***M18x - NVIDIA SLI (Driver Questions, Performance, Problems) Discussion Thread***

Hi guys. I successfully installed the 980M SLI in my M18X-R2, the only detail is the weird high temperature on the second gpu like 70/80 C in idle, I dont try games in it yet. Is there any know problems in fan tables with 980M SLI installed in M18X-R2?

 

what is the current state of the union as far as the 980m goes? R1 isn't compatible and R2 works but only with older 345xx drivers?

 

My R2 is using the latest Nvidia drivers(355.82) right now.

Update: However upon testing in games I noticed the cards are still throttling bad even with Prema Vbios mod flashed cards. The Witcher 3 for example:

I think somewhere in these forums exists a Nvidia Inspector tutorial that set the card clocks to somewhat reduce the throttling in 980M SLI but I dont find it yet.

 

i think i would be very upset if i paid that much for the 980m just to have it throttle constantly

 

GeForce 353.06 PEG MOD - Throttling FIX
http://forum.techinferno.com/nvidia/6307-[alienware]-aftermarket-upgrades-driver-support-modded-inf-30.html#post135452

At least this driver really fixed the throttling, my 3DMark scores upped from 8600 to 13580 points all stock both cpu and gpus. Now all we need is a PEG mod in a "newer" driver but I'm not complaining =D

Update: Is seems that Nvidia Inspector is also needed, the P0 and P1 state clocks must be matched to prevent gpu working in low power mode while gaming.

 

where can i get the modded inf for m18x r2? currently running single gtx 780m

 

Just follow the link in my last post, the inf mod suits for all alienware notebooks and works for 6xx, 7xx, 8xx and 9xx gpu nvidia series, just go for the most recent one you can find.
Well, I made this myself, its for the latest nvidia driver 355.82 enjoy =)

 

Thank you so much Tulius. Is there a guide to mod the inf ? for wiindows 7 64 bit? I want to try next time. Thank you.

 

went to connect the SLI cable only to find out the 2nd card doesn't have an SLI connector....HAHAHAHAHA

 

is it possible to run the 980m on the R2 with Windows 7?

im considering just not getting another 780m and getting a 980m instead...

 

980m = No win 7

 

so with all the tweaks and the right drivers...no throttling or still some?

 

its very interesting i haven't had ANY throttling to 666mhz issues in the last few days but i don't remember changing anything

 

With just one gpu and an unlocked BIOS to set SG mode(Optimus), you can install Windows 7 with GTX 980M in M18X-R2.

 

@anarky321 - don't waste your money. The M18xR1 = zero Maxwell compatibility. I've posted about this several times, but apparently not everyone has seen the information.

The M18xR1 will not POST if you install a 980M. It will fail to boot with 8 beeps because pure using Maxwell in an Alienware with InsydeH20 BIOS requires pure UEFI mode (no Legacy Support) and GOP-only. The same is true for M17xR3. Neither machine has pure UEFI support, only partial support with VGA protocol for graphics. Only M17xR4, M18xR2, AW17 and AW18 can use 980M. The M15x is an exception because it does not have InsydeH20 BIOS. This is due to InsydeH20 BIOS limitations or conflicts with a few undesirable changes NVIDIA implemented with Maxwell.

 

i was talking about an R2 Mr. Fox

 

at what point do you know if you need the dual power adapter mod?

 

i have a question regarding the x-bracket...which way is it supposed to point? toward which memory banks? on mine i had it pointing at the memory chips in the middle of the card but i see online some have it set to cool the memory on the edge....also the stock x-bracket for the 560m had 2 sides to it to cool both memory banks...is this compatible with the 780 m?


i have both of the brackets and they both seem to fit on the 780m but i was afraid to use the 2-sided one

^^ if you look at that last picture, i have mine so that the OTHER memory bank is covered....is that wrong? does it matter?

 

The sink-bank heat plate goes on the longer outer edge just like your last photo shows. That is the correct placement, not across the short width of the GPU.

I use the double "L" shaped heat plate on my 780M cards. They do a nice job. It is intended for the primary (left GPU) to keep it cooler. Just use an insulator (electrical tape) on each corner over the screw posts to prevent the support plate (X bracket) from contacting and shorting any components on the PCB. (You can see yellow cellophane tape where I am speaking of insulating by looking at the photos below.)

They should look like this:

 

interesting..so to play it safe just use the double bracket on the primary and the single sided on the secondary..

how safe am i running 100+ core OC on a 780m? can i run more safely? i have to wait another week for another 780m with an SLI connector so that sucks

 

mr fox can you take a look at my listing

http://www.ebay.com/itm/NVIDIA-Gefo...-Dell-Alienware-NO-SLI-CONNECT-/191691209191?

and tell me if that card will require a vbios flash to work in an m18x? im the seller

while im on the subject, the MSI 780m with green PCB have different ram from the dell 780m? i remember reading that somewhere....

 

It will probably work just fine without flashing the vBIOS. GTX 780M has been pretty forgiving that way. I am guessing that is an MSI GPU since they have historically been a goof-ball GPU with no SLI connection. Flashing the vBIOS with a stock Dell or Clevo or unlocked SVL7 vBIOS should work if necessary.

GPUs can have Samsung, Hynix or Elpida and it's anyone's guess until you look. Dell 780M cards can have any of those three. They can be mixed and matched in SLI and it doesn't matter. Clevo MXM cards usually have Samsung vRAM. In most cases it makes no difference whatsoever, although the Samsung memory is typically a little better at extreme overclocking.

 

Mr. Fox, now that I upgraded my R2 with 980M SLI, my spare 780M(SVL7 vBios) gpus could be used in a Clevo system or its only compatible with Alienware machines?

 

has anyone been running more than 120 core OC on the 780m for a while? i don't see any reason to OC the memory because it doesn't seem to be the limiting point...how do you check?

 

I don't know of any reason they would not work perfectly. I have Clevo 780M SLI in my M18xR2 and I am running them with a modded Clevo vBIOS and I use that same modded vBIOS on the Dell 780M cards in the M18xR1 and Alienware 18.

I just about never overclock my 780M SLI for anything except for benching. I don't see any reason to overclock core or memory for playing most games or normal use, but with just one 780M it might be more useful. The 120 offset is 970MHz, which not extraordinarily aggressive. That should work fine with stock voltage and the temps should be decent. I would use Precision X and make a profile for that, then set the overclock only when you need it instead of running that all of the time.

 

snatched a 780 up for $310 bucks.....nice

 

is there any tool to measure gpu memory bottleneck? i don't like overclocking blindly, i know i have a GPU bottleneck but don't know if its a core or memory one

 

driver crashing at +150 core OC, even with a 50mhz reduction in memory clocks and 12mv voltage bump =(

+120 seems very stable

EDIT: +140 is stable @ stock voltage, not sure why 150 was crashing so bad

 

gpu driver crashing at +120 now...did i damage my card or something?

 

Increase volts a notch and test again. Gtx780m run fine 1006/1500 + 1.062v 24/7 without problem. You just have to try / fail a bit. Follow the temperature during testing. Which driver you use?

 

im using the trusty old 345.20 and i did +12mv its running fine, but it also crashed at +150 with +12mv i guess ill try upping it, my second 780m is in the mail and im doing the transplant to an R2 with a 3940xm ...fun fun

i use HWinfo/Afterburner OSD to monitor stuff

 

What do you use to increase the voltage on the GPU?
Edit: EVGA precision X16 and Hwinfo/Rtss is all you need.

 

im using inspector trying out 1000mhz 1.037v havent crashed yet

since using these settings my CPU gets stuck at 800mhz, i put laptop to sleep then it wakes up and its good to go again

edit: nvm now it gets stuck at 3.9ghz after wakeup....could my AC adapter be struggling to draw enough through my power strip?

EDIT2: can i send 2 power adapters to someone to make a "dual AC" out of them? with payment of course....

EDIT3: even with AC running straight off the wall same problem cpu locks into 800mhz while playing..i dont understand how oc'n the gpu could mess with the cpu multipliers

EDIT4: at +120 +12.5mv the problem does NOT occur....so weird...no crashes either

EDIT5: gpu got stuck in 325mhz today ...had to restart..what is going on

 

After some gaming with The Witcher 3, the GTX980M SLI clocks dropped down to hell to 135Mhz GPU core and 324Mhz mem, I'm using J95's Geforce 353.06 PEG MOD drivers, Prema's 1.1.1 980M vBIOS and created a saved clocked startup task matching the P0 - P1 clocks in Nvidia Inspector, upon rebooting I can game normally again but randomly the clocks drops to 135Mhz again. Is there any real fix for this?

 

put in my 2nd 780m, shows 2 cards in BIOS, flashed bios to same as 1st card, bios flashed OK, when booting into windows there is no trace of the 2nd card, not even in device manager...what is wrong

EDIT: ok wow driver reinstall fixed it....idk how that would have prevented it from displaying in Device Manager but ok

 

question, for SLI mode do i have to manually set the clocks on each GPU in Nvidia Inspector?

EDIT: OOOOK more problems, now SLI is working fine but CPU is downlocking during gameplay to as low as 2.7ghz even if FPS is below 60....wooow..was working perfect before 2nd card was added...is this a power issue?

EDIT2: removing all OC from GPUs does not change it, will try upping voltage on cpu...(why?)

EDIT3: changing FlexVID had no apparent effect, downclocking the cards either.....cpu won't fully upclock in-game...very bad..changed AC adapters same thing..obviously since this just started tonight the 2nd card is somehow responsible..but there is no way the power adapter is being maxed out so idk what else could be the problem

EDIT4: somehow my long power limit got zero'd out and this was responsible for this ^^

 

any way to get Witcher 3 SLI to work with 345.20?

 

the heat dissipation on the cpu is worsened significantly after adding 2nd gpu....was running ~85 before now ~95..will have to likely reduce OC on cpu

 

can the throttling temp for CPU in the modded bios really not be raised above 90? - assume this throttling limit is PER CORE?

i believe my 800mhz throttling issue may in fact be thermal CPU throttling, i had to lower OC to 4.3ghz since adding the 2nd GPU bc temps went up significantly

 

can someone remind me whats the issue with the new nvidia drivers above 345.20? do they throttle the 780m's?

 

Are you overclocking the cpu at all? If you are, you might be hitting the system power "limit" I think 3.9 or 4.Ghz might be the limited for games that are well threaded, I haven't played witcher 3 yet so I'm not sure if that is the case or not. The modded vbios also ups the voltage on the cards so going back to the stock vbios might also give your system a bit of extra power back.
The other time I have experienced this is with the 347.90 drivers, performance would be great to start with and then after a while its like the drivers would "expire" and start throttling to 135mhz reinstalling the drivers, sometimes up to 3 times in a row would restore performance. I haven't have a problem with j95's 350.12, haven't had time lately to try any of the newer drivers.

 

Thanks, I disabled all BIOS overclock functions of CPU(3920XM) and now I'll test some games. It seems the 980M consumes more power than 780M isn't it? Hope the stock CPU doesn't decrease the fps in games but if it solves the problem then would be a good trade off. I'll put the results later.

Update: Right in the nail, with CPU overclock disabled for good, the GPUs are really working better now, no more downclock to 135Mhz and I even dare to install the newer DEV nVidia drivers 356.04 and its working like a charm, no throttle problems in 3DMark or games.

 

I maybe wrong but I don't think this is normal, a few degrees might be expected with an extra gpu, but 10 is a big deal, something other then the gpu must have changed....?

I have had 3920xm over 100C with no throttle... I think sandy bridge did have a lower TDP though so the R1 might be different

From memory @Mr. Fox has complained about the newer drivers throttling 780m sli but that was while overclocked

 

No, most of them cause throttling even at stock clocks. I installed 355.98 this morning on the M18xR1 and M18xR2. Both throttle in 3DMark 11 at stock clocks... both drop to 405 core and 400 memory.

As the M18xR2 system BIOS was permanently destroyed by Windows 10, my display got bricked again. Had to flash the EDID in Linux to stop the 8 beeps no boot problem. As the M18xR1 has never been infected by Windows 10 cancer, it has no 8 beep problem, just throttling. So, back to 345.20 or 344.75 for both machines to stop the throttling.

Looks like the fools at NVIDIA are never going to fix this throttling problem. Sadly, it seems time to sell off all my Alienware stuff because nobody supports them any more (neither Dell nor NVIDIA) and nothing is ever going to be fixed.

 

this is the 2nd time im running SLI in this laptop and the CPU temps spiked last time i ran 2 cards as well, ive had to reduce the oc to 4.2 and possibly will have to go even lower because im still hitting the ceiling occasionally while gaming, this is on Liquid Ultra

EDIT: 4.2 works but hits ceiling occasionally, im putting together my R2 with 3940xm tomorrow so we'll see how that works in comparison

EDIT2: 4.2 scorched out, kept tripping throttle playing arma3, had to turn down to 4.1, im considering disabling 2nd gpu for arma3 altogether to be able to run a higher clock, its a very stressful game

EDIT3: well setting it to single gpu won't work because it just defaults to the secondary which is the one heating the CPU up

 

wooowww just read the article about the new full powered 980...165w TDP...wowzers

 

i dont think anyone responded to this but is there a working way to get Witcher 3 SLI working with 345.20 drivers?

 

well i successfully migrated to R2 today, windows had to install about 100 new drivers on startup but other than that seems to be working ok

i accidentally booted to my mechanical disk windows install at first and it was really a reality check on how far we've advanced in hard drives...my 850 SSD blows it away in speed

 

what inf file has to be modded for the newest nvidia drivers? 780m R2 (0550)

the newest ones are marked win10

i tried using the inf modded v356 from tech inferno but not working so far...do i need to further mod those? (modded dispi.inf)

to be more specific, do i still need to mod the nvdmi.inf with R2 hardware ID with the newest drivers? i tried that and they still fail to install

can someone using the new drivers just export the SLI profile for Witcher 3 to txt format and maybe email it to me?