Aw17R4 7820HK/GTX1080 : kryonaut repaste & arctic pad change

@iunlock , well from your screenshot in your disassembly thread where you tried the repaste with kryonaut, HWinfos clock shows 30min, that's why I was asking because as its the same setup (appart from the fact that it was on the 6820HK that you did it), I wanted to compare by having the same conditions : so was it playing Overwatch?
If yes, I will need to find a key for this game in order to test lol, as I am pretty sure that like any Blizzard games, it's not very resource demanding.

And no, I am not using TS, just setting base values in the BIOS (core multiplier + PL1 and PL2) and then fine tune in XTU and monitor with HWinfo.

EDIT : During my lunch break I went home and uninstalled XTU to try TS out. But when opening it, it retained my 120mV undervolt even after properly uninstalling XTU.
Anyone could tell me where I should look for hidden files retaining the information?

 

I am finally done with the 3rd repaste and tests.

Here are some temps :

2 hours Battlefield 4 at 36x with a 100mV undervolt :



2 hours Assassin's Creed Origins at 36x with a 120mV undervolt :



40 mins Assassin's Creed Origins at 40x with 120mV undervolt :



I won't put more screenshots at 40x, it renders the same : It can go to 99° on max value and the average sits around the values you can see (86/81/86/81).

Regarding what I did/changed :

- I checked all pads before taking the heatsink off and checked the gaps
- I removed the heatsink, added pads where gaps were obvious (bottom part of the GPU and I remove a little square of the 0.1mm on the location that I showed before in a screenshot and added a 1mm pad, there was a big gap there)
- I used a 0.5mm pad that I put on top of the CPU, screwed the heatsink back on without pressure, removed it and checked : perfectly square imprint, looking exactly as your screenshot @Pete Light.
- Went more generous with the Kryonaut (nice even layer with a pea in the middle as suggested by @iunlock ).

Now comes the next step:

1) I have the possibility to ask for a heatsking. Should I do this and start over?
2) Am I good enough to go on and use Conductonaut? What I mean is, with those temps and differentials, will I have an obvious gain in temperatures that will translate into being able to overclock at 40+ with having my differentials be more in line with what I have idling/running (3° max running at 40x) something not as demanding as games on ultra settings?

Thanks again for all your help!

Looking forward for the next suggestions!

 

The temps are awful. If this is what you get after your 3rd repaste, then demand a new heatsink. Could you run Aida64 as the setup guide and post the results? You should try test with Dell's advertised maximum overclock profiles (OC LV 2, 4.3ghz) and (OC LV 3, 4.4ghz), but customized and with your own core voltage undervolt.

 

What do you mean temps are awful? I am still using Kryonaut, not liquid metal.

I'll run AIDA as soon as I get home (should I run it with no performance mode aka stock? or with an overclock?)

 

Up to 100C with 36-40x with Grizzly Kryonaut and proper work done, ain't good enough. Even Dell's stock paste will provide better results if you didn't got a lemon first time from Dell's factory.

 

trust me, i'm using the same paste as you and your temps are awful. from the looks of it, pretty sure you didnt learn much from the previous given vague "advices"

 

@speedsweep where do you live? I ask because ambient temperature makes a BIG difference! If you're comparing with people that live in cold-ass states you'll never achieve their temps...

Just put LM and be done with it. I also spent hours tinkering with the heatsinks and could never manage to NOT have core diferentials with Kryonaut. Temps weren't so good either. Not even close to the temps some users with Kryonaut post here.

I would say:
-Get Conductonaut LM on the CPU and GPU
-Check again every pad for contact. Don't go too thick, I get it's easier to assure contact that way but CPU and GPU temps will suffer. Use @iunlock 's guide as a starting point, not a bible!
-Download and flash the latest BIOS for updated fan curves and flash the 1080 with the old vBIOS supplied by @Zer0K somewhere around here a few months ago. The performance decrease is negligible. Specially when it's backed up by much lower temps!
-Install Throttlestop and get a nice undervolt going. @Pete Light was very successfull in getting his 7820HK and GTX1080 jamming at higher frequencies with lower voltages. Less temps - more performance!

Do those steps and you'll have a beast of a laptop! Just make sure the pads are making good contact. That will prevent future issues. And the LM will 99% for sure get rid of the core dif issues.

Good luck!

 

@alexnvidia What do your temps look like with Kryonaut? I read your thread, so I know that you spend quite some time repading/repasting and achieved nothing. So I suppose you got a new heatsink that solved the issue? If yes, which one did you get?

@Shark00n :
- I live in France, and I am in my office which is a constant 20° at that time.
- I checked everypad again during the 3rd try : everything seems nice and making contact without being too thick
- BIOS is flashed at latest version, I will have a look at the old vBIOS but I am pretty sure this will only improve marginally.
Did you really got rid of the core differential & temps with LM?

 

Yeah LM fixed all core dif issues. It's conductive enough that the differences in contact don't matter anymore. vBIOS will improve it further, believe me.

 

If the TRIPOD heatsink is uneven or warped (seems as it is, because the temp is awful even with Kryonaut and good prep work).... Liquid metal may become a disaster. A ticking bomb.

 

i achieved nothing on cpu core differential, at least not to the point im satisfied or happy with. my max cpu temp at 4.1GHz OCCT PL1&2 92W is 76C with a core differential of 10C, meaning my lowest core is 66C. this is a far cry from my original factory out of the box over 95C overheating and throttling under the same condition and room temperature which is 23-25C. i did extensive work on the heatsink, bending and adjusting and checking so many times i lost count. going after any particular heatsink is a lost cause too. they all have the same problem due to the weak top cpu tripod arm. Only LM can easily solve the cpu core temp differential problem provided your base line temp before LM is good. but i'm not willing to go down that road.

my GPU is fantastic though, witcher 3 maxed out settings, gpu temp fluctuate between 58-60C. and this is at 1.9+GHz OC and 180W OC vBIOS

 

Bah! We’re not talking milimeters here. It makes good enough contact!
I’ve had LM on mine for months on the CPU and GPU and it’s been great! Great temps and no core differences!
And my unit takes a beating. It’s always in my backpack and always on the move. No desk top princess here!!

 

What is everyone’s beef with LM? The directions say a pin drop of the stuff on the sink and the chip. If you’re globbing it on there and not masking it off, you deserved it. Otherwise if you apply a pin drop like stated, I could probably shake the hell out of my laptop and it won’t magically pool up and short my laptop out. It wasn’t easy to spread and it’s so thin the amount of pump out is negligible.

If you’ve tried repasting several times and still have bad core differential, I would say it’s worth a shot to get a new heat sink. One thing I did when I repasted mine, was I remounted the motherboard to the laptop with 2 or 3 screws. This ensures the motherboard has zero flex when you go to set the heat sink back on it. It also has a higher chance of sitting flat in my opinion.

Keep us updated!

Matt

 

Quick update :
- OC2 and OC3 profiles gets me bluescreens on startup: impossible to get to windows (I didn't try to chose the profiles and then modify the clocks and PLs, just tried by selecting the profiles and restart
- AIDA stress test result after 28 min with everything stock (should have tried this stress test first.. Now I am worried...) :



And regarding undervolting, either I am doing it all wrong, but going past -115mv at 40x makes my system crash after around 40mins playing.
It's stable at 36x with -120mv and 40x with -100mv.

 

As I said in post #55, temp is awful even with Kryonaut and proper prep work. If you are unsure how to undervolt, read the guide How to Lower Temperatures, Stop Throttling. And do not forget... Not all chips undervolt equal. Some need more, some less voltage for an fixed clockspeed. But most should manage -100mv. And 44x should be manageable, even with unlocked Kaby BGA.

 

I have had the exact opposite experience....I used Conductonaut and while the temps were ok at the start they didn't remain that way and core differentials were not stellar at 6-9c. I swapped to Kryonaut and my differentials dropped to 2-3 and my temps have never been better.

I just run stock clocks, undervolt I believe it at -120 and the original GPU vBios. The OC bios raised temps up about 5-6c for me on the GPU for negligible gains for what I use it for so I went back to the original. I would rather have a cooler GPU then a few extra points in Firestrike

GPU temps after hours of OW are around 60-62 and CPU usually hovers around 65-70.

 

Well in any case I asked Dell for a heatsink swap before doing anything else. And if it's not better I will ask again and again until they bring me one that fits perfectly or I'll ask again for a new machine.
@Papusan I already went through this guide and tried undervolting with either throttlestop or xtu. Frankly I don't see much improvement temp wise...
@MogRules before using conductonaut / kryonaut, how were your temps and core differential looking on stock? Did you repad?

 

I did not re-pad, I didn't have any stuttering and didn't feel the need. As for stock temps I honestly don't remember and to be quite honest I repasted so early on that I didn't really do much testing at stock. I knew even before I got the machine I was going to be changing the stock paste so it was more a matter of just doing it and seeing what the temps were afterwards. Even if the stock paste isn't doing a bad job from factory a good aftermarket paste is going to do a better job.

 

-100mv vcore voltage should gain you 10c lower temp. Test both default vs. undervolt voltage. Then run different benchmarks + already mentioned Aida64 stress test.
Proper benchmark tests CINEBENCH_R15 - wPRIME (Run as adm. Click advanced settings. Select 8 threads. Then run 1024M stability test). Temps shouldn't be that ****y with Kryonaut, undervolt and almost stock clocks.

 

To be sure: for the undervolt, I open TS, set CPU core and CPU cache to -100mV and hit "Turn On" right? That should enable the undervolt?
And btw, I deactivated intel HD graphic card in windows as its not used anyway...

 

Apply, then OK. Don't forget turn on TS in main window. Monitor load voltage with Hwinfo64 when you run tests. Test both default and with undervolt. You should easily see the difference in voltage this way. And run with high performance in Win power plan. Not some nasty own created, OEM created, balanced or power saving power profiles.

Spoiler: Put up Hwinfo64 as the pict

 

There you go.
Again: yesterday's test and today's test were made at stock clock (CPU performance mode disabled) with today's one getting a -100.6mV undervolt in TS (CPU Core and CPU Cache) :



If you compare yesterday's test and today's (with -100mV undervolt) :
- core VID min went down from 0.904v to 0.837v)
- max temps went down from 97/88/98/87 to 92/85/93/85 degrees (5 degrees gained on the hottest core)
- average temps went down from 82/76/82/76 to 73/68/73/68 (9 degrees gained on the hottest core)

I am of course always running with max perfs power plan...

Duration is the same for both tests (around 28mins)

If you look at average temps, yeah well, that doesn't look bad, but this is at stock clock and what really doesn't feel right are the max temps.
To me it shouldn't go higher than 85 degrees, even with an overclock at 4,3Ghz and "traditional paste" (Kryonaut).

Am I wrong?

P.S : Yesterday's test is on the previous page

 

Could you also run mentioned Cinebench R15 and Wprime test? Screenshoot in the end of test(with full load)(stock and 40x all cores).

4.3GHz should be manageable with Kryonaut and good prep work without temp mess as this. If you can manage this with 85C, I don't know. Other with same hardware and laptop should be able to give you answer on this.

 

@Papusan : I'll do those tests this evening, but franky cinebench and wPrime don't put enough load on the chips...Nothing better than a very demanding triple A game like BF1, GTA V, ACO, etc.

 

Yeah, you shouldn't see higher than around max 50C in both bench tests.

 

Yes, 50C if the fans spin up at 2500-3500 rpm. On max fans you should see below 50C.
@speedsweep Look at current temps in HWINFO.
Is the back of the laptop elevated?

 

No the back is not elevated at the moment.
No tricks, I want to have proper temps without artificially having to help it.

Btw, Dell's tech is supposed to come tomorrow with this :

1 FRPY8 ASSY,HTSNK,W/FAN,G3,K,NEW,AW17
1 V613M ASSY,FAN,CPU,65/110W,AW17R4

Let's see what the lottery will get me...

 

I really wanted to keep my mouth shut here but I guess I can't do that anymore.

Liquid metal requires a heatsink with PROPER MOUNTING PRESSURE.
This is CRITICAL.
If the heatsink is badly balanced and improperly warped or mounted, one of two things will happen:

1) temp/core differentials will be even HIGHER with LM paste than with Kryonaut or IC Diamond or Phobya Nanogrease Extreme (if you have such a bad heatsink, RMA It, or get IC diamond for badly fitting heatsinks).

2) if the laptop suffers a substantial heavy BUMP somewhere, that LM will squirt out into conductive balls of doom all over the place and your temps will jump to #prochot. If you did not do proper prep protection (foam dams AND nail polish or foam dams AND kapton tape), it will short out and destroy your laptop. BOTH situations have happened to OTHER users here. One had proper protection so all that happened was super high temps and liquid metal balls scattered all over. Another user had a shorted mainboard.

When you spend $2,000-$4,000 on hardware, don't skimp out on protection when you're using methods that have risk. Exercise static shock protection, use protection against conductive balls of doom blowing up stuff, and don't use LM on bad fitting heatsinks.

I have no beef with LM. I have Conductonaut on both my CPU and GPU and I can run *prime 95 small FFT* with FMA3 (4.3 ghz) or AVX (4.4 ghz) and stay below 90C and do OCCT Power Supply Test for hours. Can you do this on your Alienwares?

 

FRPY8 is the new part number for GTX1080 heatsink assembly. I've had two of those and don't see much of a difference compared to the previous one. It's still a lottery and you still need to work on the heatsink metal.

You don't need V613M at all. That is just a small cpu fan which already come with the main heatsink assembly.

 

I mean Phobya Nanogrease Extreme will work out well for badly fitting heatsinks as well. Maybe second or equal as ICD.