Robbo

I think so, you'll need to make sure you get an NVidia X-bracket, don't use the AMD X-bracket, and I'm not sure if the screws you have will fit an NVidia X-bracket (so you'd have to find out about that or order the screws and X-bracket as a set). Then you rearrange the pads on the heatsink to match the 780M, as well as removing the black tape on the heatsink contact plate - the bit that contacts the GPU core. You remove any traces of AMD drivers before putting in the 780M, and change any BIOS settings as necessary. Think the install procedure has been listed in this thread.

(That's interesting, I've not noticed a difference in my OC stability with .09 driver, nor any drivers in my experience, although I know some people say the driver affects their max OC - maybe I'm not running borderline enough.)

You can't really know how much voltage he needs to run his overclock, it varies for each individual chip, so you can't really say that +50mv isn't enough for him to run those settings - only he himself can test what is stable & what isn't. Of course you want to use the lowest amount of voltage possible for the overclock you want to achieve.

Thanks yes, that makes sense. Are you going to try a single 780M in Optimus mode in your M18xR1? If a single 780M does work in that scenario, then it's less likely that a 980M will work in the M17xR3, because then the reason for the 980M not working wouldn't be that Optimus doesn't work in your M18xR1.

Yes, but I suppose you understand my point I made that if your 780M was to work in single GPU mode with Optimus in your M18xR1 & 980M didn't, then that would mean it would be less likely for a 980M to work in an M17xR3. Do you see what I'm getting at? (I appreciate that it's not 100% proof that won't work in the M17xR3 though). EDIT: I'm a bit confused about the bit in bold, you just tried it with a single 980M, so that statement confuses me.

Thanks for the testing Mr Fox, much appreciated! Ah right, so you think it's possibly a problem with the M18xR1 not being able to run Optimus with one card installed - I take it that goes for one 780M too (rather than one 980M that you tested)? If your M18xR1 does work in Optimus with one 780M, but not with one 980M, then that would not be a good omen for 980M in M17xR3.

Maybe try using a different overclocking program like NVidia Inspector, or check you have the latest version of Afterburner. - - - Updated - - - I thought I remembered that there was a reservoir of stock vBIOS files listed somewhere on Tech Inferno, but I couldn't find them right now. Maybe have a look around here on Tech Inferno to see if you can find it. Or hopefully someone with the same model of laptop & GPU can post the vBIOS up for you - you'll have to list your laptop model though.

[MENTION=119]Mr. Fox[/MENTION], thanks a lot, that would be great for M17xR3 owners to know. Although I do remember when I installed this 670MX all the guides said to disable 'IGD' in the BIOS and run it as 'PEG' - I just don't know whether that means that the R3 won't work in Optimus mode with a non-standard card. Do you have any thoughts on that particular point? If I become serious about upgrading to a 980M further down the line, then I'd reconfigure my BIOS to Optimus mode first to see if my 670MX would work in Optimus mode - to check.

Good point chap, I guess it would be ideal to have some kind of infrared temperature sensor you could point at different parts of your heatsink so you could get an estimate on VRM temperatures, etc. I made a point of giving my heatsink the 'finger touch test' (my own creation!) to see how hot the rest of the heatsink was - I ran it with the base off and in a way I could touch the heatsink in different places while running Heaven Benchmark on a loop, I could hold my finger on it for about 1 second, so wasn't insanely hot - if you imagine a metal pan at 100 degC (boiling water), you could barely touch that for an instant without pain! haha, that was my reference!

@Mr. Fox, given what you know about 980M compatibility in Alienware, would you expect the 980M to work in the M17xR3? I believe it came out in the same era as the M18xR1, so I have doubts that it would be compatible. Obviously I have Optimus capability in the M17xR3, but I was under the impression when I installed my current non-standard card that it wouldn't work installing a non-standard card in Optimus mode in the R3 (so I installed it in dedicated mode)? Just wondering what kind of upgrade possibilities I might have for my laptop in the future. (I would prefer installation in dedicated mode rather than Optimus, but just exploring all the options).

Zombiewarpig (on Geforce.com) looks like he's given you instructions that should work, which originally came from j95. Here's a guide to modding inf files for installing drivers on non-standard cards: http://forum.techinferno.com/nvidia/5249-%5Bguide%5D-modding-nvidia-oem-inf-files.html Have a look at this one too: http://forum.techinferno.com/nvidia/6307-%5Balienware%5D-aftermarket-upgrades-driver-support-modded-inf.html

I think you should choose an overvolt & overclock value that is stable in all situations - both in games & in benchmarks, and then use that value as your 24/7 overclock. You'll obviously want to achieve that figure with the minimum amount of voltage necessary. I think you agreed that you wouldn't go over +100mv for now, so you should just run a core overclock that is 100% stable both in benchmarks & games while at that +100mv. This is assuming your temperatures are below 80/85 degC. (Did you ask what my temperatures are, note I have a completely different card: 65-70degC depending on room temperature 20-26degC.)

Maybe stick at a limit of +100mv, and whatever the core is stable at with that voltage until you see other users pushing them further with no failures - assuming you don't want to be a pioneer!

True, but I forsee compatibility problems in the future with GPUs in older machines, particularly because Pascal seems like it's going to be quite different from previous cards - it's gonna have an ARM CPU included on the PCB for example, which makes it seem more 'different' to the GPUs that have come before it - perhaps more incompatible with old systems. Also, seems like there's a move away from MXM, so maybe in a couple of years they'll be no MXM, just soldered?

No, I don't think so, because it's only using 1.05V, and other Kepler cards like the 750M use 1.156V max (780M = 1V), and the temperatures are basically under 70 degC. Current consumption is only in the region of 90-100W anyway, which is also less than other MXM cards. It could probably handle another 50mv up to a used voltage of 1.1V and 1200Mhz, which would give me another close to 10% performance, but unfortunately the voltage control won't go higher. I don't think it's killing the card in my situation. (I've been running that overclock for all gaming I've been doing since about nearly 1.5yrs so far - there's not been any decreased stability since then.)

Good decision Steve, at the moment it looks like the M18xR1 is not compatible with Maxwell unfortunately. You'll be fine with 780M sli, especially if you (overvolt &) overclock with svl7's modified vBIOS, that's how I managed to nearly double my performance with the 670MX, here's the link to the forum where you can download the vBIOS files for Kepler: http://forum.techinferno.com/general-notebook-discussions/1847-nvidia-kepler-vbios-mods-overclocking-editions-modified-clocks-voltage-tweaks.html Yep, so 670MX is only 600Mhz core at stock, and I'm able to run at nearly double that for the core with added voltage (1124Mhz @1.05V, which is only 125mv over stock) - so that's why my 3DMark11 result is nearly double of a stock 670MX. No fan or heatsink mods, apart from laptop cooler which lowers temps by 7degC (65-70degC core temp depending on room temp) vs just using it on a desk - of course I had to move the pads around on the heatsink to match the chips on the new card, but I wouldn't call that a performance mod. Alienware has always had really good cooling, so this really helps, although their new thin & light rubbish that they have coming out now will probably have worse cooling! You should be able to run your 780M's for everyday gaming at 1006Mhz (up from 850Mhz stock) - that's an achievable overclock & temperatures should be reasonable, you'll probably need maybe just 1V to achieve that figure, all depends on the luck of the draw though.

No problem Player2!

Yep, that does seem the case then. I don't know what else to suggest bar all the reinstallation procedure I listed earlier (you know: DDU to uninstall, Dell Intel GPU driver, etc). (If after trying different things you could just do a complete reinstall of the OS or restore a working image from before, and think about resetting BIOS to default settings).

You could load up something like GPUz and run a game. Then afterwards view the Sensors Tab to see what the GPU frequencies & GPU usage was. If you're not getting any GPU usage on your NVidia card as seen in GPUz, then you know it's your integrated graphics being used instead (think you can select and view Intel GPU usage in GPUz too - on first main tab). That's just one way of double checking your understanding of what's going on, and which graphics processor is running the game.

A rubbish move from Alienware. Either a Clevo or I'll build a gaming desktop next - either way 120Hz screen with good GPU!

Yeah, I can understand why you're trying to bottom out the Optimus issues before buying a 980M, no point in buying a 980M if you can't get it to work!

You might want to try the uninstall & install procedure I listed anyway, unless you've already followed that process I listed.

780M sli will be pretty potent for you, even if it is not the latest tech - if you're gonna do that though I think it's only worth doing if you can find them really cheap, because a single 980M is not far off 780M sli performance. Just make sure that buying 780M sli is gonna be cheaper than a new 980M notebook. Actually, having said that, it's bound to be cheaper getting 780M sli, and the rest of your notebook is still good, so I think it's a good plan! My notebook is going well with my upgraded overvolted & overclocked (thanks svl7!) 670MX. I don't have those 2 games you mentioned, but I can give you some game benchmark scores of relatively recent games, some of which are quite taxing even by today's standards, note that I use a small 1600x900 screen so that helps: Tomb Raider (max settings apart from FXAA enabled): 67fps Batman Arkham Origins (max settings apart from x2 MSAA): 48fps Batman Arkham City (max settings apart from DX9 and 8CSAA): 65fps F1 2012 (max settings, 16CSAA): 83fps Metro Last Light (Original Vesion, Max Settings apart from Quality High, Blur Low, Tessellation Off): 52fps Bioshock Infinite (max settings apart from no DDOF): 78fps Titanfall: can run max settings if OK with 40-60fps, but I run reduced settings for 78fps (screen refresh rate). 3DMark11 GPU Score: 6740 Firestrike GPU Score: 4733 (6123 when run at my native 1600x900, same as 880M using 1920x1080) (I keep a record of the results I get when new drivers come out - don't want to use a driver that gives me less performance!) Yep, so it runs games OK at the moment (like an 880M with a 1920x1080 screen in terms of fps), so I'm not upgrading my GPU yet - in fact I'll either buy a new laptop or build a desktop in a year or two. (120Hz screen and good GPU for definite!)

Possibly there is the same option under "Global Settings" in the NVidia Control Panel where you can select "High Performance NVidia Processor" - there's definitely such an option under "Program Settings" in the same control panel, but click on the "Global Settings" instead to find out. - - - Updated - - - Try the following if you like: 1) Uninstall NVidia drivers. Then run DDU to remove any remnants of NVidia drivers. 2) Uninstall Intel Graphics drivers. 3) Install Intel Graphics driver - downloaded from Dell website specific for your model of laptop. 4) Install latest NVidia graphics driver (or recent NVidia driver of your choice) from NVidia website. 5) Retest for Optimus functionality. The idea of all of the above is that sometimes you need to use the intel graphics driver from the Dell website in order for it to work properly. It was like this with the M17xR3. It was important to install the Dell Intel Graphics Driver first.

Definitely not, it's a completely different card! One is Fermi, the other one is Kepler, different architectures even.