Upgraded to GTX675m, overheat/throttling. Heatsink?

Hi,


I've upgraded from GTX460m to GTX675m in my M17xR3 and are having loads of problem with overheating/throttling (still using 460 heatsink).
I have custom bios, and have tried everything regarding fan profiles to the point that I no longer can communicate with the fan chip without the application locking up.
Today im using a fan board under the computer, but I want to be able to bring the computer to friends and play without the need for external fan.

So my question is:

Im guessing I need to get another heatsink?
Is there any difference between the 675m heatsink that is "kinda" available on Ebay, or is is basicly the same as 460m heatsink?
I can't see any difference really...
What thermal pads should I use?


Regards
0verseer

 

Bro. Check your temperatures and post them.

Double check if the heatsink is attached properly and GPU DIE and Rams are making good contact with it.

460Heatsink should be good enough but it have to make a proper contact, and i think contact is the problem here.

Also, what thermal paste did you use ?

Stock pads should be good enough, but generally 1mm pads are most commonly used.

peace !

 

I reseated/reinstalled the card twice, however it was quite some time ago and used the
best paste according to some charts/tests on various sites like TomsHardware or whatever.
Cannot remember its name though, but I bought it after checking the tests, Im using stock pads

Im unable to communicate with fanchip if the driver is installed, but temps on idle is quite ok, around ~47-55c.
But on heavy load with high details for example in BF3, it goes all the way to 73c,
throttling kicks in droping all stats on the card (checking while it happens in GPUz), and I get around 2-4fps.

Is there any other workarounds instead of using fanprofiles to counter this damn throttling? I mean 73c is nothing!

Do not know if it is related, but im experiencing a smiliar problem while Idling (after about 5-10m).
In windows when im not touching anything, the GPU just shuts down for some reason... -_-
For example when im listening to spotify or watching netflix, after like 5-10 min audio/video starts stuttering/lagging until I move the mouse.

Im about to throw it out the window....

 

Might be the Vbios also, i have heard some cards were set to throttle at....63'C

Also, with the Old GPU everything is ok Right ?

 

I have the latest vbios, updated to see if there was any changes on throttling temp.
Don't think I hade the idle problem before update and no problems what so ever with 460 (except the fact that it sucks).

 

I also heard that R3 Motherboards have serious problems with chipset. That is why R3 users are getting R4 Mobos and put them IN.

I do not know if disabling ULPS can help ?

@sangemaru guide

 

Wow, thanks. Im definatly going to try that.

I got one of those crappy computers. Motherboard has already been replaced once, and the GTX460 also (after vbios messed it up).
Im thinking about maybe prolongong my warranty, so that I can get a new motherboard again, because clearly this one isn't working as it should.
One should not have the problems i've been having with this computer after a fresh windows installation (gazillion bsods, drivers issues, application issues).

 

If I were you I'd put the 460m back in, and see if you can get the machine replaced entirely.

 

Have you considered repasting the card with some Arctic Silver compound?

 

Hi, jumping in here with a few questions.

1. Did you get a new x-bracket with the 675? If you are using the old one it may not be contacting the core correctly. When you open it up there should be barely any paste left on the core with most squeezed out.

2. Was there any black tape on the heatsink surrounding the core area? If so, remove it, it gets in the way of a slightly larger core.

3. For the lockup in windows, this has been a common problem with all the NV drivers after 314.22. 320.xx started to fry older cards and mine died along with a few other NBR members. Stick with the dell one and just for test purposes I recommend getting their earlier one first and not the one they released a few weeks ago. Disabling hardware acceleration inside any apps that are running (browser usually) cleared up the random lockups and stuttering video for me, but the stuttering in games remained right up to the latest whql, where it is hardly noticeable. Running 'latencymon' will show if the issue is NV related (TDR) and if it will affect video/audio.

This 675 runs at around 50C~55C in windows, 65C in games and 68C max in high end games when overclocking. My last 675 would lockup solid at exactly 72C, they appear to not be the most reliable of cards. I'm on my third if you count the 580m, which is exactly the same card, in my R3

 

I've done this 2-3 times, and I think its arctic silver im using, can't find the syrringe thingy right now though.

hmm, I only got a card, using the old gt460m heat sink. I noticed though that there are some black parts on the so called 580/675 heat sink that is missing on mine.

No, it was clean.

This idle thing also happens in game, I was for instance playing War in the North once, and went out into the kitchen as I heard the sound start stuttering. But as soon as I touched the mouse, it stops. So its just not in windows.
But is there any official dell drivers for 675? The latest one I can find is "NV Graphics Driver Update for Win 7 & Win 8 for Alienware M17xR3, M17xR4, M14x, M14xR2, M18x, M18xR2 / Alienware Nvidia Win 7/Win 8 306.97 driver", but it doesn't say witch card its for.
Maybe this one will work?
Driver Details | Dell US

 

So, you used the old x-bracket. The height of the posts is what sets the pressure of the contact with the core so it may well be that that is causing poor contact. It will only be a tiny amount but may be enough to see the problem you are having. If you open it up and take a pic of the paste remaining on the core I may be able to say for sure?

NV drivers are able to support all NV cards. It's just that sometimes you have to modify one of the install files when the card is not listed by dell. The install will fail with a 'hardware not found' error.

If you search for the video drivers for new M17 you will see a new NV driver. Checking the compatibility list all the AW's are pretty much listed.

Driver Details | Dell UK

Stuttering in-game is also another big issue with a lot of people see, the interrupt table is getting hung up and moving the mouse wakes it up.

The driver you listed is probably a good starting point as it's earlier than all the problems were introduced. There is one earlier one also under the M17 drivers that is later than the one you found.

Good luck.

 

Yes i've disassembled it now, and took some photos.
Im aware of the modified inf thing, to install some drivers.
I will put it together again and try a reinstall... again, and perhaps the old drivers will help with some issues.
I am using the Arctic MX-4 paste.

 

It does look to me like there is too much left behind on the surface (not what is squeezed out, that is normal) but it is not obviously that. Other opinions welcome

All the pads look like they are in the right place. On mine the blue strip across the last pic is a kind of putty. Make sure the pads are not interfering as this is for the voltage regulators (VRM's) and are the hottest single part of the card. Also that it is making contact on both surfaces.

My 675m replacement heatsink from dell: The black tape area is what I would remove if I was fitting it to a different GPU

For the x-bracket I'm at a loss as to where you might get one. They should come with the card but maybe it is included in the heatsink/pad kit specific for the 675m from dell?

Hopefully you find a driver that gets rid of the stuttering and other stuff so you can concentrate on getting the temp down

 

Yea seems to be quite the same, except the blue one on yours seems to be a bit thicker. I have a thick pad in the middle left.
Would be nice to know which part of the card getting the hottest.
Maybe it could be one of my pads are overlapping each other...

 

Could be an overlap, this is what caused a lot of overheating problems for the 580m owners. The strip at the top is the putty I mention and looks like a blue pad strip to me on yours?

This thread ( http://forum.notebookreview.com/ali...x-580m-78c-throttling-see-post-191-later.html
) is a real drag to trawl through but there are things like thermal images of the card in use that you may find helpful. This showed that the VRM's are the hottest part.

The core is the important bit but if other areas are getting too hot it can have a knock-on affect.

 

Wow, thats quite the thread.

Yea the blue one at the top is a pad strip for me. I added one extra strip now to increase the thickness, also, the long thin one at the far top, some of my components aren't even touching that, don't know how hot those tiny ones get.

Which ones are the VRMs?

 

Not sure if there are two or three VRM's but they will all look the same and are quite small (about 5mm along their longest length) and along that strip. If they have no contact at all then they will quickly overheat, but you won't see it in temps (except as a knock-on) as only the core is measured.

 

mm ok.
Well, I've put everything together again, I bent the screw points upwards a tiny bit and also removed the distances for better fitting and have now a fresh install of windows 8.1.
I've been running Spotify for about 30 min now, without any lag. So as far as the lag issue goes, its something with the drivers alright.

In the meantime as Spotify was running, I read some pages (around 8-9) of the thread you posted.
As I don't really have the time to read 250+ pages, did they ever manage to find a solution to the throttle problem?
Seems like the Yoda guy, and another were working together with dell on that one.
Judging by the last pages I fear the worst?

 

Good news re the lag, fingers crossed.

Basically it's all in the first post, edited at the end of all the discussions around the conference call. Boils down to the fact that although dell could have come up with better thermal solutions like fan profiles or raising the limit, they would not do so because of affecting the whole thermal characteristics requiring re-certification.

The focus was on the fact that they should never get that hot at stock clocks. Dell offered to replace the heatsink of anyone having overheating issues and from what I saw many of the problems arose from poor paste/pads from the factory.

Haven't seen it happen again so a bad batch could have been part of the issue.

 

Ok, thats to bad -_-
Guess I would need to get a new heat sink from ebay then, since my warranty ended over a year ago.
I actually raised a ticket with dell on prolonging my warranty, they wanted proof of purchase and so on. I was also stupid enough to say that I've upgraded the graphics card....

Anyway, neither of the drivers I and you posted recognizes my card, so can't install.
I thought drivers provided by dell would actually work "out of the box" (its an Dell OEM card, so yea). Guess I would require some kind of modded inf still...

 

Dell will only supply you with one for the 580m but it's exactly the same card. It was not that expensive when I was quoted, around 24.00 £/$

Very strange re the dell driver failing. It did happen to me once time when after I had used a modded inf I always had to even with the original version . Which one did you get to work?

You can get the drivers and inf file from here: LaptopVideo2Go: Drivers

The driver files are simply a self-extracting zip and contain the original files. The inf link opens up the inf and you have to manually save it in the browser's 'save as' (make sure text document is NOT selected)

 

Here is my old one and the replacement. A bit fuzzy but I think you can see what was missing!

 

Im checking ebay right now, and the 675 heatsink does look abit different, hmm.
Can't find a fitting 580m heatsink though.

Dell Alienware M17x R3 R4 GeForce NV13E GS1 GTX675M GPU Heatsink MT1DN | eBay

So damn expensive...

 

That's a crazy price! Did you try dell for a 580m one? RFYN9 N12E-GTX2 GeForce GTX580m Kit (Video card 3MF8R + Thermal Assemblies V1K2G). From the earlier post link.

 

No, I havn't tried that yet.

I finally managed to get the 327.45 driver installed, It idles around 52-56c.
Ran FurMark burn in test in 1920x1080, 8x AA, for about 25 mins. It maxed out at 77c, but didn't see any throttling.

The idle lag seems to be gone, must've been the specific driver I was using at the time.
Is it possible, and would it help to under voltage the card? Would it make the card run cooler? I guess it would be some performance loss too.

Im thinking of maybe highjacking the cable, and creating my own fancontroller so that I manually can increase fan speed, anyone tried that yet?

 

Yeah, the driver you got to run is actually getting better feedback than the latest WHQL. We're overdue for another one, a lot of people having the same issues you did.

Messing with the voltages requires specific vBIOS files with the settings you want in them so unless you see one it's not possible.

Your temps are around 10c higher than mine so it looks like it's just poor contact caused by the minor size differences. One more degree and you'd hit the throttle!

You can use HWInfo to build your own fan table but it takes control of both fans so it can get noisy! If you want to try it search it out here on NBR, setting up fan control needs some instruction.

 

Yes, I am aware of the fan profile thingy, however this is where I hit the wall regarding this computer and mainly the reason I turned to this forum again.
Doesn't matter if I have CompalEC driver installed or not, or using safe mode, all communication with the compal EC fan chip results in HWiNFO locking up, fans behaving crazy like, dying completly or running constantly on max speed even after application crashed.
To remedy this I need to hard reboot the computer because a normal restart does not work, and if I try and kill the process I get the message saying "You do not have permission to kill this process".
So, basicly it feels like I got a $1900 paper weight atm...

 

Please don't use Furmark on your cards, it's nasty program, your GPU might die! Especially given that it's a 580M / 675M! Furmark creates unrealistic loads on the cards, and NVidia have been artificially limiting how their cards run on Furmark since the Fermi Generation. Unigine Heaven Benchmark is a good one to use to check if your card is OK, as it loads up your card to 100% with realistic loading that you'd see in games, it also runs on a loop so is useful for long tests.

I don't think manually increasing fanspeed will help, as at the higher temperatures you've mentioned your laptop will be running at 100% fan speed. On the M17xR3, the GPU kicks up to maximum fan rpm at 65 degC. Actually, it doesn't seem that you have a temperature problem given that you didn't see any throttling in Furmark. I'd play some games or run Heaven Benchmark and see what your temperatures & clocks are like under those situations.

 

I can't play demanding games more than 10-15 mins before throttling starts. Dayz for example starts throttling after 10 min, and that engine is old as hell. The settings I used in that game wasn't even demanding, and I had around 45-60 fps.
So I am pretty sure I have an overheating issue, and an issue with the Compal chip since I can't pull data from it.
Also, fan does not go 100% at 65c, more like 75%. I've heard the fan going alot faster the few times I actually managed to get HWiNFO fan profile to work.

However, Ungine gave me aprox 69-70c on 1920x1080, extreme settings. Seems a bit more normal, but still...

 

That's strange, I get 4000rpm on my GPU at 65 degC, which is near-as-damn-it the 4100rpm max that I can make it spin at when I set it to maximum in HWInfo, so not sure why your's would be different.

Anyway, yes, if you've noticed throttling in games then it does look like you have temperature issues. MickyD's given you some good advice, and I don't really have much more to add than he has already suggested. I think it's all about making sure the heatsink is sitting flat & close against the GPU core, affected by: pad placement & thickness, X-bracket used.

 

Have you watched the temps/clock in-game to see the trigger point. I'm guessing around 78c but worth a look using afterburner OSD. It might be the power throttle.

The fan not getting to 100% is the same as mine although even with a 710 OC it still never gets above 67c.

If it's the power throttle where the clock drops to 410 (or thereabouts) then you could try the vBIOS from my new dell card. I recently posted it for someone and it sorted their problem out (now if only I could remember the problem). I'm sure it's a newer one than I had in the original and it hardly ever goes to the P1/2D clock when gaming any more...

The 'mod' one has P0 and P1 matched at 620/0.87v so a power throttle has no effect. You can use nvidia inspector to adjust P1 but without the vbios mod there is a small voltage drop and this can make it artefact.

 

Yep, and I'd forget the whole Compal Fan Control thing you're talking about, it's a red herring, just let the laptop control it's own fans like it was set up from factory - if it's functioning correctly you'll be getting maximum fan speed well before the temperature throttle point of the GPU. If you're concerned about temperatures I'd focus my attention on the padding & pasting of the heatsink. And I guess the VBIOS flashing that MickyD talked about might be worthwhile, as well as the Afterburner OSD to monitor things first (just so you make sure if/when/what triggers the throttle - no point trying to diagnose a problem if you don't know the triggers).

 

Yes I would look into the paste job and make sure no thermal pads are folded over themselves along with the obvious cleaning etc.

It could also be a heatpipe is damaged which is impacting performance.

 

Thanks Micky, I've been able to run the system pretty stable now with the 675m drivers from dells page, it reaches about 69-73c in BF4 single player.
The game lets me know that the drivers are not recommended every time I start it though, so I do not know if later drivers will do much difference in terms of heat.

Im talking to the dell support on the side of this, and they are wondering if there is any official statement from dell regarding this "change to R4 motherboard" for R3 users, do you know?
Im thinking of disregarding their troubleshooting, since they will require me to change back to the 460m and flash their A12 bios (which will enable the damn optimus again...). They also want med to run with integrated graphics and so on. That wouldn't be so hard since the troubles I've been having with optimus is that the graphics cards refuses to change to secondary card when entering any game...
Anyway, Im going to try to get a hold of a heat sink from dell.

 

I don't know anything about the MB change you mention? The BIOS for each model will only support cards listed with the machine and the R4 is a different CPU family to the R3. That's not to say they can't be made to work it's just not going to be plug-and-play.

For the driver you can just ignore the warning. I have no idea why they did this but all I can think is that a new feature has been implemented in BF4? If I'm right then a newer driver is not going to do much since this is probably only valid for the latest cards. I tested BF4 on 314.22 and the last beta. All I saw was an improvement of a few FPS but horrendous stuttering! Temp was a few degrees lower though.

Maybe try this dell driver: Driver Details | Dell UK

Good luck with getting the correct heatsink, please let me know how it goes

 

Wow, you seem to be making this very complicated for yourself. So many grey areas. I would say something more constructive & positive, but it's all been said before - sorry!

EDIT: thought I'd add something constructive afterall. From your posts it sounds like you're approaching this in a kind of haphazard way. If it's a problem with your GPU, why don't you monitor the clocks & temperatures while gaming so that you know for certain how your GPU is behaving. From your posts it looks like you've monitored the temperatures, but not the frequencies at the same time. Once you have done this, and you know for sure that your GPU is throttling (dropping it's clocks), then you'd want to work out why. You'd compare the temperature at the point that the GPU throttled. If it was at the 'throttling temperature' when it throttled, then the problem is with inadequate cooling. If it throttled and the temperature was say 70 degC, then you know it wouldn't be a temperature issue for instance. So, then, if you've worked out it's a temperature problem, then you take the heatsink apart from the GPU, and repaste & check the pad layout, blow out dust from the heatsink, reassemble & retest. If temperatures are solved, then Bob's Your Uncle - you've sorted the problem. If temperatures are still a problem, then either your heatsink and/or mountings are damaged & need replacing OR you've just pasted it & padded it up badly (you need the correct X-bracket too). That's just the kind of approach that I think makes the most sense to take. But you seem to want to bring drivers, motherboards, Compal EC's, and the kitchen sink into the equation - leave that in the kitchen!

 

Sorry MickyD, it was Flingin who said the thing about R4 mobos (post nr #6).
However, Im surprised that the 675m is actually working "so well" with R3 mobo/cpu, so I don't think it would be a problem with R4 mobo, if that was what you meant.
Yes I did download that driver, going to try it out next, just need to get a hold of modded inf. Got errors on laptopdrivers2go when I tried to get the inf files for whatever driver I choose.

The driver was only for the stutter problem, which is solved now.
As I stated before, I've repasted and reapplied the pads 2-3 times, still same results.
I have done the monitoring by running GPUz at the same time as I've been running various games. The throttling occurs as the card drops all/some core clocks which state in one of my earlier posts, I think.
I've come to the conclusion that the 460m stock heat sink does not fit the 675m.

 

Ahh, I see it now. I believe that refers to the problem with SATAIII which dell couldn't get to work correctly (and is provided by the CPU chipset). Dell changed the spec of the R3 and removed all reference to SATAIII following a downgrade in the BIOS (R08 was the version I believe).

It should work perfectly well in an R3 since it's a 580m, but the change of name fails the BIOS 'authentication'. In theory you should be able to flash a 675m to a 580m and get optimus in an R3 to work as well, but I wouldn't recommend trying it!!

 

Ah yes, I remember you posting that you'll be buying a 675M/580M heatsink specifically. Well, that's fair enough, at least it will be padded up correctly & all you need to do is screw it down; however, I don't think there are differences between the 460M/560M/580M heatsinks, I think Dell offer different part numbers for those parts because they're just padded up differently. At least if you're buying something specifically marketed as a 580M heatsink, at least it's guaranteed to work. I'd get the correct X-bracket too while you're at it.

Or you can throw your 675M out of the window, keep your 460M heatsink & use it on a 670MX/675MX/680M! (That would have been better bang for buck than buying a 675 from the Fermi)

http://www.ebay.co.uk/itm/Dell-Alie...t=UK_Computing_Laptops_EH&hash=item4179404ffe

 

@0verseer you could try selling the 675m and get a 670MX instead, 580M heatsink (100W) V1K2G Genuine 0V1K2G Dell Alienware M17x R3 GPU Video Card Cooling Heatsink V1K2G | eBay

BIOS settings unlocked A12 M17x R3 BIOS A12 - unlocked & 'SATA fix' disable the integrated graphics, Second Advanced Menu>Video Configuration>integrated graphics disabled>set display to PEG>F10>reboot.

Download GeForce 331.65 - WHQL - Game Ready Driver

- Extract driver with 7-Zip, Browse Display.Driver folder replace/overwrite nvdmn.inf (unzip)
- Under ...\International\... folder delete the following folders:
Display.Optimus
Display.Update
GFExperience
GFExperience.NvStreamC
GFExperience.NVStreamSrv
LEDVisualizer
MS.NET
ShadowPlay

-Within the extracted Nvidia folder run setup.exe




- Windows 8 - disable driver signing.
Command Prompt (Admin) Win key + X Type:

Code:

bcdedit /set {current} testsigning yes

"The operation completed successfully">reboot>Install drivers.

Exit test mode, cmd (admin)

Code:

bcdedit /set {current} testsigning no

nvdmn.inf [Guide] Modding Nvidia's OEM INF Files

 

If you do decide to get the 670MX I think you'd be alright reusing the 460M/560M heatsink & X-bracket (like I have done). But, if you do get the 670MX or 675MX, then to notice a big performance difference from the 675M that you have, then you will have to be happy with flashing a modified VBIOS and overclocking by a large amount (they overclock well normally). That 680M that I linked you seemed like a pretty good deal, but you're probably not based in the UK. They're all good cards though, and they work in the M17xR3, albeit without optimus, but you're kind of stuck without Optimus on your 675M anyway, so no difference.

 

Thanks guys, my budget right now does not allow me to do any buying/selling though.

However, I've been in contact with Dell support a couple of times now and I finally decided to buy the heat sink kit. It costs 275SEK, which is cheaper than the ones on ebay.
According to Dell, the 460 heat sink does not fit the 675.

Im really excited now and I will report back as soon as I've installed it

 

let me know if it doesnt work, and u want to sell the card.. im interested

 

Great, that price is about the same as I was quoted, very reasonable. I saw them on ebay as well for around 3 times the price!

I am interested to see if it includes an x-bracket?

Enjoy

 

Yes, probably because the pad arrangement is different (they would have different part numbers for it - don't think the actual metal heatsinks are different), but I'd be surprised if it wasn't possible to take a 460M heatsink and re-pad it to work with a 580M. Good luck though with it once you get your 580M heatsink, hope it works for you.

 

the 460 is capable of cooling up to 75watt, the other model is used for 100watt

the shape is almost the same, its mostly the pad layout.. like robbo said.
thing is, the 100w heatsink has more copper in it

 

I read this also from others. Another thing is the x-bracket. Just because it fits nicely doesn't mean the post heights are the same. Especially between core technologies.

I believe the MXM standard specifies the component locations so they will always look very similar and be able to work with some pad modification. You wouldn't think an AMD heatsink would work well for an NV, but I've seen a lot of people saying it works fine (change the screw retainers and screws).

 

Actually, I think you might be right about the 100 watt version having a bit more copper on the heatsink. I just checked on ebay, and there are some versions of the M17xR3 heatsink on there with a little bit of copper off to one side, whereas on mine it's just 'normal' metal there. I've seen some people upgrade from 460M to 780M without issue on the stock 460M/560M heatsink, and temperatures were still good, so I wouldn't think there's a massive difference in cooling performance between them. I think padding it up & pasting it up correctly is going to be the biggest factor in temperatures in this instance.

This is the one on ebay that I saw with the copper in it:

http://www.ebay.co.uk/itm/NEW-GENUI...ther_Computing_Networking&hash=item2a31a38a78

Haha, and in post #23 of this thread, MickyD posted a pic of his with the extra copper in the heatsink too - hadn't noticed that!

 

Yes, important to have the correct X-bracket. But for people reading this the 560M/460M X-bracket (& heatsink) works well with the GK104 Kepler chip (670MX/675MX/780M). I've tested it with the 670MX only, but same chip used in 675MX & 780M (just cores disabled in lowlier chips), so I would think they would be the same height too.