*** Official Clevo P37xSM-A/Sager NP9377 owner's lounge ***

Certainly looks like there's enough Liquid Ultra. Could be the pad, one quick way to test is to remove the pad and see what the temps are like. Just don't stress the machine too much in that state.

If it's indeed an issue with the pad, get a 0.5mm pad (stock pad is 1mm). Or maybe get a new heatsink altogether.

Probably also a good idea to request some pressure sensing film and obtain a pressure map.

 

Ugh. What a nightmare. That pad doesn't even cover those components completely either. It covers one and a half rows at the most.

I'm going to assume that my CPU package itself is fine then? It really doesn't look good. Big black line going across from half of the 1st all the way to half of the last core. Ive never seen anything like it.

Sent from my HTC One_M8 using Tapatalk

 

Here are some thoughts... strange deal.

 

Black line could just be residual junk from the previous TIM. If the die actually cracked I doubt the machine would even POST.

The pad needs to be aligned with the outer edge to cover most of those mosfets. If I recall correctly the Alienwares don't use mosfets for power delivery to the CPU, so could probably get by without a pad.

 

Well I added more and core 0 is at 71C @ 3.4GHz which isn't great and it incrementally increases across the chip - core 2 is 73C, core 3 is 80C and core 4 is 84C. Those are huge increases but I didn't do the slow tightening of the screws because I wanted to just test it. Right now it's performing worse than IC Diamond on all cores except the first which is 3C lower than what IC Diamond gave (I have my chip at 47W TDP right now). And that's with forced max fans. Let me try removing the pad and redoing the whole thing. This tube has a lot of paste in it, it came at 15 and its down to 13.

The interesting thing is that my idle temps? The first and last core are the hottest at 30 and 32 minimum with the second being 30 but the 3rd goes down to 27C. Gonna clean this up and I'll take a picture of my core. It isn't pretty.

Sent from my HTC One_M8 using Tapatalk

 

When looking at the CPU die, which core is 0 in relation to the VRM position?

I went and respread the paste (because its still liquid, I don't think there would be a problem with doing this but feel free to correct me if I'm wrong) and now the cores are running the opposite as to which is hotter... This tells me there is sufficient contact pressure but my technique needs work.

Last run, it was core 4 that was hottest, now the 1st core is the hottest then the second, third, and fourth in that order. Core 1 is 11C hotter than 4, core 2 is 8C hotter, and core 3 is 3C hotter.

The lowest core is 75C @ 3.6GHz which is an appreciable drop if I can get the other cores to match. The CPU idles in the mid 40s now when it was idling in the 50s before.

This time I spread the paste and then did the incremental tightening following the screw order on the heatsink.

But yeah if I can figure out which core is which I can probably make this work. I'm going to unscrew it and tighten the screws at the other side right now and see if that works.

 

(IC Diamond has been known to leave scratches in chips & heatsinks, due to the hard particles, maybe that's what caused your line across your CPU. You're not having much luck when it comes to laptops Ethrem it seems.)

 

Usually you would clean the heatsink and start again to avoid air bubbles.

 

That's certainly one really nasty scratch if it's from the IC Diamond... It goes across the whole chip. Figures that I was just saying the other day that I have no idea what people are talking about with scratches and then bam, big scratch across the die.

Well it looks like my results with liquid ultra won't be that great on the CPU... I'll have to order some contact paper. I did it again only reversed the screw pattern and I've got 90C on the first core, 90C on the second, 83C on the third, and 77C on the last. The only core that is better than what I would get with IC Diamond is the last one and this a 3.6GHz bench - the 90C doesn't show up with Diamond until 3.9GHz.

Maybe it will do better on the GPU though. That slave GPU is my biggest priority for it. Since the heatsink assembly is separate from the memory sinks, it shouldn't have to worry about the pads. Let's just hope it has enough contact.

Sent from my HTC One_M8 using Tapatalk

 

How are you going to get air bubbles in a thin layer of metal though? It was completely respread til it was flat with the little beads which I moved into the middle of each core.

Sent from my HTC One_M8 using Tapatalk

 

My advice to you is to put away the Liquid Ultra until you can resolve the mechanical failure of the heatsink not directly contacting the CPU die. Until you get that fixed thermal paste selection is a very distant second.

As Robbo said the ID7 can scratch the surface of the die. My CPU abd GPU both have scratches from that stuff.

 

Lol I can't even get this stuff off now. It's just moving like foil across the die. What weird material

Sent from my HTC One_M8 using Tapatalk

 

There are dark spots and then small scratches. The dark line that crosses the whole CPU is what concerns me. It is pretty bad.

I don't really think there is anything I can do about the contact pressure at this point. Gonna stick GC Extreme on it and see what happens. Hopefully it will be close to IC Diamond because I don't ever want to use that stuff again.

Sent from my HTC One_M8 using Tapatalk

 

When you sandwich two surfaces together air gets trapped unless the paste spreads out and clears it. Hence why the blob in the middle method is preferred.

 

That doesn't really work with liquid ultra though since it has to be spread out before the heatsink goes on.

Sent from my HTC One_M8 using Tapatalk

 

Ah surface tension may do it with the metal, never mind then.

 

WOW! Mind. Blown.



That's GC Extreme. You mean to tell me that my CPU die is screwed up because of an inferior paste? -_-

That's with auto fans... Its 8C lower at max than IC Diamond was doing with the same settings. Idle? Mid 40s. Previously? Low 50s.

I thought I screwed up and didn't connect the fan or something and the system was going to shut off because it didn't kick on when Windows started like it always did.

I have seen the light and IC Diamond is going in the trash!

EDIT: Here's my 3.9GHz overclock with max fans that IC Diamond hit 92C on



I bet I can run 3.9GHz as an every day overclock now.

I have to say I'm incredibly impressed with this material, lets see if it improves over the next 24 hours or so.

Its too bad CLU didn't work out for the CPU but I'm going to try it on the slave GPU now.

 

My GPU and CPU look like yours. I would not worry about surface scratches and move on.

 

That's what I'm doing lol.

So far the liquid ultra is doing much better on the GPU.

I wonder if I'm being too light with how much I'm using.

I did lose a lot to the damn cap too. I see what people were saying about it. It really does go all inside the cap. Such a weird material...

Sent from my HTC One_M8 using Tapatalk

 

I guess I'm going to have to RMA my machine. The liquid ultra has the cards rising at the same temps now but I'm still throttling below the stock 954MHz almost immediately.



Sent from my HTC One_M8 using Tapatalk

 

Are you sure there is anything wrong with your machine that requires you to RMA it? I mean we know the 880m's have a throttling issue, and its basically up to NVidia to fix it (or a vBIOS mod maybe?). As far as your temps that's part of the deal if you're overclocking and using extreme cpu isn't it?

For comparison sakes, what were your stock temps before you did anything?

I'm been impressed with the low temps on the NP9377 completely stock (factory IC Diamond paste job). Granted, coming from a few Razer and MSI machines I was used to a lot more heat. With a 4810 cpu my idle temps are upper 20's, GPU's are also upper 20's. Trying to replicate your throttlestop TS bench my CPU maxed in low 70's during 8 threaded 1024 test. That is absolutely astounding IMO.

If you bought the machine to benchmark with you probably already know that these kinds of issues go with that hobby, and the risks of either getting a poor overclocker or even doing damage. If that's the case that does suck that he machine isn't working out for what you intended.

If you bought the machine to use daily and play games on, I say go play games and forget about benchmarking! It's an obsession that you just cant win, especially with these newer GPU/CPU combos that are very difficult to overclock and get decent results.

Let me know if there is any testing or anything else I can do to help your troubleshooting. As long as it doesn't involve cracking the seal on my heatsinks

 

Lol my stock temps were higher, the CPU with stock paste was running 85C under regular load. I've never had good idle temps CPU or GPU. The absolute lowest my CPU has ever gone, and only with Max fans, is 27C. The CPU heatsink is extremely bad.

As for the GPUs, the temps were about 3C higher than after I repasted, running around 85C under load as opposed to 82C.

Ive been having issues with the 880s since the start. I still haven't figured out what causes them to randomly blank out the screen.

I've had issues with it since I got it. My tracked left button is messed up, my lid has a lot of flex, the temps are high across the board, random instability, my msata runs at 70C and underclocks itself...

I didn't want the hassle of sending it back so I've tried to resolve everything myself but there are things wrong with this machine I can't fix no matter what I do and it might just be time to ship it to get repaired.

Oh and I'm not overclocking the CPU. I actually applied an 80mv undervolt to it which caused it to overclock itself. The temps you see are the same whether it's undervolted or stock, just the clock changes to max out the TDP setting. I have a 4940MX which will eat anything you throw at it, it's not built to conserve power. In fact I discovered that my chip will run the same stock 3 4GHz clock with a TDP of 47W instead of the 67W it defaults to.

Sent from my HTC One_M8 using Tapatalk

 

Ah, ok that makes sense with all those different issues. Sorry, I wasn't trying to offend questioning you.

Again, if there's anything I can do let me know. My machine appears to be working great, and I'm not planning to overclock. Other than installing drives the only change I made was adding a thermal pad to the intel chipset. I'm pretty sure that didn't help at all, but I just hate seeing an exposed die a mm from a huge metal heatsink (keyboard bottom).

 

Yeah that line across the CPU looks normal, even mine has it and it's only ever seen MX-4 and Liquid Ultra. As for CLU going inside the cap, just make sure you pull the plunger back after using so the metal retracts a bit into the syringe.

I also received my Gelid GC Extreme and so far it's been... weird. Temps are higher than MX-4 without using my 120mm fans on the U3, but with the 120mm fans on temps are 3C lower on the master GPU and about even (0.5C higher) on the slave GPU when gaming. One thing I did notice is that Gelid does seem more viscous, and is more "rubbery" and doesn't stick as well to the die. I had some theories but they all went out the window when I saw the temp drop with external fans, just makes no sense.

EDIT: My original theory was that for low pressure applications (GPU heatsink), a more viscous paste (like Gelid GC Extreme) would actually perform worse because it wouldn't spread as much, and the TIM layer would be thicker and heat transfer would be worse. What I couldn't figure out was if this was the case, why blowing cold air directly into the heatsink would actually make it perform better compared to MX-4.

 

Well it certainly works great on the CPU, better than IC Diamond, so I'm happy with that.

The first time I cleaned off my CPU it didn't have any lines or imperfections at all which is why it freaked me out seeing that.

As for the video card, I'm leaning more and more towards it being a bad card. I will probably call Sager to start the RMA. I don't have any more patience to deal with it and liquid ultra is too much of a pain to remove and reapply so I'm not going to switch the slave and the master just to find out that it's defective anyway. It's unacceptable to have the card dropping below it's stock clocks. I was able to make an excuse for it when the temps were so far apart but they are literally the same between the slave and the master now that I put the liquid ultra on the slave but it's still down clocking. Flashing the modded vbios is not an acceptable solution because it won't drop to 2D clocks which causes the fans to run all the time and feeds the cards 1v 24/7 needlessly.

Sent from my HTC One_M8 using Tapatalk

 

Well the liquid ultra has definitely improved even more. My slave card now takes longer to heat up than the master and runs 1C cooler than the master does when the fans kick in.

The core is dropping to 888 though.



I'm going to back up everything, switch gaming to the desktop, and prepare to send the machine off to have the card replaced. A card that can't maintain its stock core is defective and its been seen on this forum that people who have SLI 880M don't drop below 954,they just don't boost often. Then there's the fact my slave has always been hotter than the master which apparently is a red flag too.

I guess I get to see how Sager is with warranty repairs. Any guesses how long I'm going to be without my machine? And should i clean the ultra off and put on some GC Extreme? I don't want them to be sloppy cleaning it up, short the card, and then say it was my fault.

Sent from my HTC One_M8 using Tapatalk

 

It will be interesting for sure. I wouldn't be surprised if they denied warranty due to over clocking, vbios changes, or thermal repasting. Keep us informed.

 

All of that is allowed except the vbios but they void the warranty if you actually kill the card with a custom vbios, its not dead and its on a stock vbios it's just not operating at its rated specs.

You're allowed to overclock (in fact the Clevo control center has an overclock button that links to XTU), you are allowed to open up the machine and change the thermal paste. These are enthusiast machines, not your run of the mill consumer machines. Even Dell had me open up my Alienware and check ribbon cables when I called their tech support. There is no hardware damage and they already have a log of these issues that I had when I got the machine. Doesn't matter if I RMA it then or RMA it now, the situation is the same as it was before (although thanks to the liquid ultra the temps on the slave are better).

Sent from my HTC One_M8 using Tapatalk

 

Out of curiosity what are your load temps on the 880s with maxed fans? (hit FN+1 to force max fans) I'm curious if your slave card is hotter than your master like mine is.

If you don't know which is which, set a game to single GPU with the nVidia control panel and use hwinfo to see which card is being used.

Sent from my HTC One_M8 using Tapatalk

 

After 30 mins of Watch Dogs on ultra at 1920x1200, my master maxed at 74 and the slave at 75. The average temps agreed with my max that the slave was about 1C hotter. I know that card 0 in hwinfo is the master because I have been playing Cryostasis off a single GPU.

 

74 and 75?!

Do you have any special cooling? Laptop fan for example?

Do you know how to use hwinfo and RTSS for OSD to monitor the core clocks? I'm curious if you get the slave throttling like I do.

Sent from my HTC One_M8 using Tapatalk

+rep btw

 

I am using a cooling pad with fans, and the fn+1 max fan setting. IC diamond paste from Sager direct. I bet I am getting throttling like most are. I will try to do a longer session tomorrow, I was hoping within 30 mins of Watch Dogs I would hit max temps but that seems low to me.

 

It takes more like an hour. If you want to set up an OSD like I have, let me know and I'll dig up my post on it. I have mine set to show temp, core, and memory of both cards, the CPU temp and the core speeds and the framerate. RTSS is quite useful I pretty much never turn it off

Sent from my HTC One_M8 using Tapatalk

 

Those temps sound about right. Watch Dogs actually doesn't heat up your cards like some other games (*cough*Far Cry 3 *cough*) do. It's what I use to test my cooling mods too because I don't like cooking my cards for no reason.

 

Watch Dogs cooks my cards... What settings are you running? Watch Dogs burdens them more than anything else I have

Actually i take that back... Bioshock Infinite burdens them the most if I don't turn on adaptive vsync.

Sent from my HTC One_M8 using Tapatalk

 

Watch Dogs everything on Ultra and 2xSMAA with all cooling mods. Valley benchmark for 10 minutes without U3 fans. Mind I'm running 780M which runs cooler by default. Using auto fans btw.

So I just did a second 30 minute run with Watch Dogs, and the results agree to the first run to within 0.5C with Gelid Extreme, so I know it's not an issue with curing. (picture above shows second run) Master 780M peaked at 68C and averaged 64C, while slave 780M peaked 72C and averaged 67.5C. The master 780M temps improved by 2-3C compared to MX-4, but the slave temp stayed the same and the average is actually 1C higher than with MX-4.

I *think* the master heatsink exerts more pressure due to having 2 washer-like tabs on 2 opposite corners. The other thing is I lapped my slave 780M heatsink. Starting to think the higher viscosity and less pressure with the slave heatsink is why temps are slightly worse compared to MX-4, since the paste can't spread as well or thin.

What's really interesting though is that in the Valley benchmark, the temps are higher compared to playing Watch Dogs, because I did not use the 120mm U3 fans in the Valley benchmark. If someone can give me a reasonable explanation as to why Gelid performs better than MX-4 with external cooling but worse without that would be awesome, as I've been scratching my head the entire afternoon over this.

I'll stick with Gelid for now since I'll always be running the 120mm fans when gaming. Just need to see how long Gelid can hold up before losing its effectiveness. Wonder how the ShinEtsu will fare given that they're even more viscous than Gelid.

 

I would imagine it's just hard to control every variable in between your separate tests: room temperature, exact position of laptop on cooling pad.

 

Sure, but that's why I ran the Watch Dogs test twice, and both times the temperatures were remarkably consistent and within 1C of each other. This is also why I report the average temp, because that's less susceptible to fluctuations, and averaged over a long enough time period (30 minutes) a 1C different becomes statistically significant.

 

That doesn't take into account room temperature changes, nor precise position of the laptop on the cooler that I had mentioned.

 

I have to agree with Ethrem, Watch Dogs has always been hard on my machines. I've played it on an 870m and a single 880m previously and it always got my card the hottest. I'll try a longer session today with the 880m SLI.

Normally I use hwinfo64 and just log everything to review temps and clocks for throttling. I am about simplicity, one program that does it all, hehe. If that OSD is easy to setup I can give it a go (and replace FRAPS). I know years ago I fought with other OSD programs that would quit working after updates and it became a huge pain to keep going.

********************************************

I did longer session and looked at the data, observations below:

I did an 80 min log this morning with watch dogs and got some pretty cool info from it.

Test conditions:
room temp 76F
on cooling pad, sorry I don't know model or brand its really old lol
Max cooling (fn+1)
4810 cpu stock clocks and 30mv undervolt
880m SLI, stock clocks/voltage/vbios, 337.88 drivers
1920x1200 resolution on external display using mini DP, ultra settings, ultra textures, FXAA

The game played better than it did on my single card (870m/880m)laptops previously. I had not played with the SLI laptop much since I completed the main story on a previous single 880m laptop.

After 80 mins I had 2122 readings from HWInfo64. Some things that skew the readings and I cannot account for, so I can only note it here to keep in mind. I task switched to the desktop several times during the test to check temps and logging, and looking up some things online. During those times the GPU's lowered their clocks since they weren't being utilized so some of the low clocks are not the bad, gameplay affecting kind of throttling. Also realize that my averages listed below are skewed slightly from those times I switched to desktop and the temps dropped for a few seconds, as well as when I went to the menus and the FPS skyrocketed way up. One last thing, I had a fraps issue mid game when I was trying to take a screenshot and accidentally started recording which lowered my FPS for a while until I figured out what I had done.

In 2122 readings, only 4 reached the max of 77 on the master GPU, and only 1 reached the max of 78 on the slave GPU. This says to me those are very rare and realistically. 76/77 had more readings (about 100) and many more were 75 and below. The average was 70/70 for the entire duration of the logging.

Average core/mem freqs were 975/1228 master, and 975/1229 slave which is great and says to me there is not much for throttling if I could somehow exclude all the times I task switched to the desktop. 1890 readings were at 993mhz core on master and 1840 readings were at 993 on slave. 111 readings were below 954 on master and 105 on slave. 2020 readings were 1248 memory on master and 2025 slave.

The GPU memory allocated max was 5240mb, and average was 4332mb. I realize that doesn't really tell you anything directly since I don't know how much was actually being used.

Take this with a grain of salt due to the super high menu readings and the 0's from when I was on the desktop, but the average FPS was about 80 for the duration. If you exclude the 97 readings above 150 FPS (menus or facing a wall I'm sure) and the 28 zero readings while on windows desktop (Aha! that may be a good number to exclude to get a rough throttle count!) the average comes to a more realistic 64 FPS which seems reasonable to me from the game play session.

Lots of numbers, and while I enjoyed going through the data, I wouldn't want to do this every play session, hehe.

 

I tested the positioning and it's fairly forgiving (1" offset on either of the 4 sides gives virtually identical results), so that'll account for 1C at most if even that much. If anything, room temp was lower (24C) when I tested Gelid compared to MX-4 (25C).

And the Valley benchmark test was done without using the cooler fans, so I feel very confident in saying Gelid performs worse without external cooling.

 

Doesn't sound feasible to me, I can't imagine the paste type causing the unusual effects you described, I still think there are other external factors that have not been accounted for. Just use the paste that you're most happy with.

(The paste is just an interface between the GPU chip & the heatsink - it can't influence the effectiveness of external cooling from a laptop cooler - all it's doing is transferring the heat from the core to the heatsink.)

 

Wow that's a lot of work lol.

Grab RTSS and install it, launch it, check show own data to turn on FPS, open Hwinfo in sensors only and hit setting then OSD and click on the values for core and temp on each card. Put the master in line 1, the slave in line 2.

You can add the CPU or whatever was you want as well.

Now when you launch a game, you just open those two programs and you have OSD. I would change the color to white instead of Magenta in RTSS as well.

I doubt it's compatible with Fraps running at the same time but you can try it. I'd turn off the Fraps FPS counter.

I would be really interested to know your settings that so I can try to duplicate it, it does sound like you are boosting most of the time and for the most part, your slave and master cards are clock synced which lends more to the possibility that I have a defective card after all.

Thanks so much for doing this!

I backed everything up to my desktop last night and played a few games. It amazed me how my 780 Ti was faster in most games but took a total beating in Watch Dogs. It used to run smoother than that so I'm guessing that the patch broke it.

It is interesting that your average mem clocks are so low though. I would add their values to the OSD too.

Sent from my HTC One_M8 using Tapatalk

 

What I'm getting at here is whether it's possible for a paste to trap heat locally for whatever reason, thus leading to higher temps. But once you apply external cooling, it actually becomes more effective since that trapped heat can now be efficiently carried away by cool air blowing over it.

Also, it seems that I would need to use Gelid on the master GPU, and MX-4 on the slave GPU to achieve the best results lol. Hopefully the Shin Etsu will get here soon because I'm really curious to see how those would fare.

 

Haha, sounds like a load of hooey to me!

 

The thermal conductivity figure will be fairly fixed, different pastes may work better or worse depending on pressure.

 

That's what I was thinking: lower pressure = less viscous better as it will spread more easily/evenly.

 

Now, that I agree with. There was an article on Toms Hardware that said as much too.

 

Of course the caveat here is after a period of heavy usage, the less viscous paste will pump out faster and so the temps will start to creep up. So longevity won't be as good compared to a viscous paste.

Although given the difference I'm seeing between Gelid and MX-4 on the slave GPU, I may as well use MX-4 because by the time it's pumped out enough that temps start creeping up, it'll be on par with Gelid from a fresh application anyway.

The proper solution would be to increase mount pressure, although there's the very real risk of cracking your die if you don't know what you're doing (and I don't), so for now I'll just keep searching for a better paste.

 

I've heard some good things about mx4, including it's easy application, which is half the battle - if you do a good job applying the paste that's the biggest factor for good temps (unless the heatsink mounting system has some damage or poor manufacturing tolerance). I'm not sure that paste 'pumps out' though. I've not experienced that. I've used Arctic Silver 5, using the spread method that they list on their website, and I never see it running outside of the GPU core (I use a very small amount). The temps can gradually creep up though, but often times I think that's due to the paste becoming harder over time - and I have a suspicion that on the odd occasions I've transported my laptop around when the paste is harder I've seen the temps shoot up after the transportation. I think the vibrations during travel can maybe crack the seal of the paste on the heatsink & core interfaces leading to poor performance, whereas that is less likely to happen when the paste is younger & more pliable (or if you don't transport your laptop) - just a theory though. (I've only ever used AS5 though which is quite viscous, so maybe the other softer pastes do run out over time, I don't know).