[Review] Sager NP9873 / Clevo P870DM3 Quick Review by HTWingNut

This might very well be causing the gap :

 

Ok, success! I think.

I swapped GPU's back to original. I used double the amount of thermal paste and temps are now what I'd consider reasonable:

Time Spy Benchmark:

CPU @ -170mV: 77C
GPU Master: 82C
GPU Slave: 77C

Fans are near maximum throughout though.

If I drop GPU boost speed by 100MHz temps drop by about 6-7C with only about a 2% decrease in performance. So if temps are a concern, that is an option as well. Wish they allowed undervolting.

I agree that a huge gigantic heatsink is not a good idea, because getting alignment is a pain the rear. They need to at least make the touchdown surface somewhat adjustable independent of each other. But for now, using a good quality paste and an excessive amount of it for now seems the best bet. I think IC Diamond is a good choice because it remains thick and has good overall thermal conductivity and lifespan. Liquid metal isn't really an option because the gap is too large and inconsistent.

In any case, now I can do some more downtown benchmarking.

 

Naw, I removed all the thermal pads and the thermal paste spread pattern didn't change. I did even fix that pad so it wasn't overlapping.

 

Gotcha. But the new temps seem way better with that excessive application.

Looking at the heat sink though, seems like Clevo took pointers from something like this :

Spoiler: :P

 

Yeah the actual cost of this vapor chamber heatsink must be at least $100.

 

The vapour chamber means it all has to be one piece, there is no other way of doing it.

 

And to add to the above... this part is so new it is not yet in stock as a stand alone part...

Sent from my SM-N930T using Tapatalk

 

I understand. But there's always a way. How do you think flubber was invented.

 

But it's good to note that skimping on the paste for this one is a bad idea.

 

Personally I would not compare stock and OC when comparing two different devices, okay for the same device though. Anyway why can you not undervolt? Should be able to flatline (for instance if wanting no higher than 0.9V then set all points from 0.9V and above to same frequency) or pick a single voltage point, a least for testing purposes.

 

Haven't tried too hard but not readily available as an option with Afterburner or Inspector.

 

To be quite frank, this is only a Clevo problem in general. Other laptop heatsinks are generally non-warped. Lapping may help someone else slightly, but lack of contact is generally a non-issue on other laptops. Whether or not fans get air, are pasted properly, or have good fan tables? That's something else entirely, as you know with your having to force your AW fans to a higher speeds in the past, yes?

If MSI made this exact model for example, I guarantee you contacts would be good. Maybe not perfect like some top quality desktop stuff gets, but it'd not be having any issues like this.

Prema said that auto fans cap out around 60% this time, rather than 80%+, so you should try with max fans.

Since Pascal to a lesser degree has the same voltage variance stuff as Maxwell does, undervolting is a death sentence for stability. Maybe with Prema mods if voltage is forced to a constant we might get somewhere, but then that might also shoot up temps a bit, which... we probably don't want. *shrugs* Who knows maybe it'll drop temps, lol.

 

We need MSI, Clevo and NVIDIA to all sit down together and take another shot at Pascal, including their sinister choice in form factor. And, this time actually think about how their actions are going to affect every single laptop owner in the world with normal MXM GPUs

 

Thank you for investigating all of this. Now I know what to do if I have similar issues when I get my P870DM3. Some follow up questions for you.

When you say you used double paste, can you be more specific? Normally I use a pea sized amount of IC Diamond when repasting, placing it in the center. So would you say two peas worth, or a kidney bean or small grape if we're using food as an example? I don't want to use too much as I know that can cause problems too.

Also, I was planning on paying extra to have HIDevolution use liquid metal for the CPU, but it sounds like this may not be a good idea as pastes with different thickness could cause the heatsink to rest unevenly. Not sure why, but all the resellers only offer the liquid metal for the CPU. Is there a reason that stuff isn't offered for GPUs? Why not just use it on everything?

 

Look at his video of repasting. Almost double the amount. Regards Liquid ultra on cpu ... There should not be a problem on processor(Separate heatsink). The reason they do not use Liquid Ultra on gpu's... Because there is not much to gain by switching to Liquid ultra. If you or others are not used to dealing with Liquid ultra... What happens when you shall clean and repast gpu's yourself? Probably will be a problem for some. Easier on the processor.

 

Meaker has his machine running at 65-70C tops on GPUs. While 2 owners reported this high 90C temps, think about the 2xTIM volume later but first give it a run if temps go north then you might be needing to check for alternatives. IF you are hardcore bencher and enthusiast go for Liquid Ultra I'd also prefer to do this myself so that I can gain experience with the Liquid metal TIM jobs, Consider the worst case - If the heatsink is warped and the CLU spills out ( Read this happen to the P7xxZM owners) that could cause a havoc since it's conductive and breaks AL on contact. Thermal Grizzly Kryonaut / ICD / Gelid GC Exteme should do the job.

 

As I said I have had very limited testing with mine. I've had hardly any time with it so please take my shirt benchmark runs with a grain of salt.

I only just migrated to the SSD and have found a position I like on my desk. I want to do some long heaven or valley runs and we can compare balance points of frequency and temperatures.

 

I dont get why people have problems with CLU. One really has to do something out of the normal to spill it. The only problem I ever had with CLU is that dam brush keeps soaking it.

 

Urban myths and fear of the unknown more than anything else. It's just silly. Crossing the street with your eyes closed is dangerous. Applying CLU in a careless manner is not a good idea. It does not "spill" or "run" from where you place it.

CLU is the absolute best option to use whenever possible. Unless a person is very careless and haphazard there is no reason to fear it. When applied in a proper and deliberate manner it is not dangerous.

 

Two long lines the length of the die. If I repaste again I'll take a photo.

Liquid Metal for CPU should be fine. It has a small heatsink and independent attachments from the GPU's so it should sit pretty parallel to the CPU IHS surface. For the GPU forget it. You need a large amount of paste to get proper coverage and heat transfer and that just won't work with CLU.

 

While I obviously trust Prema, while benchmarking and fans sounded loud I hit the max fans and barely any audible noise difference or temp difference. I did not check to see if a sensor for fan RPM is readily readable though.

 

let me politely disagree there @PrimeTimeAction and @Mr. Fox i consider myself neither careless nor a total noob when it comes to hardware. still, this exact thing happened to me and my beloved dark knight due to non perfectly fitting heatsink. clu application was just fine for several weeks / months before little tiny droplets started seeping out from under the heatsink. this was especially noticeable after transporting my machine in a shoulderstrap bag and it thus being in an upright slanted position for longer periods of time. it definitely sucked, luckily there was no permanent damage to my internals through a short or smth. in any case, point being it can happen to any of us if murphy is involved... (ask @ajc9988 about his ram socket adventure with clu!)

however, i completely agree that clu is the absolute best when it comes to cooling performance! it lowered my load temps by a whopping 10C for the cpu, thus leaving sufficient headroom for very nice overclocks

Sent from my Huawei Mate 8 NXT-AL10

 

Maybe you have OC fan profile active and not Automatic.
HWInfo can read out the fan speed of one gpu fan and the cpu fan. When put to max it should be around 4300 rpm.


Gesendet von meinem SM-G928F mit Tapatalk

 

To be fair, I had a gap in contact that it slid out, also due to elevating at a huge degree for the back (around 60 degrees) which allowed for it to come out. If good contact is present or not shocks, etc., it stays in place...

Edit: but, if you are highly active and carying it vertically while riding a bike or walking and contact issues are present (which seems to be an issue with Clevo heatsinks), then weigh your options and verify contact!

Sent from my SM-G900P using Tapatalk

 

In that case, never use Liquid Pro, Phobya Liquid Metal or Conductonaut. Those work as well as CLU, but they are very watery and have no viscosity at all. If your laptop was subject to enough movement to displace CLU those other options are definitely not going to work out, at all.

One thing to consider in a rough use scenario is laying a bead of automotive RTV silicon around the perimeter of the IHS to create a gasket or dam as a barrier. That should not interfere with heat sink contact and it is not difficult to clean up if you need to later on.

 

Oh! that really sucks. I was not aware that people actually had problems. I was thinking that if a total noob like myself has done multiple times without any issues then it must have been very simple thing to do.
Having said that, I still dont think it is dangerous enough that people should be scared of attempting it. Things can go wrong with anything.

Just a word of caution, Liquid pro should be avoided in any case, its more difficult to apply and more difficult to take off. If you leave it for a long period of time, it can make solder like bond which is really difficult to take off without risking damage to components..

 

Ultra has been on my 13" light notebook for several months of it being shoved in bags and moved around with no issues plus due to the 128MB of cache on the chip and the large surface area there is a fair amount of it. Plus there are surface mount parts uncovered around the chip.

It's for perfect mating scenarios only.

 

Then something is most certainly weird there.

 

I have an audible gap between the fans on auto max and on manual max.

 

OK, I'm a dolt. I had overclock profile set in Clevo Control Center that's why fans were so high and that's why temps were so low. I went back and ran regular auto fans and results were as follows:

CPU Set to -170 mV in all runs

Auto Fans:
CPU: 86
GPU1: 88
GPU2: 80

Overclock Fan Profile:
CPU: 77
GPU1: 82
GPU2: 77

Full Fans:
CPU: 78
GPU1: 84
GPU2: 76

So for some reason overclock fan profile did about same or slightly better than full fans. Maybe with a few runs they may be about even.

In any case, the extra thermal paste seemed to help. Just I don't like GPU temps so close to the thermal throttle limit. Granted thermal throttling means minor drops in clock speed. When I underclocked by 100MHz GPU temps were a solid 5-6C less. So it might be a fair compromise if you want a reasonably quiet machine and not bordering on thermal throttle limit.

 

Btw. can someone check the max RPM of the fans in hwinfo? (Fn+1)
I think you need the EC communication thing enabled for HWInfo to show that.

 

4.2k.

 

Sweet!

 

Any info on the ESS Sabre HiFi DAC ? Couldn't find any info with the 775DMx lounge too on this matter..
Thanks !

 

I don't see it anywhere unless it's the unknown device that has the Synaptics vendor ID. In any case, there's certainly no dedicated output jack for it.

 

That's what pascal has been designed to do, that's what it does on the desktop too.

 

NI will not help with Pascal voltages. Perhaps AB 4.3.0 beta 4 doesn't like your PCI-ID of 1BA0, have you tried creating an MSIAfterburner.oem2 file with the PCI-ID?

Code:

;OEM

[VEN_10DE&DEV_1BA0&SUBSYS_1BA01558&REV_??]

; Clevo GTX 1080

Desc       = Clevo GTX 1080
VDDC_Generic_Detection   = 1

Just making up that VEN string, use ctrl-i in AB beforehand to get the proper one. Hopefully then ctrl-f should give you the voltage frequency curve to use. If you are still having trouble I can write a rough program to set voltage lower if you wish. 1.062V seems way too high for 1.8GHz.

Forgot to mention that you would need to set voltage to 3rd-party in prefs.

 

+4k RPM?!
My P870DM1 has max 3,3k RPM on F+1... That DAMN much... :O

 

They have new 12v fans, automatic mode maxes at 60% rpm because clevo found the fan noise too loud too explains the high temps in auto fan mode.

 

There are no fan sensors anywhere that I can view. As far as I can tell fans keep kicking into full on their own even with auto profile set. I only can tell by running a GPU intensive game or benchmark and then hitting Fn+1 and listening for audible cues. Fn+1 is barely louder than auto set, and it's really loud.

 

HWInfo64 picks them up for me as long as I enable monitoring the EC.

 

should i cancel my order and get the gt73vr or gt83vr advice pls

 

I have EC Support checked in the Safety Tab of HWinfo settings but still no fan RPM.

Edit: Nevermind, I found it in sensors and activated it. I swear I did not see that before.

Also, I think the GPU OC app may be a little glitchy. I had both GPU's set to OC off, but in the system tray it said GPU OC is ON. I set them both to ON and saved, then back to OFF and saved and now it shows they're off in system tray and fan speeds no longer kick in full on their own.

 

I wouldn't cancel. Others aren't having issue I am, and in my case, it may be operator error at the moment. Lack of sleep, exhaustion, and too many other things going on in life at the moment I think is making me not think clearly. Usually I have this stuff sorted in no time.

 

@HTWingNut In my experience this can show and hide itself for no reason in HWinfo64. Usually triggered by shutdowns and reboots.
Your fan problem sounds like you've been tweaking with XTU and it's goofed your fan settings as it's been known to do.

 

I went through and double checked everything. No XTU running at the moment. It seems everything is stable now (finally) and should be good to go to crank through a bunch of benchmarks.

 

do a benchmark video plssssss

 

nice review video.. but as i suspected even with under volt, the need to cool off two 180w class GPUs and direct heat to CPU result in super hot CPU... clevo should just go straight water cooling and not being scared of it lol, either that or simply have 4 FANS in the laptop and two fans work on the CPU instead of 1.

 

Better just to make it thicker if you are going to do that.

 

Seems like Clevo didn't address the goofy fan tables I experienced at times. I wish they could be set from the bios.