Just got my 880M twins!

Not trying to derail the thread here or anything, but anyone happen to know what size thermal pads to use on the AW18, since we're on the discussion of cooling?
I remember it being said that the chips on the GPU had thermal pads on the clevo units but not the AW units.

 

0.5mm and 1mm if I remember correctly.

 

I answered this in another thread already. These measurements are from brand new Alienware heat sinks with the thermal pads that come pre-applied from the factory.

 

That's all well and good but that doesn't mean anything as 3DMark is throttled as soon as it starts by the vbios. You would have to flash your vbios and put your fans on forced max then run an actual game that can stress them (Bioshock Infinite does it within about 20 minutes for me) and record your maximum temperatures. I'm willing to bet you'll still pass 90C unless Clevo has fixed their heatsink design for that machine versus the rest of them.

I'm considering buying a slave GPU heatsink from R&J Tech to see if it cools any better than the one I have now. XOTIC told me 95 for a converter so I'm pissed and debating whether I'll bother.

 

3Dmark dont throttle anymore on my Clevo. 5300 points in Firestrike Extreme on 4.3Ghz and Stock GPU clock.

 

So long as the dimentions are all shrunk at the same rate (unless it's unbalanced to begin with) it's always worth trying getting the thinnest pad at 0.5mm.

 

Yes, thinner is better as long as there is no contact interference that might tip or tilt the heat sink and cause it to not sit flush on the die or other components. You can check this visually with no thermal pads in place. If there is no such condition without any thermal pads you can use as thin a pad as you wish as long as you do it uniformly (not mixed pad thickness) as suggested. For those that don't feel confident about verifying that simply using the stock thermal pad thickness will definitely work fine.

 

I don't buy it. Use a tool to monitor your clocks. Your score is going to be significantly higher on that system because of your overclock and much faster CPU but that doesn't mean your graphics score is going to match the rest of the system. Putting cards with a vbios that throttles in a stronger system isn't magically going to remove the throttle that is there. The same load is going to cause the same throttle no matter what system you put the cards in.

I use HWINFO64 with RTSS to get an OSD going... the second I start 3DMark, it throttles the cards to 954MHz. It has nothing to do with heat at that point. I don't have a heat throttle with the stock vbios, just the modded one.

This extreme run just now had a maximum temp of 70C and the fan was on auto, it was no higher than medium speed the whole time yet I watched as both cards were throttled. Primary to 954, secondary went from 954 down to 889 at its lowest with both cards starting off the second benchmark at 993MHz before dropping again.

NVIDIA GeForce GTX 880M video card benchmark result - Intel(R) Core(TM) i7-4940MX CPU @ 3.10GHz,Notebook P377SM-A

5423 graphics... Because my machine is significantly weaker than yours, my total was 4866

 

Im sorry, I ment that the laptop dont throttle as hard as it did when I bearly got 2800 points in FireStrike Extreme.

NVIDIA GeForce GTX 880M video card benchmark result - Intel Core i7-4960X,Notebook P570WM

This is my score now.

NVIDIA GeForce GTX 880M video card benchmark result - Intel Core i7-4960X,Notebook P570WM

 

I just buy the recommended Fujipoly themal pad brand of 0.5mm thickness to install on my new 780M. Seeing that photo Fox posted here showing the themal pad layout in the Nvidia heatsinc there's a side with 3mm of pads.... must I pile 6 pieces of 0.5mm thermal pads to fill this gap or can I put less with 0.5mm pads? Thanks.

 

Hmmm... Its interesting that your result reported the boost clock.

Thanks for sharing though, I'm no longer thinking that having the P570WM would have been a huge benefit over the P377SM-A. Your score even offset by the much more powerful CPU is not that different from my own, at least not enough to warrant how much it would cost me to get the exchange done (15% restocking fee, shipping, time lost, etc).

 

Well, if all you care about is graphics performance it would be a major mistake to buy a P570WM. On the basis of graphics performance your system is already on a level playing field with it because both systems have 880M SLI. I take it you did not notice the relationship of physics score is treated as being irrelevant even though it absolutely obliterates your P377SM-A. The problem here is Fire Strike sucks as a benchmark because it does a HORRIBLE job of reflecting overall system performance. This is very misleading if you look only at the overall score. Fire Strike basically ignores the CPU and doesn't count physics performance when calculating an overall score. If that was a 3DMark11 or Vantage run you would see a massive difference in benchmark results.

 

I didn't ignore it at all. I know the CPU blows my score away. It should - 4.3GHz on 6 cores vs 3.7 or 3.8GHz on 4 - but my graphics score and combined score both beat his by a few points which tells me that game performance would be pretty much the same between the two machines. His machine would likely be better with the modded vbios than mine though. There's still part of me that is thinking of getting a 9570 barebones from RJTech and putting my 880s into it but it would run me another 2.5k when everything is all said and done (new hard drives and processor) and I don't know if I want to give up my 120hz screen.

Either way the point remains that his scores have made me much less annoyed about having a 9377.

On a side note I found out what is freezing my computer. The nVidia 337.88 driver actually crashed on me earlier and it froze up the machine for about 2 minutes before the driver crashed. It was just sitting on the desktop. Well it froze again later and I waited again and sure enough, driver had crashed. So the freezing mystery is solved :|

 

Well thats good it's not something more serious

 

Does it freeze with your overclocks enabled? If it does, then your drivers, OS or system simply doesn't like your overclocks. It happened to me before on this machine, I'd crash in like Orcs Must Die 2 or CoD BO2 but I could play BF4 all day erryday with the clocks straight up. Odd as all hell.

 

I had the same problem with Titanfall, had to back down from 1124Mhz to 1097Mhz to get rock solid stability - but just on that game. My theory is that Titanfall GPU usage fluctuates heavily between 100% and 50% due to fps locked to my refresh rate - all other games my GPU usage is pretty much 100%. You mentioned Orcs Must Die and COD where your overclock was a little unstable, I would hazard a guess that those aren't that graphically demanding & you might be seeing high fluctuations in GPU usage in those titles, whereas BF4 is probably 100% GPU usage the whole time - maybe you might have instability for the same reason that I think I have in Titanfall - overclocks don't seem to like wild fluctuations in GPU usage/load.

 

I don't have anything overclocked at the moment, CPU or GPU because every time it would freeze, my overclock gets reset anyway. The heat is too high with the modded vbios so I am on stock GPU clocks.

I wish it was that simple. I have never had so much trouble pinning down a freeze before and I've been building machines since I was 10 and I'll be 30 this year. It's totally random when it happens - it froze while I was typing a document in WordPad of all things the other day and Bioshock Infinite freezes so much I stopped playing it. Watch Dogs freezes but has a lot more checkpoints so it's not a big deal.

It's annoying the hell out of me though. I really don't want to put Windows 7 on here, especially since I have a RAID configuration going with 8.1 (I have never had RAID before and I have no idea what it would entail getting the two to cooperate and if Windows 7 ends up not freezing at all I'm going to want to stick to it and dump 8.1) but it seems like it might be the only option I have right now.

 

Memtest86+ on the RAM? Might be a system RAM instability problem.

 

I'll try that. Right now I just got annoyed with trying to get Windows 7 to install in UEFI mode after Windows 8.1 so I just cleared my RAID array and made a new one, installing windows 7 first then Windows 8.1. I'm very thankful I opted for two storage drives in addition to my msata drives now.

I can tell you it's highly unlikely it was my RAM. I did the aida64 stress test which is far more RAM stressing than Memtest and it ran for hours with no issue. When I say the freeze is random, I mean it. It can be doing nothing and freeze or be stressed to the max with a game and freeze.

 

That kind of behavior is sometimes an indication of not using ideal CPU power settings. Haswell is unreasonably finicky and good settings for one system may not be worth a darn on an otherwise identical system. I think this inconsistency is causing problems for the OEMs because using "one size fits all" BIOS defaults for CPU power settings isn't working as well as it used to. To further complicate matters, some people are trying to undervolt to mitigate Haswell thermal issues. Taking this too far is probably causing even greater instability for some folks.

Haswell-powered Alienware laptops also have issues with system freezing, but this is almost always an indication that the Core Voltage, Core Current Limit and/or Processor Current Limit are not set for the correct values. It usually means one or more of those values are too low and you need to either increase them or, alternatively, reduce the core clock ratio so less is required. Having voltage and power settings too high is better than having those settings too low as far as stability is concerned. Unfortunately, system stability comes with a price tag of more challenging thermal management where Haswell is concerned. Nice job, Intel.

 

Thanks for the tip. I'll bump the BIOS setting up from TDP Nominal (57W) to TDP high (67W) which will bump everything up higher.

 

That may help. It may be easier to use XTU or ThrottleStop. I would start with bumping core voltage in 10mV or 20mV increments and leaving TDP alone initially. Or, lower the 4th core multiplier by one and see if that helps. If you experience clock speed drops under load, increase the TDP, Core Current Limit and Processor Current Limit. The core voltage being too low is the greater contributor to freezing. The other limits being too low impacts performance more than stability.

 

It would be easier to just flip the switch in the BIOS. Doing so bumps up the settings on the processor to the ones XTU says are default for my chip, including bumping the amperage (I can't remember the values, in the midst of full system setup - by the way I've got to say that Clevo has their crap together when it comes to their driver disks. Everything is listed in the order it needs to be installed. Man I don't miss those days of figuring out the list with my M17xR1 to not lose Bluetooth!) and it sticks even if the system crashes unlike everything else.

 

My m18x R2 is alive with 780M's and 3940XM!!!!!!!!!!!!!!!!!!!!!!

 

Yay! That's awesome. Congratulations.

 

Sweet brother Astonished!

 

Nice work bro. You got yourself a Stallion

 

Did we ever figure out what the deal is with those 880M cards? Do they act all goofy in the Clevo as well? Is the vendor going to refund or exchange them for 780M cards, or did they leave you high and dry?

 

Based on upgrademonkey's reputation I would certainly hope they took care of the situation. Although 3 bad Clevo 880M cards in a row really does raise an eyebrow, and makes one wonder whether it's due to a bad batch of cards or some sort of more serious quality control problem...

 

Arotished has been using them in his P570WM without issue.

Sent from my HTC One_M8 using Tapatalk

 

This is really freakin strange.

 

Yeah he pulled them from the M18x and put them in the P570WM then took those 780M cards and put them in the M18x.

Maybe the M18x doesn't fully support the 880M? It's the only thing I can think of since they work fine in the Clevo - Clevo machines are much more finicky than Dell machines when it comes to what video cards they will work with.

Sent from my HTC One_M8 using Tapatalk

 

Upgrademonkey have provided good support the past 2 months. Just a correction, I didnt get three bad 880M card from Upgrademonkey but only two. My third card which I got from Upgrademoney which replaced my first card that blew was the same revision as the card I bought from Eurocome. My second bad 880M card (first revision card) is on its way to china as we speak and Im awaiting what Clevo has to say when they receive it for test.

Conclution was: My first two card from Upgrademoney was a different revision (Older?) and none of them worked in my M18x R2. First card blew a resistor and second card just made the hole system unstable when I had it installed, blue screen and unable to install Windows Updates and Framework 4.5. The 880M card I bought from Eurocome (to replace the first 880M that blew) was a different version than the two first 880M's I got from Upgrademoney and it kinda worked (4400 points in Firestrike) and the second card from Upgrademonkey was the same revision as the one I got from Eurocome. I had problem with my Alienware when I installed the replacement 880M card with the one from Eurocome to enable SLi (Revision FF on GPU-Z and Error Code 43 in windows device manager). I bought that my MOBO was broken so I acutally was a few click away from ordering a new one but then I just putted it aside.

This is correct, I installed the 880M's in my P570WM3 (just need to flash the BIOS from P570WM3 to P570WM to make it work with 880M). I did have some PSU problem first but once I got that sorted, ther 880M's worked smooth in my P570WM3! Then I just trieds to install the 780M's in my M18x R2, as I knew my 780M card worked. Result is that my 780M's work perfect in my M18x R2, have tested it all night and no issues at all.

 

Upgrademonkey have provided good support the past 2 months. Just a correction, I didnt get three bad 880M card from Upgrademonkey but only two. My third card which I got from Upgrademoney which replaced my first card that blew was the same revision as the card I bought from Eurocome. My second bad 880M card (first revision card) is on its way to china as we speak and Im awaiting what Clevo has to say when they receive it for test.

Conclution was: My first two card from Upgrademoney was a different revision (Older?) and none of them worked in my M18x R2. First card blew a resistor and second card just made the hole system unstable when I had it installed, blue screen and unable to install Windows Updates and Framework 4.5. The 880M card I bought from Eurocome (to replace the first 880M that blew) was a different version than the two first 880M's I got from Upgrademoney and it kinda worked (4400 points in Firestrike) and the replacement card from Upgrademonkey was the same revision as the one I got from Eurocome. I had problem with my Alienware when I installed the replacement 880M card from Upgrademoney along with the card already installed (from Eurocome) to enable SLi (Revision FF on GPU-Z and Error Code 43 in windows device manager). I bought that my MOBO was broken so I actually was a few click away from ordering a new one but then I just putted it aside.

This is correct, I installed the 880M's in my P570WM3 (just need to flash the BIOS from P570WM3 to P570WM to make it work with 880M). I did have some PSU problem first but once I got that sorted, ther 880M's worked smooth in my P570WM3! Then I just tried to install the 780M's in my M18x R2, as I knew my 780M cards worked. Result is that my 780M's work perfect in my M18x R2, have tested it all night and no isses at all.

Who els have tried a M18x R2 and 880M SLi?

 

and now it happend again...

 

Could be the motherboard spiking the cards then.

 

Yeah, it may very well be the motherboard Aritoshed. I feel sorry for you brother. I hope you get this resolved.

 

Late reply, but I did a bunch of rendering this morning and I decided to show the HWMonitor min/max. Even though it's at -80mV, when it needs power it seems to simply throw up the drawn watts, as I usually operate with 38-40W max draw on the IA cores, but if I do a lot of streaming or heavy rendering it can hit 42 or 44W (at +/-0, it doesn't usually pass 47W however). Here is the screenshot, and you can click on the dots at the top right and select "original file" if you need to see it full-size.

 

Wow, that's terribly low wattage draw for rendering. There is no way your CPU is going to be able produce amazing performance results with low power utilization like that. Are you getting high temps even though the watts are so low? My 3920XM hits just under 100W when doing rendering benchmarks and the 4930MX about 130W.

Have you tried adjusting the TDP higher with ThrottleStop or XTU, or did Intel lock it down at a low value like that for the 4800MQ?

 

The 780M cards works in my Clevo..

 

I think the original goofed up 880M cards did some kind of damage to your motherboard. It may not be something you can prove, but I think you're going to need another motherboard before your M18xR2 will ever function the way it used to.

The 880M cards that Johnksss and svl7 used in their M18xR2 and M15x, respectively, did not have any of the issues you encountered with those screwed up GPUs. I haven't seen many examples of M18x systems that have been upgraded to 880M yet. Probably too many people scared of losing a ton of money on something that does not perform well after see all of the drama going on in thread like this one. Plus, even systems that ship with 880M from the factory don't seem to perform as well as they should.

I know I'm one that got scared off by everything negative about 880M that has been posted. Other than amazing results posted by svl7 and Johnksss, I have not seen one single 880M benchmark that was impressive and most the ownership experiences that have been shared publicly are negative. I certainly don't mind spending money when I know the outcome is going to be amazing, but I'm not very adventurous when it comes to spending a lot of my own money on something where the outcome is so uncertain.

 

That's with -80mV adjustment, and my performance is (as far as I can see) better than +/-0mV, due to it keeping the 3.4-3.5GHz clocks a lot longer under heavy load. It also runs 4-5 degrees cooler than if I left it at +/- 0mV adjustment.

I have no overclocks on it though, and I did the adjustment through the bios, as Myth's bios lets me. Intel XTU had a problem with my computer; every time I used it, when I restarted the machine, it'd shut off before it even POSTed. I had to unplug my power cable, start the machine on battery, then plug in the power cable every time I did a restart or shut down/turned on if I wanted to keep XTU working or use it at all, so I simply ignored XTU.

I didn't want to allow for extra power headroom/drain on the CPU because before I started propping up the back of my machine, I'd hit 86+ degrees streaming or rendering (or even running a virus scan, as the scan uses about 60% of my CPU when running through the ~500GB of data on my SSDs as fast as the SSDs allow) so I kind of wanted it cooler and not hotter XD.

If there's anything you want me to check and report back with, post it here and also in a PM as I'll be shipping my machine back up to mythlogic sometime this week for a board replacement if all goes well, and I won't have the ability to check it for about a month (I'm assuming).

Edit: I should also note I do not have a second adapter. I've been mostly able to run -80mV and +100/500 on each 780M on this single adapter (though I don't OC them much anymore for gaming) so extra power drain is likely a bit of a problem for me XD.

 

Nothing I can think of right now, but thanks for asking. That just seems like really low TDP even for 4800MQ. The dual AC adapter is only beneficial for heavier GPU overclocking and the Alienware 18 does not utilize the extra power that is available due to a power handling limitation on the motherboard. But, the real power drain comes from the GPUs. You should be able to max out the CPU without depleting the AC adapter as long as the video cards are not overclocked heavily.

I hope the replacement works out really great for you, bro.

 

86 degC sounds very hot for a CPU running at 60% usage during your virus scanning. 86 degC sounds more like what it might be during a P95 run. You might want to repaste your CPU if you've not done so already. (Apologies if you have, I haven't checked back through the thread).

 

I had repasted before, and I got SLIGHTLY reduced temps, but then the paste quite literally EVAPORATED on the heatsink. I was getting near 90 degrees doing almost anything on it and we tried repasting and it had almost nothing left of the paste. When I sent it to myth to originally get rid of my retarded shutdown issue, they repasted for me then and the temps were okay enough, but if I don't downvolt it like that, it usually runs at 3.2-3.3GHz when under full load for anything more than a few minutes (though I hadn't discovered propping the back up like this back then). But yeah, I usually sustain 3.5GHz constant with 38-40W drain on this machine. I dunno if my 4800MQ is special or not XD. But everything else is stock.

 

The 780M card actually got the Code 43 error after I tried to do a 50/100Mhz OC on the cards while doing a 3Dmark11 run. The PC froze and after the restart, the second card was "damaged".

 

Keep in mind his room is a blazing hot 34C, so those temps are pretty amazing actually.

As for power draw, I think it's purely because the default settings were used and no tweaking was done. I know I can definitely get my 4900MQ to pull 90+W when rendering if I up the TDP limit.

 

Isn't the IA Core value an estimate of what the chip itself is pulling based on the VID? Mine doesn't pass 45W but I couldn't tell you what it is at the wall as I'm too cheap to buy a kill-a-watt

Sent from my HTC One_M8 using Tapatalk

Here's what my CPU does.

Stock:

Spoiler

-80mv undervolt in BIOS:

Spoiler

I took them both at 99% so it could easily be seen what is going on with the clocks and temps. Stock, it settles down around 3.4GHz while the undervolt pushes to 3.6GHz with all 8 threads loaded. Both of them hit 80C, the stock had a 1C difference which is basically a wash.

I didn't make any other changes to the BIOS and I don't have any tweaking programs installed on the machine so that's everything stock but with an 80mv undervolt in the BIOS.

 

I don't get how your -80mV gives the same draw and such as your stock... I know mine hit 47 and 48 almost constantly without the undervolt.

 

He has an MX chip, which will turbo as much as it can until it hits either the thermal or TDP limit. So when an undervolt is applied, it will just apply more turbo until it draws the same amount of power.