the .92 vbios. I was playing a game last night and the fix seemed to be working fine, but when I tried overclocking a bit the temps on the gpu shot up to 91C then the laptop shut off.

 

I have and it's spreading evenly over the card after removal. The thermal pad I was referring to is on the bottom (opposite side of the GPU side) for the row of 4 RAM chips, I believe. The new heatsink didn't have those and it never originally had them.

 

@LV The bit you are referring to is the backplate or x-bracket/x-plate. The regular one covers the single row of chips to the left but not the rear ones. You have the correct configuration with additional thermal pads on the other rear ones that are not covered. It is very strange your card just seems to be uber hot. You sure your fan is working properly?

 

I'm struggling to remember the back of my card. I recall that the x-plate had a piece that extended over some chips - and had a pad but no extra pads were needed or used. If you added pads then are they in contact with something else to dissipate the heat? Grabbing at straws here but could they just be acting like a blanket keeping the heat in

 

There's a row of 4 chips on the right side which the x bracket covers. There's a row of another 4 chips at the bottom which nothing covers. I will say this though, since I have received my new heatsink and playing just Borderlands I have NOT had the game hit 78c throttle. I can still hit the throttle VERY easily in FurMark. What are yall's temps in FurMark and what settings are you using?

 

Well, the pads are for transferring heat to something not as a cooling pad in itself so I would say that if there is no contact point for the chips then take those pads off? They will cool better in air than having a pad air cooling without something to transfer to.

Just tried furmark at 1920x1080 no aa and got 68c max, most of the run was at 67C.

 

It says in your sig you are running an R4 and a 675M which is significantly different. The 675M and 580M are different but the BIOS's happen to work...Not really a good comparison

 

The R4 is a replacement for my R3 580m which went through all the problems.

'Significantly different' I was lead to believe that the 675 is identical to the 580, just a rebadge?

 

675M=580M

You are correct, it is just a rebadge. 580M owners can flash their cards to the 675M vbios without any issues (that I know of)...

 

Thought so, thanks.

It is good comparative data but from my experience with 2 x 580's and now this 675 that there is some subtle difference - mainly with the way it handles P1 throttling and less able to work an ovecclock. But other than that I haven't seen anything.

 

Yes, my mistake. I believe it has a slightly better tolerance for heat since it is clocked higher though.

 

NP, same base clock (620) though . It really does not 'like' an OC and maybe that is something they did keep them cooler. Anything over ~710 and it artefacts. A voltage increase would probably help this (matching P0-P1 voltages helped a lot) but I'm over this card now. It works just fine in every game I try so now to consider the next level - 680m

Temps I am seeing now are virtually identical to when dell replaced both the card and heatsink (580m) - I had a different HW issue that prompted me to ask for a replacement.

 

Hi Guys,

Just checked back in to see if any progress with this issue, after just arranging a support call for my ongoing issues (games stuttering when the card hits 77degC, while the card cools down after ~half hour of playing then more frequently after that). Replacement card and heatsink assembly coming out, not confident that will make any difference though...

Anyway just from reading the end of this thread, is if fair to say the 675m doesn't suffer from overheating in the R4? If that's the case what is the difference between the R3 and R4 that make the card work properly (I gather the 580m/675m are similar if not identical)?

As my support contract is due to finish the end of this month I want them to get if fixed finally and properly, whats the consensus on what **fixes** this problem, do some of the 580m card/heatsink assemblies actually perform properly (i.e. allow you to play games on your system without throttling in normal use, and for the life of the system) or is (as I believe) the system is flawed and will never work properly (got a feeling the aircon's been on in the lab when they test them or something...) ?

I must just mention Bill (and to a lesser extent Mark) have been very helpful thus far, but being Dell USA and I'm in the UK, not able to help with replacement heat sinks as suggested.

Cheers,

Matt

 

from what I've seen a lot of the problems are down to poor quality control on the R3. I've see quite a few pictures of misplaced pads, missing pads (as I had), and the most common one is the thin strip of compound that covers the VR's. It gets moved during assembly and ends up making things worse by getting in the way of other pads.

With both a card and heatsink on the way you should be good. Just hammer it as soon as you can to test it out. I found temps way down after both were replaced on mine - although it started to hang if the temp got above 72c (rare) - hence the replacement

 

Announcement:

My final post on NBR as a Dell representative.

 

Sure is Dell's loss! Good luck with your new job and thanks for all your help here.