GMP

This memory is rated at 1.35V / 1.5V out of the box. It runs at 1.35V on auto, at least that is the voltage I can see in various software, see screenshots a couple posts above. If I change it to 1.5V or 1.55v in bios, nothing changes, the volages does not change.

I just tried something else. I had a 3840qm before the upgrade to the 3940xm. I just replaced the cpu and put the 3840qm back to see if that one is able to handle the 2133MHz modules. It was the same as with the 3940xm, errors running memtest86, at frequencies 2133 and 1600. Didn't make a difference, although it seemed to generate a couple errors less at 1600. But errors nonetheless. I don't know if you're supposed to use an extreme cpu to be able to run 2133MHz RAM, I don't think so? So from this I can derive that the integrated memory controller in the cpu is not the problem as both the 3940xm and 3840qm have the errors? And still, running one stick gives no errors at all. Only from 2 and up. I truly think the RAM is fine. The 2133MHz modules are now back in the laptop, but running at 1600MHz. So far no BSOD's but I'm sure it won't take long before it starts happening again. I'm running out of ideas here :(

It automatically takes 2133MHz. I just now did some other test, setting the memory to 1600MHz. I checked and it actually applies this setting. After that I ran memtest86 again with the 1600MHz setting and it still reports errors when running more than 1 stick. I figured I'd test that as my samsung ram works just fine and is 1600MHz. Although the Kingston ram has quite a bit tighter timings running at that speed compared to my other ram, so I can't completely duplicate it to run at the samsung ram settings. But yeah, still errors at 1600MHz also.

Could be, any way to test this or know for sure? Don't really feel like buying another 3940xm. This laptop upgrade has cost me a pretty penny already lol. I'm gonna insert the Kingston modules again and try to set the memory to 1600MHz in bios in stead of 2133MHz (if I can) and see if I still get errors. If not, then something might indeed be wrong with imc. Or maybe it's a voltage related issue somehow. I really don't know.

Update: Unfortunately I had to put my 1600MHz RAM sticks back in the M18x. I ran the Kingston 2133MHz sticks for a little over a day, only had 2 BSOD while running memtest64 scanning for errors. But all of a sudden this afternoon I got one BSOD after the other out of nowhere, could hardly even boot to windows before crashing. I really don't know what to do now, these sticks cost me almost €300 and I can't use them for some reason. Testing the sticks seperately gives me no errors, so I'm quite sure the sticks are fine. It's when I run more than one of those sticks on top of the DIMM A slot that shit hits the fan. @ssj92Any idea what could be going on? :/

New highscore after changing to Kingston HyperX 2133MHz RAM, around 300 points increase on the CPU score: https://www.3dmark.com/3dm/80175707 This puts me in 4th place worldwide with my configuration (3940xm + gtx 1070), not bad.

I did some more testing, using memtest86. First I put everything in bios on stock: - Memory override support (voltage): disabled - Frequency: auto - XMP Profile: auto Then ran memtest86 with 1 Kingston stick in DIMM slot A (the lowest underneath the laptop). No errors. I tried all 4 sticks this way and no errors at all. I ran the test only for a couple of minutes each time but that was enough to get a basic idea. So from this I derive that the memory is just fine. Then I put 1 stick in only slot B, immediately a lot of errors. I did the same for slot C and D with only 1 stick, immediately a lot of errors. Don't know if it's normal to get errors if you're running 1 stick and don't put it in slot A, but ok. The fact is, I tried the same with my Samsung 1600MHz RAM, putting 1 stick in slot A, then B, then C and D. That did not give me any errors, so that feels strange that it only happens with the Kingston type memory. Then I put 2 Kingston sticks in slot A and B, immediately errors. And in the end I put all 4 in and immediately errors, although the errors came a lot slower, there were less. But still it yielded errors. I'm not sure what's going on here... That being said, with everything in bios on stock settings, windows seems to run quite fine (for now). I'll keep monitoring the next few days how many BSOD I get. But the memtest error situation, that I don't understand at all. EDIT: Also, when running 3dmark now, I'm getting really low scores all of a sudden. Framerate looks fine butten suddenly "glitches" and drops really low, then goes to higher framerate and then back down. Some sort of throttling. Weird things are happening because yesterday I was able to benchmark just fine with the new ram a couple of times. Just as a test I put back my old ram and ran a 3dmark run, but it's still the same. Now the GPU seems to throttle heavily at intervals, although I didn't change anything on that part. Even tried without OC, just stock GPU and CPU clocks and still throttles on the GPU almost as soon as the benchmark starts. And it's not a temperature issue. Anyway I'll sort that out, my main issue is with the RAM right now. EDIT2: Did a bios reset, ran 3dmark again and didn't throttle. Changed the memory again and the GPU throttled again. Apparently every time I change the RAM in the laptop I MUST reset the bios or the GPU throttles. Happy I found the solution. That still leaves me with the memory issue. I tried testing in windows with memtest64 and allocated 15360MB of memory to be tested, so a little over 1 stick of 8192MB. It runs fine without errors while testing the first 8192MB, but as soon as it goes above and thus the second memory stick kicks in, I get errors in memtest64, or even a BSOD sometimes along with it.

Ok the Kingston HyperX 2133MHz sticks have arrived but I'm really struggling with them. This is the exact RAM: https://www.kingston.com/datasheets/HX321LS11IB2_8.pdf I put memory voltage on 1.5V in bios, set frequency on 2133Mhz and then tried several things on the XMP profile. Tried auto but getting crashes while booting, or I get in windows and crashes after a while. BSOD, even orange screens of death. Once it even rebooted and automatically started running the windows memory test. Then I tried XMP profile 1, pretty much the same problem. I opened Thaiphoon Burner and checked what it should be able to run at: Then I went back in bios and used a custom XMP profile where I entered all the bottom row values (with cas 11). This seems to run a little more stable, but after a while I still get a BSOD and when I checked bios right after, the XMP profile was automatically reset to "default". When I run memtest86, I get errors already pretty soon. I'm not sure what's going on here. I'll post here some screenshots here of several software sources. Aida64 (different SPD than Thaiphoon Burner?): Thaiphoon Burner: Lots of things are Undefined like Manufacturer, Part Number, Jedec Dimm Label and everything XMP related. Is that normal? It also displays different SPD like I said compared to AIDA64. CPU-Z: 1.35V, shouldn't it read 1.5V? Not really sure what to do now. @ssj92, you are running the same RAM right? How are you running your 4x8GB sticks? What are your bios settings for the RAM? Voltage, frequency and XMP profile. My sticks are not new, they are second hand, all I could find. Not sure if they are faulty (which I doubt) or if it's because something is wrong in my settings. Any help is truly appreciated! My situation right now is that if I enter these settings from the SPD I see in Typhoon, that I can run "ok" but in the end it will also crash after while, where XMP profile in BIOS gets reset? Anything else I run crashes really fast. Is it necessary to do a CMOS reset for example for the new RAM? Should I test every single stick with memtest seperately? (In the bottom slot under the keyboard I presume) Because I don't think they are faulty, must be something else imo...

New Firestrike highscore. Huge thanks to @Expresstyle for helping with a new vbios! Card can now consume up to 125W power and runs almost 10°C cooler under full load! What a crappy stock vbios that was. With my configuration of 3940XM + single GTX 1070 I'm now in 10th place worldwide. And I'm sure I'll be able to do better with some tweaking and extra cooling mods. https://www.3dmark.com/fs/28453134

Yeah, from what I can tell, the card just doesn't seem to want to clock much higher than what it was designed for at the moment. Standard clocks: When running benches, the card hoovers around the boost clock of 1645MHz constantly, with max GPU power consumption around the design TDP of 115W. So that's all normal. But when I add 200MHz on the clock like this: Then it still doesn't want to go any higher than the clocks before while benching. It still hoovers mostly around the 1645MHz and TDP of max around 115W. My scores do increase marginally with a few 100 points though. But it doesn't seem to want to go above the 1645MHz much, besides a spike of a fraction of a second that I can see to like 1750MHz. But 99.9% of the time it stays around the 1600-1690MHz mark while benching, as far as I can tell. At +250MHz it crashes, always. Am I running into the 115W TDP limit and that's the reason it doesn't clock higher? Do you get higher clock speeds than 1645MHz? I also have the non g-sync version (Rev. 1.2), vbios on mine is 86.04.2F.00.3A Is that the same as your original vbios? Where did you get the 125W TDP bios version? I've been looking through some MSI forums, but couldn't really find anything much helpfull.

What are your GPU settings here when you check your firestrike results? I can't go any higher than this before 3D mark starts terminating in error. My average clock frequency is 1649MHz, but of course this is based on the card being clocked lower because of temperatures I guess (Nvidia boost 3.0). And max voltage is 1.06V, but that is by design and can't be altered afaik. I think I should be able to do better in graphics score somehow, maybe I should check the thermal pads and paste, but temps seem acceptable. Your graphics score is quite a bit higher, done at normal room temperature? What exact msi version do you have? The version 2.1 card? And g-sync (hardware ID 1BE1) or non gsync version (hardware ID 1BA1)? I'm hesitant to flash any other bios, wouldn't know which one anyway. And I'm not sure if the heatsink and fan (mine is 5.5CFM now) will be able to handle the heat dissipation. I've ordered a couple of other fans also, to see if I can do a custom job and get a >10CFM fan running.

My best so far for the M18x R2: https://www.3dmark.com/fs/28345537 I think there's still room for improvement when I get my new RAM and do my planned mods for better cooling.

Yes I saw that thread. And I do have one of those original alienware blanking plates. But I figured I'd use that one in my other M18x which is an R1. I just didn't want to make a custom one myself since I'm afraid of putting anything in the MXM slot that can short pins if it's even the slightest bit conductive, even with static. Even that alienware blanking plate scared me a little to put it in, maybe it's some special material which is not so static sensitive I don't know. Or it might be just HDPE, I have no idea. But like I said, I prefer keeping it for my R1. Maybe I'll change my mind who knows :) I'll decide when I have all the materials I ordered. Maybe I end up using the blanking plate in the end, we'll see. I'll post pictures when it's all done. Could take a month or so though before I can do the mod.

Holy crap. I didn't find it in my search results on ebay since there was no "DDR3" in the product title and they even made a mistake in the manufacturer product code so that gave me no result either. Glad you found it somehow. Thanks! Update: I just bought them. I hate customs fees (I live in the EU) but hey, that's life :)

I can't find them anywhere. 4GB modules I can find, but I need 8GB modules. I work on this laptop and I really need 32GB as I'm running applications that need a lot of RAM :/ Even when looking for the manufacturer code HX321LS11IB2/8 I can hardly find any hits.

Work how exactly if I may ask? I thought multiple GPU's didn't work with SG mode? Or does that only apply to SLI and Crossfire. You're thinking that somehow not only the GTX 1070 will be used but also the AMD card for extra performance? Kinda lost what you mean here or what I would get out of it...

I tried to find it online but nowhere to be found?

Hi all, hope you don't mind me asking some help without spending another couple of hours on this. I'm close to finishing my M18xR2 build, I've spend countless hours into researching the other stuff and all that is left is the RAM. Before I actually decided to bring this R2 up to date, I already bought 4x8GB sticks of RAM but it was basic RAM. More specifically Samsung RAM sticks, M471B1G73DB0-YK0 With pretty bad timings as you can see, 1600MHz 11-11-11-28-39: Can I do anything with this RAM if I want maximum performance? When it comes to overclocking this RAM I found this: Or would you guys advise me some other RAM at 2133MHz for example out of the box. I've seen people using Corsair and Kingston RAM. So couple of questions: 1) If you had to buy something, what would it be? 2) Am I better running lower frequency if I can get latency down that way? Does higher/lower frequency in combination with lower/higer latency make any difference between gaming and benchmarking? 3) When I buy them, can I just plug them in all at once, boot to BIOS and add the frequency and timings in a custom XMP profile and that should be it? 4) What happens if I can't boot anymore, I need to reset bios by clearing CMOS? Or is that something that won't happen. 5) If I buy some 2133MHz RAM with tight timings, I can try to tweak everything some more but the frequency? Since the R2 only supports RAM up to 2133MHz. Or can I try to overclock that too. 6) Do I really need software like Thaiphoon Burner to write stuff to the RAM sticks? Or can I just do it all from BIOS? Or is there any other software I should use or any guide I should follow? Again, sorry for the questions, but I don't feel like spending another 10 hours going through every single forum and NBRchive, hope you can forgive me :) Thanks!

I've looked a bit into the MXM standard and the pins. Always fun to learn new things :) Here are the pins we all know, front and back: Everything that is power related seems to be in the "big" pins E1-E4 and the couple pins after that. But the main power to the MXM GPU seems to be PWR_SRC E1-E4. There are 3 types of power going to the MXM card: the main power, 5V and 3V. All through the MXM connector coming from the motherboard I assume: Pin 6 PWR_GOOD is an output signal on the card. When all 3 power sources (main, 5V and 3V) are detected a signal is given to the motherboard from this pin. The motherboard in turn will send a signal back as an input on pin 8 PWR_EN, which in the end enables the power to be allowed on the MXM card. The Power Sequencing below also implies that first the main power, 5V and 3V must be detected on the pins. When that is detected, and output signal PWR_GOOD on pin 6 is sent to the motherboard, which in turn then sends an input signal back to the PWR_EN pin 8, after which the module internal power is applied. The "NOTE" under the graph also states it specifically that no voltages to any of the other more than 200 pins will be applied before the power is enabled to the MXM card. So if I'm able to mess with the PWR_GOOD signal, that means the card should be competely "dead" and powerless forever. So from that I derive that there are some options: 1) I cut the PWR_GOOD pin 6 so that the signal is never given to the motherboard, which otherwise in turn would send a signal back to actually apply the power on the module. 2) I cut any of the power pins, main power are big pins so that would be the easiest. That way not all 3 power sources (main, 5V and 3V) are present and the MXM card never sends any signal from the PWR_GOOD pin 6. 3) I cut the PWR_EN pin 8, so that even if the power is present to be applied, it can never get an input signal to actually apply the power to the MXM card. Bottom line, I think it's easiest to cut the main power pins E1-E4 with a dremel or something, as the other pins are way to small to do any work on them. That should do the trick and never allow the card to be powered on, and thus for sure never produce any heat: Or I just cut the whole piece containing E1-E4 and the pins 1 to 8 after that. That way I take away anything that is power related. The rest of the pins can stay on, which will allow me to still plug it into the MXM connector and get a snug fit for the GPU card so that I can mount the heatsink on it on which I can solder the heatpipes coming from the CPU. I am curious though if the BIOS will still detect the card. I will test it out when I get the heat pipes and low temperature solder which is on the way. If I'm saying things that don't make sense, please tell me lol :)

That was actually the first thing I thought about, cutting the pin edge off. But the card being snug in the MXM slot gives it the best stability. I want there to be the least amount of movement on the card because of the heatpipes mounted on it to the CPU heatsink. It's still some distance and the heatpipes moving could tilt the CPU heatsink and cause imbalanced temperatures from less than optimal contact. At least I consider that a possible risk that I'd prefer not to have to deal with.

Since I'm now running a GTX 1070 in my M18xR2, I decided what to do with the empty secondary GPU slot. I decided not to use it for a MXM to NVMe Adapter, since I think a normal SSD gives all the speed I need. What I did decide was to use the secondary GPU slot for extra cooling for the CPU. I'm gonna use the secondary GPU heatsink and fan, and solder some heatpipes from the triple piple CPU heatsink to the secondary GPU heatsink. That will allow me to keep the 3940XM even cooler and push it even more. I checked and it can certainly be done. I can easily add an extra 3 heatpipes. When it's done, I will post some pictures. I ordered some low temperature solder (melts at 138°C), that's low enough to keep the heatpipes from bursting/exploding while soldering. The second option is a thermal conductive epoxy, but I'm having trouble finding a good one with good thermal conductivity, can be cured at room temperature and has a glass transition temperature that is high enough (this is the temperature at which the epoxy starts to become liquid again). That being said, that is not my problem. The thing is I prefer not to make any custom plate that fits in the secondary GPU slot on which I can mount the GPU heat sink. I have an old AMD GPU from before that fits and which is perfect to mount the 100W AMD heatsink on. My question is, will this interfere with the GTX 1070. Will I have boot problems with the 2 different GPU's? Or is there a way to completely disable the secondary GPU slot from the unlocked bios so that the AMD card doesn't get detected and powered? That way it can act like a dummy plate on which I can mount the heatsink. I checked but can't really see an option that fits disabling it in bios, unless I'm missing it. Or can I just boot in windows and disable the AMD GPU from there in the hardware? Does that mean the GPU won't be powered in that case as soon as windows loads the next time? Not really sure what the best option is here. Of course I don't want any heat originating from that dummy GPU. If someone has any experience with this... Thanks!

Do you know what the best method is to determine if a refresh rate is stable or not for everyday use?

Good to know, thanks a lot!

Thanks for that. I just edited my previous post as I figured it out myself in meantime. But it works so I'm very happy :)

Yeah I figured out in meantime that it's not possible with Ivy Bridge CPU's. I read something on NBRchive (also form @ssj92btw) that exporting the EDID bin file from CRU and using linux to flash it to the display. I was looking into that, but your option sounds like a lot less hassle if I can somehow make an windows executable with reboot. Can't wait to hear from him how it's done. Fingers crossed :) EDIT: I actually found how to do it! In Custom Resolution Utility, change/add everything you want. Then click the "export" button and save it as an executable. Afterwards, close CRU and run the executable. Just do Install EDID. It will tell you to reboot. I rebooted but got a black screen when reaching windows. I rebooted again and got to windows without a problem. And eureka, all higher refresh rates that I set were available. And they work :)