Robbo

You'll have to stack up the pads to the correct height for whatever chips you're covering - so that the heatsink sits level. So it can't be answered whether you need to use 0.5 or 1mm pads. My 560M heatsink only uses 0.5mm pads in one place - the place where they were before when I had the 560M installed - I'm guessing you can leave those ones there based on the fact that fits with my 670MX. I used 1mm pads everywhere else & stacked them up to make them the right height where needed. I re-used the VRAM pads from the 560M, and just shifted their position slightly to better cover the VRAM chips of the 670MX. Your 880M won't be much different (same chip layout as the 670MX apart from more VRAM chips).

Yep, I noticed there were 2 versions of the 765M vBIOS on here: one for the Alienware 17, and one for the Alienware 14. Historically, it looks like Dell, MSI, and Clevo vBIOS are often interchangeable based on svl7's first (or second) post in this thread where he says for instance that the 880M vBIOS is suitable for Dell/MSI/Clevo. There is a note about P1xxHM Clevo users needing to be careful about using a certain 680M vBIOS version, but there's nothing there about your Clevo model. I would google you model number along with 'modified vBIOS' to see if there are any issues that other users have had by switching to a different vBIOS, otherwise I would have thought either of the Alienware 765M vBIOS would work on your laptop.

Yeah, that's not a massive overclock, but it's still 100Mhz faster than the stock boost of 1126Mhz, and those 980M's are already clocked pretty high at stock. I don't think many people are gaming at much more than 1300 odd Mhz anway, so you're probably only 100Mhz off what some people are gaming at. 100Mhz increase in clock speed up from your 1230Mhz clock is only a very small increase in performance anyway (max 8% performance improvement).

There you go, you're getting to the bottom of it!

Have you only just now flashed & tested svl7's vBIOS, or have you been using it a few weeks? I've read that some people find that stock clocks on svl7's 980M vBIOS can be unstable, this is because it runs the card at max boost clocks all the time, and some cards can't sustain that at the stock voltage of the svl7 vBIOS. You might have to just increase the voltage by a notch or two to get it stable (without adding an overclock). Did you compare your measured voltages (using GPUz for example) between your stock vBIOS and svl7's vBIOS, could be that your stock vBIOS uses a bit more voltage when the card is at max boost?

Well I didn't realise your GPU was getting up to 93 degC on the core, that's a big problem. Although I think you also showed throttling happening when you were at a lower core temp as well. I don't know, either way you need to sort out the ridiculous 93 degC. Need to repaste for sure, and sounds like thermal pad placement is wrong if you're getting 93 degC (not sitting flat on the GPU core). I think you need to redo the whole thing. I don't think I can suggest anything else apart from all the things I've already advised in all my previous posts anyway - good luck, I hope you manage to solve it.

I also think that one scratch won't make a difference, but if you have a lot of scratches on the chip then you get less efficient heat transfer, that's why some heatsinks have mirror finishes for better heat transfer. Although from what people have been saying on here, if you remove IC Diamond carefully then it doesn't create any scratches anyway. I'd buy Gelid GC Extreme though next time as on Toms Hardware it performed the best when they did their roundup of different pastes (well Liquid Ultra was better, but I wouldn't use that due to the risk of it being conductive). In the meantime I'm getting by with an ancient tube of Arctic Silver 5, it's lasting particularly well on the CPU - haven't repasted it within the last year and temperatures are just as good.

I think you should leave the big thermal pad on the motherboard, it's there to cover the VRAM on the bottom of the card - check that it's positioned to contact the VRAM chips ok, should be - I left it there when I upgraded to 670MX. Yes, that's correct the chip you circled in red is the problem chip, you can see from my pic I showed you yesterday how I covered that chip. You could try something similar to what I did before buying a new heatsink. If you do decide to get a new heatsink try to get a Dell M17xR4 heatsink that's been used on a 680M, because it's more likely that the red circled chip will be covered - same layout between 880M and 680M.

That seems strange that you have that heatsink part number for your M17xR3, as oj9h7x is not listed as a M17xR3 heatsink in the lists I've been able to find. In fact, 0j9h7x only appears as an M17xR4 100W heatsink on ebay (there may be other occurrences on the internet, but a quick search only yielded the following ebay listing): Alienware M17x R4 Primary 100W Heatsink | eBay I have heard I think that the M17xR4 heatsinks are compatible with the R3 though, I think it's because the chassis of the R3 & R4 are the same, just the motherboard is different. If you have that 100W heatsink oj9h7x, then I think you're best off just trying to cover that chip as best you can. The heatsink for the 680M for the R4 is the best most compatible heatsink I think you could get for your 880M, and it looks like that's the heatsink you have based on that ebay link I showed you.

Are you sure the V1K2G heatsink covers that chip we're talking about? Yep, about the HD audio, follow j95's instructions in Post #92 at the following thread: Alienware M17x R3 460M to 780M - Questions/Concerns | Page 10 | NotebookReview

I just tried to cover that chip as best I could, and that was the only way I could work out. I was wanting to fix a strip of metal across it but the clearance between the heatsink and that chip is less than 1mm, so there wasn't enough room to use metal strips & thermal pads, just had to stick with the thermal pads. If you can think of a better way to do it, then do that by all means, because my job is a bit of a botch job, but it does seem to take some heat away from that chip because the top of the thermal pad stack does get hot to the touch during use, so it is drawing heat away from that chip. (I've been running it like that with a heavy overclock for nearly 2 yrs now - no sign of failing card or instability)

Perhaps, he might be having problems with this one chip that is not covered by these heatsinks. I had to use layers of thermal padding to 'extend' the heatsink to cover one of the chips. It's possible that the 880M places more of a demand on that chip than say my overclocked 670MX, and considering he probably hasn't even covered that chip at all, then this might be causing issues. Here's the pic of the thermal pads I layered up to cover the 'problem chip' (layout of 880M is the same), I have the same heatsink as him (560M): (I hadn't remembered to consider this as a possible reason for his throttling problems when I recommended he buy the 100W heatsink) @Kedawg, see if you can cover that chip that's in my photo - that might help your issue.

Cool, well I've given you the part numbers for the 100W heatsink, I think you should get that heatsink & make sure it's padded up properly to accurately fit your 880M - that's the crucial part (the padding) because it allows the chips to be covered, but also determines whether the heatsink will sit flush & flat on the GPU core - if the padding is uneven (too thick in places) it won't be able to sit flat on the GPU core.

So 327.23 fixed your power throttling for the most part (you could experiment using some of the later drivers too, to see which ones don't throttle but yet give you good game compatibility with the latest games). Your 3DMar11 score seems a bit low still though because I get the same score as you with my 670MX - might be the thermal throttling you mentioned. What was the Graphics Score you got in 3DMark11 (not the P score)? You should get in the region of 8500 for the Graphics Score. What card did you upgrade from - that will determine whether you have the 75W or 100W heatsink. The only difference between the two is that there is a bit more copper on the 100W heatsink, and less aluminium - they're both 3 pipe heatsinks. I'm using the 75W heatsink on my overclocked GPU, and it's fine, but your 880M will be drawing more power. I'd make sure you've got 580M/675M heatsink, part numbers found in the following thread, and you'll need to make sure you're using an NVidia X-bracket, as the post heights are different to the AMD X-brackets which will affect the amount of pressure received on your GPU core from the heatsink: *OFFICIAL* M17x R3 Owner's Lounge Thread - Part 3 | NotebookReview The heatsink part number I think you want is V1K2G

the modded inf's include all the latest GPUs so your machine should be covered. See if you can find some guides on how to modify your own inf - there's some here on Tech Inferno by j95, and there's probably some on notebookreview forums. I taught myself how to mod them by just knowing what my strings are - seen in Device Manager, and then working out where to paste them into the inf. EDIT: You have the R3, the same as me, I modify the nvdmi.inf file and replace the M17xR4 680M (11A0.0551) strings with M17xR3 670MX (11A1.0490) strings, you could do something similar, the section numbers will be different though from driver to driver, I don't change them when replacing the strings: %NVIDIA_DEV.11A0.0551.1028%= Section345, PCI\VEN_10DE&DEV_11A0&SUBSYS_05511028 replacewith: %NVIDIA_DEV.11A1.0490.1028%= Section345, PCI\VEN_10DE&DEV_11A1&SUBSYS_04901028 %NVIDIA_DEV.11A0.0551.1028%= Section346, PCI\VEN_10DE&DEV_11A0&SUBSYS_05511028 replacewith: %NVIDIA_DEV.11A1.0490.1028%= Section346, PCI\VEN_10DE&DEV_11A1&SUBSYS_04901028 %NVIDIA_DEV.11A0.0551.1028%= Section347, PCI\VEN_10DE&DEV_11A0&SUBSYS_05511028 replacewith: %NVIDIA_DEV.11A1.0490.1028%= Section347, PCI\VEN_10DE&DEV_11A1&SUBSYS_04901028 NVIDIA_DEV.11A0.0551.1028= "NVIDIA GeForce GTX 680M " replacewith: NVIDIA_DEV.11A1.0490.1028= "NVIDIA GeForce GTX 670MX "

You could get a modified inf from this site that I've used in the past (but this very second the site seems to be down, was working yesterday): http://www.laptopvideo2go.com/ Of you can mod the inf yourself by adding the strings of your card and system to the inf file. I do that, and I just find a similar system in the inf & then replace my strings into it. Laptopvideo2go is a simpler solution though.

Driver 327.23??

Try 327.23 driver too, then if that's ok, then try one of the later drivers (ideally you wanna use the later drivers for compatibility with games, but for testing purposes useful to try 327.23 driver if later ones create throttling).

Good stuff. That's very strange though, Perfcap PWR means that it thinks it's reaching the TDP limits of the vBIOS, but that shouldn't happen during 3DMark11 and shouldn't make the clocks throttle dramatically while at 100% load as is shown in your GPUz screenshot. Some of the latest drivers have proven to have weird power limitations & throttling issues with some systems, I suggest you try an earlier driver, try using DDU, and then install 327.23 drivers - it seems that those don't include any limitations based on this thread I found: https://forums.geforce.com/default/topic/754922/nvidia-crippling-the-gtx-880ms-with-there-new-drivers-look-for-a-nvidia-response-/ I believe you should be able to use some of the later drivers with no issue too, but I have read that some of the latest drivers have been causing strange throttling problems, but I'd try 327.23 driver after a DDU clean to just rule out that it's not the drivers causing your issue. Yep, so install that driver & do the same testing, see what happens.

I don't think you listened to my post, you have temperature problems (amoungst other issues) with your card. GPUz is showing about 70 degC when idle in some of your previous screenshots - that's too high. Investigate repadding your heatsink like I mentioned to you in my last post. Secondly, don't use Furmark to validate what's happening with your cards - it's a power virus & NVidia automatically throttle the cards when they detect it running, so it's just not a useful test. Thirdly, your card is behaving weirdly with throttling clocks, I suggest you revert to the stock vBIOS you were using then retest things like 3DMark11, etc, while monitoring your clocks & temperatures with GPUz. Actually, reverting to the stock vBIOS you had is what I would do first, then I'd test 3DMark11 while using GPUz, and depending on the results and what you saw in GPUz I would then repad the card. Klem gave you some good advice about how to install the NVidia drivers properly using DDU. EDIT: did your card ever work properly when you were using the stock vBIOS? If it didn't then there's probably other reasons for why your card is not working properly, like too much heat, but you would need to show that using GPUz when at load.

It sounds like using a different vBIOS is not your answer, I think you need to sort out the heat problem. The heatsink needs to be padded up accurately so that the chips are covered, but the pads need to be the right height to allow the heatsink to sit flat on the GPU core. I think you should try re-padding your heatsink & repasting the GPU core. (Obviously check you haven't got dust blocking the heatsink - use a can of compressed air to blow out the dust). Also use GPUz to check what the temperatures and clocks are doing during gaming or running benchmarks so that you can uderstand what's going on.

From reading around on threads like these there's a few things you need to do in order for a 980M to work on the A17 when you've got a 120Hz screen. You need Windows 8, Windows 7 won't work with it with the 120Hz screen on the A17. You need the BIOS set for a 'pure' UEFI environment, no legacy boot option chosen. And I think the disk needs to be GPT partitioned or something like that - bit hazy on my memory of it. This is my recollection, but read around in the forums here & on NBR to confirm what I've said.

Reminds me a little of the 580M scenario where the VRMs (or was it MOSFETS?!) were getting extremely hot & then killing the cards - hope it's not a card killer effect!

True, I see that now that you point it out, not sure what that is. Either way, if I had your cards I would take the approach I outlined in my previous post. If you want to have 2 cards for sli maybe you can still do the testing I mentioned & then send back the card that doesn't overclock as well and request a card with the identical design of the card you're keeping.

I've heard conflicting info on this topic - regarding the number or MOSFETS that are along the top of the card. I think there was a user @Khenglish that was saying that the cards with 2 MOSFETS had MOSFETS that were more 'advanced' and efficient somehow, therefore requiring only 2 in contrast to the previous 4 that were common. If you're willing to test them to their limits then you might be able to answer that question for yourself - by flashing one of those modified vBIOS from Prema or svl7, then overclock and see how stable & high they overclock. Sure, there's a silicon lottery between the two cards anyway, but I'd probably just keep the one that overclocks the best.