XPS 15 9550 temperature observations (undervolt + repaste)

The first 2 ambient sensors are about 20-30C hotter than the others. I have just put themal pads over the VRM chokes, they peak at about 79C.

I believe the thermal throttling happens at 78C of the first ambient sensor.

 

78° looks more than fine to me (considering they can withstand much more).

What are your gpu/cpu temps (current) after some minutes of unigine? (asking because from what i know no one ever noticed throttling during unigine even without repasting/undervolting)

 

If you keep it looping it will eventually happen. Also, my ambient room temp is about 25-26C (no AC), so that's something. CPU 73/GPU 75 right now, ambient is under 78, and I'm still throttled. It just needs to hit 78C once it seems and then your machine will be throttled until you restart... So unless Dell increases the temp limit, it looks like you'll have to keep your GPU and CPU under 78C at all times...

 

Cutting the black shielding away from the copper just in the GPU/CPU/VRM pad areas has made a big difference, I think.

 

Oh you went for it?

Any ideas on before and after?

 

I'm trying to figure out how to avoid that first sensor hitting 78C. It hit 78C without the GPU or CPU hitting 78C, which means it's just heating up from the voltage alone and not the heat from the GPU/CPU. Meaning, lowering the CPU/GPU temps aren't going to stop it. I wish I knew exactly which VRM it was so I could pad more.

 

This is a 15 min ungine loop, maybe you can compare it with your graphs and see if you can spot the key difference; ambient sensors are pretty fresh (i can upload my vrms cooling setup when i get home).

 

Wish me luck

 

What were your CPU clocks during that time? Also, did you use CLU? I'm using Gelid GC. Curious your ambient is much higher (but if you have AC going...)

 

CPU multiplier was at 34, yeah i used liquid metal, room temperature is around 22-23°.

 

Hmm... well I'll have to see your padding config. Maybe you're just able to keep the temps so low with CLU that it doesn't heat up the VRM? that means it's not voltage-related heating them up...

edit: my VRM sensor hit 80 and no throttling... so what is the exact cause? If it isn't any temperature, what is it?

 

Power draw usually... hits the limit. My VRM get upto 96C and still don't throttle.

 

Why is mine hitting power draw limits when godless's isn't, though? I have "maximum performance" set in the nvidia control panel. My CPU is generally consuming like 14-18W during the test. And why does it throttle after 10 minutes or so, but not after 5 if it's just power draw? I'm using SST (1-255) set to 64 with TS.

 

This is the vrms/chokes padding config (two 1.5mm thick stacked on the vrms and one 1.5mm thick on top of everything) it doesn't touch the heatsink but it does touch the back cover.

 

GPU temp is too high. You'll see the downclocks from the pascal temperature limits... that's pretty normal. It's gotta be temps thats causing the CPU to throttle. Do you have it undervolted in Intel XTU?

I'm going to redo the pads on the VRM this weekend. Going to use a combination of thermal pads and paste...technically it should help my temps. I'm also planning on making 1 larger pad rather than all these little ones...Yea it might reduce airflow around some of the components but my highest temps right now are the VRM so all other components should be ok at this stage...

 

The GPU temp causes CPU to power limit?

 

This is another 15 min loop adding the graphs that you showed and were missing in my last run (trying to fit them all into the screen) this way you have a fair comparison for the next repaste/repad.

To add to the topic: when i had average gpu temps of 77+ during unigine runs (because my heatsink didn't make good contact with the gpu die) the gpu memory clock shown in the top right corner of the benchmark kept jumping from 3500 to 1500 and back up (and the final score was 20-30% lower than it should have been).

You may check for this behaviour, if it happens it means gpu temp is the issue (78° is high anyways).

 

Lower your Igpu voltage... I undervolt mine by like 0.070 right now with no problems. Seems to help with all around temps...

 

Not that this is worth anything, but my i7-6700HQ does -165mv on the GPU without issue... You may try for more.

 

I'm -75mv on the igpu and -125 on the core/cache.

I looked at the evenness of my heatsink and while the GPU/CPU areas are flat (on a ruler), the pipes off to the side of the GPU look slightly warped. Could be causing higher GPU temps?

 

It's at -50mv. Moved at -90mv some time ago but i didn't notice improvements.
I noticed a slight improvement by setting it to maximum battery life in the power management options but it doesn't matter much anyways, it just takes 20-30 seconds more to throttle in realbench stress test.

I don't think it's that easy to bend the heatsink if you're careful when you remove it; just apply a thin layer of thermal paste and mount the heatsink, remove it again and check if on the heatsink and dies the paste is totally squished (as in 0.01mm thick) or not.

Edited a picture to show what the contact should look like.

 

What is the voltage stock vs undervolted? Do you happen to know?

 

I don't sadly, and it's not apples to apples of course. I was just mentioning that because you may be able to go lower as well.

 

Well in other news... I can easily play BF1 at 60fps on Medium preset no problem... That's promising! It was actually pretty butter smooth. I was quite impressed. I should nearly get 60fps on high preset but med is good for now.

 

They make even contact- it's that one of the pipes out from the heatsink is warped up by maybe 2-3mm.

the paste is a bit thicker than that- but I really did use only a little. Like, I'm wasting a lot of paste by squeezing too much out. Your .5mm pads on the VRAM probably allow a better seal than the stock ones. I'll squeeze them as much as I can and try a really thin spread of GC. I feel like I've done everything I can to keep the temps down, I think. The temp just slowly creeps up until it hits 78 after a while

 

I'm using k5 pro paste now to replace thermal pads on the vram modules, by using it i noticed that even 0.5mm thick pads on some modules prevented the heatsink from mating perfectly with the gpu.
Each [xxx] represents a module:

[>1mm gap] [<0.1mm gap]

[>1mm gap]

[<0.1mm gap]

Placing basically any kind of pad on the <0.1mm gap modules will result in mating issues (maybe with the exception of very squishy silicone based pads that i didn't have to test).

How much time does it take to reach 78°?

edit: since the forum breaks the formatting of my incredible scheme here's an image.

 

Notice how I did my thermal pads...well more like paste... the 2 closest to the heatpipes were so difficult... I used the original pad but added thermal paste between the 2 surfaces... the far right one had half thermal paste on it because of how uneven it was...the rest of the VRAM chip had the thermal pad on it... The small gap VRAM had just paste... That seemed to be the best solution at the time...still is IMHO. Those VRM pads are getting replaced this weekend...

 

Are you sure those mounts go into the palm rest? i think they go into the motherboard. I had the same issue, i noticed it when a tech came to replace fans so i contacted dell they asked to send it back to them and i just got it back today and this is what they replaced:
Fans, cables, heatsink, battery... I didn't need any of those things and my laptop was working perfect, only issue was those two screws not screwed in and they are so ****ing dumb over there who ever worked on this laptop. How could they not fix this? They saw it when they did the heatsink.. wtf? Dell keeps dropping my calls they're not giving me an answer and my warranty expired feb 26 but i sent the laptop in on feb 22.

 

Sorry my bad, of course they are or I wouldn't have taken my board out with the heatsink still attached

Same fix applies though, you would need a washer and thin threaded machine screw or something along the lines of a thin thumb nut or some sort of rivet nut to allow you to screw the heatsink down and have the nut on the other side of the board.

 

I think they're actually warped at both sides, because the MB level where the chips are is lower than the level where the heatpipes end. Though on one side the warp is smooth because there is enough distance in between, while on the other the chip is near the fan so the heatpipe is quite bent just above one vram which usually has poor contact. And the alignment indeed doesn't seem good. I wouldn't use liquid metal before testing with tiny grains of normal paste would confirm perfect alignment.

 

I recall folks have complained about too hot bottom plate just by using quality pads from the VRMs to the case bottom. I guess a thicker copper strip might help get heat away from the VRM area towards the fan exhaust area, considering that there is no space for an extra heatpipe.

 

I hit 78 after about 15 minutes of OW with fps at 120.

If you test fit paste for the VRAM do you really need to repaste it like a CPU? I have HDT paste, mx-4, and gelid. Which do you think I should try instead of the pads?

 

I had been thinking about that as well, yes it would work as the heat will be drawn to the cooler exhaust. I would probably do it to both exhausts if it comes to copper plate only. Stick some mylar on to prevent shorts and it could work well.

 

But it doesn't work well enough...

 

I replaced the vram pads that were squeezed tight by the heatsink with thermal paste, and my GPU hit 78C after just a few minutes of Heaven. However, it took 10 minutes to throttle.

edit: added a bit more paste to VRAM #1 and #4. Hasn't throttled after 10 minutes now, GPU temp holds steady at 78C but not over.

Also, it seems that Bios 1.0.3 has slightly higher fan RPMs than 1.1.3. they're 50xx rpm with 1.0.3 and went into high 4k's with 1.1.3 unless I'm not remembering properly.

 

He was using like 14 W/mK pads on the chokes.

As for vrms temps did anyone show a test where the ambient sensor reaching a certain temp caused throttling?

The highest i've ever seen reaching any of my ambient sensors is 85° (and much less on average) but maybe i'm running a different test.

Well i used a paste made to replace thermal pads (it's much thicker than normal paste and probably has lower thermal conductivity).

If you test fit it with some kind of thermal pads (very squishy ones) and the contact area on the gpu satisfies you there's no need to use something else.

Also, if you don't use an external monitor (at 120hz), capping fps at 120 does nothing over capping them at 70-80.

 

I am using an external monitor

 

I was hitting the 90's on aida64 FPU test so I had another go at the heatsink, I noticed 2 tabs that help it locate on the CPU side. bending those tabs up a little has let the heatsink mate better with the CPU and temps are down 10 degrees. Fans have gone from 5000rpm to 3700rpm.

I also decided removing part of the copper heatspreader form the base will do naff all as it is directly above the CPU so will only end up drawing in very hot air. No doubt why it has been covered.

I see no realistic way of trying to cool the VRM's with copper plates as the clearance is so small it isn't worth fighting with it.

 

Yea. Ideally, you're starting to see laptops, especially gaming laptops with VRM cooling...or at least a plate covering the VRM attached to the main heatsink... I agree that there just isn't enough room... Our best bet is to cool it via the case then possibly use a thermal pad to connect the case with the heatpipes...hopefully to carry away at least a little heat from that area...but its more going to just heat everything up instead. lol

I'm gonna pick up a cheap metal mesh desk organizer for papers and mod it into my cooling pad. I've got a 5 channel desktop fan controller and a bunch of fans... I've already tested the controller with a cheap AC C power brick and it works great. Should turn into a nice custom cooling pad. I tested it with just 1 fan blowing on the bottom middle area, VRM didn't go past 80C so that's a 15C improvement!!!

 

When you have a chance could you indicate those 2 tabs on a photo (you could just steal a photo on the last few pages and indicate that way.

Thanks for taking the time and effort to look at those copper plates on the center of the bottom case and for measuring the clearance of the vrm to the bottom case.

You should just drill some holes in the bottom case below the VRM!

 

I would ideally like to find somewhere to cut out slots to match the ones already on the base.

And as for the tags, the left 2 mounts for the cpu cooler also have a fixed locator under the spring, it is those. Can't miss them as it only has 2.

If I keep it tbh, getting a bit pissed they have still not fixed the redshift issue and the build quality is still a lottery.

 

Vanity Fair!!!!

Just pull out that construction impact drill in your garage and have a go at it...

Extensive and consistent history is a good inicator of the future...

 

GoNz0: if you can wait until May, you could roll the dice with Lenovo 720...

 

Looks promising!

 

Well my work is done...wow!

Check it out! So another good mod is adding electrical tape to allow the air to actually flow through the fins of the heatsink... Not sure if it was my total overhaul of the VRM cooling but my temps are WAY lower and fan speed is low too.





VRM are a heck of alot cooler and I can't wait to check out gaming temps! Check out my flickr album for more photos. Don't want to bombard the forum...
https://www.flickr.com/photos/30149337@N04/with/33242850065/

 

Awesome hack!

 

Very interesting. How does this affect the VRM and ambient temperatures? Does this by chance work by leaking some cool air from the vent exhaust before the fins (due to more resistance) towards the VRM area?

 

It does leak some air. However, the air from the blower is BEFORE the heatsink fins... you'll see a gap after the fan housing...don't cover that. I didn't...

Just ram Prime95...max VRM temp on the CPU is a crazy 51C!!!! Definitely fixed all my VRM issues! DANG! Even ran BF1 at 1866MHz on the GPU! Pretty freaking awesome!

Looks at these VRM temps! WOW!!! I used thermal paste between the pads and the VRM... guess it helped along with the news pads...


Who knew prime95 was such an easy test for this laptop..Maybe Kaby Lake is really optimized for computing prime numbers???

 

Tape over the vent? you do realise you will have to replace that VERY often as the dirt will stick very fast. Probably not worth the effort considering it has pads each side of the fan to direct the air.

And why cool the wifi card?

 

Is this mod to put a piece of ~5cm electrical tape on top of each fan from the beginning of the heatpipe to the back black grille?

Is the idea to:

1. Send as much hot air from the radiators out the back black grille and ensure none of that hot air circulates inside the laptop

2. Increase cool air blown inside the case (by not taping over the gap between the fan housing and the heatpipe?

I have seen people using the 3m super 33_ tape around here for liquid metal. Is this 105*c tape what you are using or did you find a higher temp spec tape:

http://solutions.3m.com/wps/portal/...l-Electrical-Tape?N=5432987+3294355633&rt=rud