[CPU + GPU Temperatures + Benchmarks] - XPS 15 [9560] Kaby Lake

Check my album: https://www.flickr.com/photos/30149337@N04/albums/72157680615207736
Don't do it! It's bad!! I didn't think it would be till I took off the cooler a year later. It's bad news! I'm amazed my laptop didn't die.

Looks like the pch temp is really really high! Put a thermal pad on it! Check my xlbum https://www.flickr.com/photos/30149337@N04/albums/72157680615207736
The pch chip is right next to the right fan by the battery. Looks like a 2nd cpu basically.

 

Hey

Finally everything was fine with my XPS after all the changes, a good fresh install and good settings for Nvidia GPU & ThrottleStop.

Just a few days ago, nothing changes since everything was fine, I got CPU clock drop down to 0.45Ghz while playing but it doesn't come from temperature or utilization as I have no PL limits getting triggered.
And it's not the usual drop down to 0.8Ghz but 0.45Ghz for a few seconds.

No consistency by the way, had it two days in a row, then nothing for 3 days, and now it just happened again.

Any idea ?

 

Do you have any HWinfo logs?

 

At the moment no, but I'll generate some at my next session.

Except for this ones, which informations could be useful ?

 

Use High performance power plan in Windows power plan(aka you shouldnt see clock drop). Then run Hwinfo64 in the background the whole time. Follow up... All those sensors (performance limits reasons) showed below + clock speed and temp in Hwinfo.

 

Ok this is the ones I was thinking about
Will do next time !

I'm already in high performance mode

 

Ok, it happened again, here is the files, I don't really know what you'll be able to see :

https://ufile.io/n5q5y
https://ufile.io/u9o5z

 

So, on a new batch of logs ( https://ufile.io/benbd) what I've observed is this at the moment the CPU drop down :



The temperature looks good :



PCH Temperature is @ 66C.

What do you think ?

The weird thing is that it's random. Sometimes the problem doesn't appears.

 

So according to my Wemo power controller which measures power consumption, under PRIME95 maximum power consumption, I see a total system draw of 59W. After undervolting to -0.135V I reduced the consumption down to 49W. The GPU though is the killer for power consumption and its quite surprising how much it does use... Running Valley benchmark, I get 84W at stock... After undervolting the GPU down to 0.900v I get a total down to 68W which is quite good I think!

 

Any ideas for my case ?

 

Hey guys, i am tinkering with my XPS 9560 to get it not throttle ( as are we all here ).

For now i have tried almost all the tricks in the book:
1. undervolting cpu and gpu to max stable voltages i could

Spoiler: CPU and GPU undervolts

2. Repasted GPU and CPU with Grizzly Kyronaut + repadded everything under heatpipes.
3. Thermal cooling air solution as in iUnlockers first post (turned copper plate+air flow)

Spoiler: Thermal pads solution

4. Disabled Turbo
5. Set Max pre-rendered frames to 2

But still i get thermal throttling after 10-15 mins in game (rainbow six siege 1080p LOW graphics)
I would underclock my CPU to get better thermals - but it seems like Rainbow6 doest like it (freezes even with 100mhz downclock)

In Throttlestop i can see Limit reasons:
CORE - PL1 PL2
GPU - PL1
RING - EDP OTHER

As i understand the mean "power limit" (and i think its because something still gets too hot).

I also got HWinfo logs (while gaming and throttling) and from there i see that CPU temp reaches 80C (whitch i think is the problem), but i cannot understand why this happens....

HWinfo LOG 1 (Didnt have pads directly on GPU and CPU)
HWinfo LOG 2 (Added pads directly on CPU and GPU)


My next steps will be:
1. Get heatsinks instead of thermal pads (i would think that heat sinks would be better solution)
2. Repaste CPU and GPU (maybe i failed there?)
3. Maybe get rid of copper plate and stick everything directly to back plate (?)

Any user input would be appreciated

 

The PL limit throttle is caused by the VRMs overheating and being throttled (subsequently being able to provide less power to the other components). But your padding solution isn't actually contributing to cooling these VRMs.

You are padding everything directly to the backplate, dumping the heat of the CPU, GPU, SSD, VRMS and VRAM all on a single aluminium plate that is not actively cooled, nor has the surface area to be properly actively cooled. You are going to overheat all your components together then.

 

Hmm actually i am padding everything to the copper plate thingy, but from your reply i understand that even when i would pad everything directly to aluminium back plate it would overheat...
Im still a bit lost , should i pad less chips to backplate maybe ?

Ordered heatsinks too, will apply them when they arrive, but meanwhile any help would be appriciated since as i understand with right padding it should be possible to game without throttling

 

I understand iunlock paid a lot of attention to redirecting airflow towards the added cooling surfaces that takes heat out via the central grille.

 

Actually it takes heat from center bottom grille and sends heat to VRAM-s and sends it out from top (below the screen), i have tried to do the same, but atm i think that i have too many chips connected directly to the copper plate thingy, i would think that padding them directly to aluminium backplate is a better option (and also de-padding few chips for less overall heat to the backplate).

For now i have de-padded CPU and GPU from copper backplate (and it gives me a bit more time without throttling)

EDIT:
Found out today that it still overheats following sensors (they have "yes" in maximum row in HWinfo):
IA: Package-Level RAPL/PBM PL1 (i believe its the same as in Throttlestop "CORE - PL1")
GT: Package-Level RAPL/PBM PL1 (i believe its the same as in Throttlestop "GPU - PL1")
RING: Max VR Voltage, ICCmax, PL4 (this must be Throttlestop RING - EDP other)

 

Take a good look at iunlock's mod. A bit of cooling air is diverted from the main fins to the additional VRM cooling surfaces, then the air exits via the small grille below the display.

 

Hmm, i still dont see difference between air flow compared to mine ( i too take a small portion of air from fans to VRM-s), the difference i see is that i have too many chips padded... Am i missing something ?

 

I understand he shaped the pads near the fins to get more air.

 

Cracked my laptop open again and tried to redo "wind-tunnel" a bit better.

Spoiler: Pics

Also removed some thermal pads from chips and managed to undervolt everything to -0.1650.

So far still throttles in game, but i have noticed that it comes out of throttling alot faster - so this means i am getting somewhere
HWinfo Logs

I think im gonna leave it like that until i get the heatsinks (unless someone has any good ideas)

 

Have you changed the paste? You also aren't undervolting... you need to undervolt. I've observed a 10W difference with undervolting on the CPU... just FYI.

You also may want to do 1 big pad instead of small little ones near the VRM. Who knows what is getting too hot... I don't have any throttling at all when I game...
This is what I did:

Spoiler: PICTURES

 

Hey, i am undervolting everything to 1,65 ( as seen in my first and last post ).
I would think that padding everything wouldnt be such a good thing, but when it is not throttling for you then it seems to work :O (will try it when i get new thermal pads)
Still waiting for my heatsinks to arrive to test them too...

 

Not sure, because the idea is to get the heat out by convection, so there must be room for air flow.

 

Today i finally got my heat spreaders and installed them, all pictures of what i did are in imgur link : https://imgur.com/a/ScrxJEA

My conclusion is that this does NOT help much and it NOT worth the time and hassle.

I will still try to remove copper plate entirely and allow thermal pads to contact back plate directly, maybe it will help a bit, but i doubt it

TLDR: Would not recommend since the gains are not noteworthy, just buy a decent laptop

 

I understand iunlock padded the heat spreaders to the cut part of the copper sheet, and made some effort to redirect more cooling air towards the VRMs.

 

Yeah, thas exactly what i tried to do too.

 

I find it odd that you don't have any improvements. I have tons of logs showing great results from several different 9560s. What is the main issue you are having?

 

Hey, i need to clarify myself here:
I am sure there are minor improvements, but they do not stop throttling while running games (that is the whole reason i tried this mod).
I could run hwinfo to post results here if anyone is interested.


EDIT: i really think that removing the copper plate and trying to do wind tunnel directly in contact with back plate would help overall cooling, will try it tonight.

 

Only way to get rid of throttling is to uninstall Intel DPTF and monitor temperatures by yourself.

 

i wass not aware that i could remove that, will give this a try definetly !

 

Yeah, Look up reddit theres lots of posts about that.

https://www.reddit.com/r/Dell/comments/7byhjq/disabling_intel_dptf_for_good/
https://www.reddit.com/r/Dell/comments/6wlbyx/how_to_prevent_your_xps_15_95509560_from_tdp/

 

Well Sir, this fixed the throttling issue and after 3h of full load - max temp was 78 celsius - that was GPU, all others were lower.

Thank you again !

 

Hi what if I told you my cpu temp increased after applying liquid metal (grizzly conductonaut). Applied decent amount and was able to cover fully(but not like pool of water) and a little bit on heatsink as well. Now cpu is idling 75-85 lol got worse. What could have gone wrong?

 

Sounds like you've either misaligned the heatsink or bent it during removal.

 

Nah i removed the heatsink to look and it was looking ok. The heasink is ok, too. This left me clueless. And when i run prime 95, the temps reachead 102. It never did with noctua paste lol

 

You did something wrong. If you put anything over the SMDs with electrical tape or something similar you may have increased the gap size and the heatsink is no longer making good contact with the CPU (or it was never making good contact to begin with). You may also not have put enough LM on there to make good contact. The degree of difficulty when applying LM goes up by an order of magnitude IMHO, but getting worse temps = you did something wrong. Having said that, I don't recommend LM on laptops for a variety of reasons. My temps were better with Conductonaut than Kryonaut, but it wasn't worth the 2-4C to me.

 

Do u think 1.5mm pad creates so too much gap between the heatsink and vram?
I put 1.5mm pad on the vrams next to cpu under the heatsink that probably increased the gap and doesnt make good contact. Maybe i'll apply my left over noctua on them and see how it will go

 

That very well could be the culprit. Put the stock pads back in and see.

 

Dont have the stock pads. Thermal paste could work? I've seen some people do that

 

Not ordinary thermal paste.... K5-PRO

 

Yes did it. i got jaw dropping 20C cooler . Now the cpu is around 78 when gaming and 77 on the gpu. All the ambients pretty cool as well thanks to the pads. Their temps are hovering around 60-70.

 

Can anyone point where is PCH temperature sensor please?

 

That's the PCH right there....

 

Is it worth padding? Light use my CPU sits at around 36-40C, PCH at 55 - 60, which is like 20C difference.

 

It doesn't appear urgent in any sense. Some of us have tried aiming to make the fans spin less (seem to trigger below fifty), with inconclusive results. Mind there is a hot voltage regulator nearby, near the battery connector on the other side of the metal plate.

 

Yes,

It will help with your fan...but I guess its a doubled edged sword...if you want higher speed and your fans on more often then leave it uncovered...if you want no fan when idle...I suggest padding it... I padded mine. My fans NEVER come on when on battery + the power setting......but overall it does stay cooler. Doesn't produce much heat but it will climb into the 50's and 60's like you said..

 

To add to the info available, some users with eGPU enclosures on the eGPU forums claim they got faster speeds by having it padded. They included benchmarks showing an increase around 25%. But that was the 9570,and I don't think I've seen that with the 9560.

 

Hey guys, I have a Dell Precision 5510 which is similar to the XPS 15 9550, and I have liquid metal my laptop with Thermal Grizzly Conductonaut. My CPU temps when stressed is 81 celsius and I don't seem to get temps anywhere near you guys though it is 10 celsius cooler than stock thermal paste. Any suggestions as to why I am getting such an anomaly?

CPU: Core i7 6700HQ
GPU: Quadro M1000M

I cleaned the heatsink of the old toothpaste stamp with alcohol wipes and applied liquid metal on both surfaces (the silicon die and the copper heatsink), and somehow I am still getting high temps. Am I doing something wrongly?

 

Perhaps you didn't clean enough. I usually don't use alcohol wipes but a combination of alcohol, cloth and cotton swabs to really clean all gunk of until each of my tools doesn't show any grey smear. Also, maybe you used too much or too little liquid metal, preventing proper contact with the copper heatspreader. Finally, your CPU is also a pretty hot CPU with far too high voltages. If you haven't done so already, Skylake CPUs will benefit greatly from undervolting.

 

Did you blow compressed air to clear out the cooling fins? Did you carefully clean the fans (don't use compressed air as overspeeding the fans will kill the bearings).

Did you check if the heatsink is not warped? This is a common issue with the 9550, so may favour a good non-metal paste (liquid metal has little tolerance for warpage).

If you run something like Prime95 for a while, are the cores very close in temperatures (like within 2*C). Larger variance indicates either warped heatsink or suboptimal application.

If your heatsink is warped you may be able to gently sort it out but this is ultra delicate and a few experienced guys have kinked/broken them here.

Even the professional laptop engineers here typically try a few repastes to make sure they are getting optimal coverage.

 

Reply to MrBuzzkill

I think there was a lot of the old paste. I just cleaned it up 5 minutes ago, and redid my liquid metal.

How do i tell if I used too much or too little liquid metal? cus this is what it looks like after taking out the heatsink

https://1drv.ms/u/s!AvIsJ-c9AzQV9DEM-vfPnAqo37eS

https://1drv.ms/u/s!AvIsJ-c9AzQV9DJAr79ttv4oFkG_


Some of it actually spilled out a little on to the electrical tape, hence I think I may have put too much Liquid Metal.

I could undervolt, but when I do anything CPU and GPU intensive, I still hit high temperatures and throttle a lot. for this particular CPU, the lowest stable undervolt for me is -120mV

Reply to pressing

Yes, I cleaned both the fans and the heatsink fins.

The heatsink looked flat to me and all 4 cores are within 1 degree celsius of each other.
The screenshot taken below is the CPU being stressed only, with liquid metal applied

https://1drv.ms/u/s!AvIsJ-c9AzQV9DRt83UwFxMsyIta


I have tried these traditional paste
1) Arctic MX 4, bad for this laptop, have to keep repasting every 2 weeks to maintain temperatures
2) Cooler Master Mastergel Maker Nano, last much much longer, but the laptop still loves to heat up
3) Thermal Grizzly Kryonaut, best non-liquid metal paste I've used so far with the lowest temps across the 3, still thermal throttling, only 1 to 3 celsius cooler than the Mastergel Maker Nano

EDIT: Hmmm, new people can't seem to be able to post images.

Screenshot, when both CPU and GPU are stressed

https://1drv.ms/u/s!AvIsJ-c9AzQV9DVIfjWjX8_OxudS