XPS 15 9550 temperature observations (undervolt + repaste)

Hey! I'm looking to buy the pads for my 9550 and don t know which size i should take. I just re read the whole topic without having a clear answer: pour the heatsink pad, which thickness is fine? 1mm or more?
can i put the same pad everywhere (on the SSD, RAM and maybe on the chips if they don t make contat with the heatsink)?

Thanks a lot!

 

The problem is there are so many places to replace the thermal pads. Here is what I used:
For most of the ram for the GPU I was able to use .5mm but had to use 1mm thick on the section above the heatpipe (these vary widely depending on how bent your heatsink is from the factory)
From the heatpipes to the bottom cover I used 1.5mm
From the SSD to the bottom cover I used 2mm

The stock thermal pad on the vram was ~1mm thick to begin with I believe.

 

Very nice post! It is interesting to see that I am not the only one with the ambient sensor reaching more than 80°C in Prime. Although your idle temps are lower than mine, my load temperature is a bit lower. (your idle core temps of 22°C sound a bit too overoptimistic to be true, maybe the sensor reading is a bit off in that region...)

 

@ custom90gt

Thanks a lot, i ll do the same as you did

 

You could also use 2mm for the heat pipes and 1mm for the GPU ram. Thermal pads are pliable and can mold into shape (within limits of course).

 

This looks like an awesome way to cool it down! i will definitely try and repaste it! But i'm not familiar with overclocking or undevolting it....any guide i can follow to undervolt the 9350 ? with the 6560u iris 540~

Thank you so much guys

 

Just for newbies.

1) Does undervolting have any effect on performance? I guess that if you undervolt you should cool things down but lose some processing power, is this right?
2) Isnt the processor supposed to change its voltage according to what's needed, by default? Is this not properly handled? I cannot find any reason why people are undervolting their processors.
3) How easy / hard is it to repaste? I don't wanna mess up my system but would love to make the fans shut up a little

 

1. Undervolting has no possible negative effect on performance, but may increase performance due to lower thermals power consumption
2. It does, but it's set up for lowest common denominators. Most chips are capable of running with an undervolt of -50mv or more. The skylake series seems to be able to do -100 to 150mv. An undervolt is relative, it is an offset-not an absolute value.
3. You will need a decent paste, T5 + Phillips screwdrivers, alcohol, and cotton swabs. Unscrew the 6 screws holding down the heat spreader. Clean the paste off the spread and the CPU using the swabs and alcohol. Make sure no cotton strands remain. Use a tiny line of paste across the CPU/GPU, about the size of some long-grain rice.

pictures on what a repaste should look like can be found here: http://www.ultrabookreview.com/10395-dell-xps-13-9550-iris-540/

note that once you apply paste and put the heat sink back on, if you remove the heat sink you'll have to repaste, as air bubbles will have been created.

 

Great, @Eason. Thanks for your descriptions.

 

I have used the thermal pads from the SSD's and stuck them on the heatsink to see how good it is and if I agree with the heat on the baseplate. So far I am hitting 70 max but the ram sensors still going to 83, god knows where that one sits!

 

Progress, The DIMM sensor is not the DIMM sensor at all. I think it shows correct in the BIOS as it seems low but Aida64 reports the sensor on the edge of the board shown here

HWiNFo64 shows the right sensor.



Amazing what you can find with a q-tip with a drop of liquid air as it dropped the sensor from 55 degrees to 10 degrees in an instant!

 

And I am removing the thermal pads off the back of the heatsink as the whole laptop is getting warm, it isn't worth it.

 

That is a very nice find. Thank you! So it seems that this sensor corresponds to the voltage regulation for the CPU/GPU. Makes sense, these things are usually cooled on GPUs.

 

So, what does this censor do, again? Would cleaning it with alcohol make a difference, or it requires some special applicant?

 

The voltage regulators around the sensor supply the CPU/GPU with power. In that transformation process, they produce heat. That sensor simply measures the temperature in that region. Cleaning it won't help, but you can put passive coolers on the regulators.

Gonzo used liquid N2 to locate the sensor on the board. He observed the temperature in bios while cooling it and then concluded it must be that device on the board.

 

Ok. Was curious if that was causing my firmware-level CPU throttling after a few minutes of gaming. It isn't PROCHOT, temps are fine, and it isn't TDP-- it's something in the firmware. It can be overridden with TS, but I'd rather not have to.

 

I use AIDA64 and it bugged the hell out of me as I knew DDR4 wouldn't get that hot. I didn't test it in the BIOS as that seems to read correct, I knew that sensor was 10 degrees above everything else at the time.

Maybe the AIDA64 dev can get the right sensors sorted out, I am waiting for his reply after providing sensor dumps. I got the 9530 working via dumps last year so fingers crossed.

 

They should pay you for that. Nice work!

By the way: You said that the BIOS reading is correct (for DIMM I assume)? For me, that is already enough to know.

 

I repasted my XPS L502x and 9550 with MX-4. Got a little better result than Arctic Silver 5.

 

Aida64 has been updated, the latest BETA now has corrected readouts, temp4 is not the DIMM from what I can gather via HWiNFO64

 

Does anyone know how low the voltage should be for an Intel Core i7-6700HQ? I tried using OP's post as reference, but he has an i5.

By the way, if the Throttlestop icon is green in the taskbar, does that mean it's currently working?

 

Every CPU is different. My i5 ran at -170 without issue but my i7 runs at -155. It's really up to you to test it, some run better with lower voltages than others.

 

Im running -150mv on CPU and -100mv on iGPU.
160+ were unstable for me (crashes during games, prime etc actually worked fine)

 

just curious, but how long did you guys leave furmark on to get those temps? When I did it, the temperature just kept rising, but would stabilize at certain temps for a few minutes. I left it on for about 10 minutes and it eventually went up to 87. I'm pretty sure if I left it on longer it would have risen even more. Is this normal? Cause my CPU temps are pretty great. running prime95 and my temps seem to stabilize around 52C, and the fans are barely on. Although when I play BF4, GPU seems to average around 72C and doesn't seem to go above 80 (This is with the frame limiter off. I usually put a frame limiter on to about 60fps, in which case the temps would usually average around low 60s), so I'm not really sure if my temps are ok or not. I repasted with Arctic silver 5 btw. Also, is it worth putting thermal pads on the heatpipe? How hot does it make the rest of the computer compared to before? Cause it seems to make quite a drop in temperature.

 

Not long as furmarks certain to damage things if left running as it is not designed for laptops.

 

I seem to have ended up with a pretty good i7 / 512gb SSD / QHD version, generally. A month in and really no trouble I noticed (slight "squeeky" sound if I hit the right side end of the spacebar ... but not too bothersome). Only thing that concerns me a bit is that I feel it idles at a too high temperature. At the moment it's at 43-47 degrees Celsius (Package Temperature value in Intel Extreme Tuning software). Occasionally a bit higher, occasionally a bit lower, but seldom below 35 degrees, in the morning / cold days. Undervolted CPU by -150mv / Graphic Card by -130mv. And it's on my DIY still passie cooling pad, which lifts the laptop up vertically, giving it extra breething space below.

Interestingly even when taxing the system it doesn't go way hot, though. About 78-80 degrees when colour grading with Davinci Resolve / playing games. Guess a re-paste is in order sometime?

A bit of a 'noob' style question, but the last 7 or 8 years or so I've used linux only and am out of the game - what's good software to use, currently, on windows to dig a bit deeper and get more detailed readings?

 

Hey!

So I just did the repaste and added a couple pads. But I am having an issue: I tried to benchark the laptop to see if it was stable and my paste well applied. It worked on the CPU bench but the GPU made the XPS reboot by itself.
So I'm wondering if it's my fault (the temps were correct before it rebooted) or the fault of the nvidoa drivers (it made me reboot when i was playing a game before).
Do I have a way to verifiy that point?

Thanks

 

I found this thread to be very helpful for me so here are my results. I have a Dell XPS 15 9550 (2016 model), i7 6700HQ, 16GB RAM, 512 SSD PCIe, 84WHr battery with the 1080p matte display. I've only undervolted the CPU/cache. Have not touched the integrated GPU or the nVidia GPU.

CPU
- Test: Prime95 (Inplace Large FFTs - most power usage)
- Undervolt voltage: 170mv (might be able to go further, but this seems stable)
- Thermal Paste: Arctic MX-4

GPU (nVidia)
- Test: 3DMark Fire Strike
- Thermal Paste: Arctic MX-4
- Thermal Pad: Fujipoly Utlra Extreme XR-m 17.0W/mK 100x15x1.5mm

During Prime95 Test (30+ minutes of testing for each; 71F/22C ambient room temp; laptop sitting flat on a wooden desk; power plugged in; screen vertically opened)
- Original temp: Low to mid 80C
- Undervolt only temps: Mid to higher 70C
- Undervolt + repaste/repad: High 60C to low 70C
- Original speed: 3.1GHz to start but quickly ramped down to 2.6GHz
- Undervolt only speed: Bounced between 3.1GHz and 2.6GHz; about 50/50 split between the two speeds
- Undervolt + repaste/repad: Stable at 3.1GHz entire test

I also ran 3DMark Fire Strike (not extreme or ultra) for several hours to make sure the graphics were stable too. GPU (nVidia) temps were in the mid 80C and CPU temps were in the low to mid 60C. Doing an official run, I was able to get just under 4000 for a score.

I used Intel Extreme Tuning Utility (XTU) but I'm probably going to try ThrottleStop as it is supposed to have less overhead than XTU. I also found that XTU's stress test wasn't as stressful to the CPU as Prime95 was. I did run the XTU stress test for 10 minutes, and everything was fine. I don't remember the temps. They were much lower than when I tried with Prime95 though.

Pulled the back off the laptop (screws around the edges and in the middle under the XPS cover). Then there are 2 tabs on either side to contend with. Just need to be really gentle to get the back off. Once the back was off, it was quite easy. Just unscrew the heat pipe (6 screws I believe) and the whole thing comes out.

I cleaned off the GPU and CPU as well as the corresponding spots on the heat pipe. Also removed the thermal pads that were over the GPU memory chips (for me they were stuck to the heat pipe). I kept the thermal pad that was being used for the SSD.

Applied Arctic MX-4 using the dot method on the GPU since it's square and the line method on the CPU since it's a rectangle. Cut up the Fujipoly thermal pads to fit over the GPU memory chips (they are the 4 black chips surrounding the GPU). Reinstalled the heat pipe. I had some thermal pad left over. I removed the stickers that were on the heat pipe above the CPU. I cut the remaining thermal pad in half and put one on top of the heat pipe above the CPU and the other on top of the heat pipe above the GPU (so these thermal pads are between the heat pipe and the back cover). I haven't had any issues with the temps on the back cover yet. Seem fine (warm of course) to me.

About the fans. Before any upgrades, the fans were on all the time and spun up to audible speeds every once in a while. After installing all the updates (Windows and BIOS) and repasting/undervolting, the fans are used much less. In fact, just normal browsing and idle times the fans aren't on at all. They only come on at the lowest setting. It's not until I do some gaming or benchmarking that the fans turn on high.

On a full charge, just screwing around on the web and the brightness at about 50%, Windows reports that I should get about 10.5 hours of battery life. Not sure how true that is, but it's encouraging!

 

Check the event viewer, use WhoCrashed to read crash dumps and get something logging temps.

 

Hello,

I received my XPS 15 i7/512/16ram/4k screen about a week ago and it seems to be a bit warm. It is around 45-48C with light usage in ETU. I have read through this thread and it seems that repasting, repadding, and undervolting helps reduce temps. I do not have the expertise nor trust myself to do these repairs. I have found a reputable repair shop here locally that is capable of making these changes.

So I have a few questions:

1. What parts should be padded/repadded? (I see SSD, GPU, and CPU)
2. Best pads and size? I saw a Fujipoly Ultra Extreme recommendation.
3. What parts do you recommend to be repasted? (GPU, CPU, Heatpipe - anything else?)
4. Best paste? I saw Coollabratory Liquid Ultra praised.

I appreciate any help and recommendations.

 

I would not pay a shop to add the extra pads as my opinion and observation as a laptop engineer was a laptop that was becoming uncomfortably warm once the heat had dispersed throughout the case to the extent the fans were on more than without pads, it may help under load but at idle it had the opposite effect. the baseplate was so hot it was causing the air being drawn in to be so hot it couldn't effectively remove the heat so the fans stayed on more than without. Now I have removed the pads I have a cool laptop again, even raising the back up caused issues as the rubber strip is in place to stop exhaust heat being drawn back in.

It is a laptop, you have a tradeoff of higher idle temps to have a quiet laptop, a PC has more space to get rid of the heat and tends to have a fan on all the time as it isn't sat 2 feet from your ear.

Do not waste money paying someone to repaste it if you are not capable, it is a really easy job when you read the freely available service manual.

I also suggest a low pressure paste such as mx2 or mx4

 

Thanks for your reply and info on the pads. Regarding pasting what components would you suggest repasting?

Also, what is the difference between mx2 and mx4?

 

I was playing with the XPS a little more today and found my idle CPU temps are around mid 20C (25C - 27C). Again, 71F/22C ambient temp and sitting on a flat desk. Also, with the "Balanced" power profile selected, I was seeing anything from 10.5 hours to 13 hours of battery life with a nearly full charge. When I put it in power saving mode, it increased to 16 - 19 hours. Again, that's Windows' estimate and to be taken with a grain of salt. But I figured if people need numbers to work with, that's what I'm seeing.

 

about 0.5 degrees c

You only have the CPU & GPU to repaste

 

How does " Gelid Solutions GC-Extreme" compare?

 

about the same, I have some on now. it's only a 45w CPU and that stuffs designed to go to 300w+

Sitting around 38c surfing.

 

After some more gaming, I found the bottom of my XPS to get really hot. So I reran the benchmarks and took a temp reading. With Prime95 running, I was getting around 115F/46C on the bottom of the case over where the CPU is. I figure this must have been because of the thermal pads I put there. So, opened the case, removed the thermal pads that were between the heat pipe and the case, and replaced the bottom. Again with Prime95. Waited 10 minutes or so. Found the temps to be around 105F/40C. Much better. In real life gaming (used Poker Night 2 as a test), I am getting about 95F/35C for temps on the bottom of the case. Much better. How did this affect my CPU temps? Idle is about 2C - 4C higher and with Prime95 running they are in the low to mid 70C now. So a bit higher, but still not bad.

 

@ GoNz0

Thanks a lot!
After few researches, i think my nvidia drivers are not good.
Which version do you use? I tried the lattest ones and I m currently on 364.51, which still does crash after 3 or 4 minutes of bench...

 

I'm just wondering what's the standard temperature when doing nothing on your XPS 9550. I mean you are standing still , watching the desktop, just the startup apps in the background, what's your package temperature (I guess you have to have the intel XTU open to check this though). Mine is around 45 C, something which still sounded like a lot. If I hapen to open a browser it suddenly goes to more than 55C, then goes back, and goes up and down with a decent variability.

Is this normal? What about you?

 

I'm on this week's release. What benchmark do you use and is it just a driver crash?

High 30's low 40's.

People need to remember the bios is juggling heat and noise so results will vary depending on reading from other sensors. Under 50 is acceptable

Sent from my SM-G920F using Tapatalk

 

I ran the 3D occt bench.
Yes it seems to be the driver which makes it crash.
It happenend with a videogame too.

 

Hi, while running XTU grahpics stress test the maximum processor cache and processor graphics frequency is without a doubt not what XTU is reporting. XTU is reporting cache frequency at 137,097.12 GHz and 43,621,810 MHz at the graphics processor. Is this normal to get numbers as high as these? Anyone know if this is a software issue related to XTU?

 

1. Is the 9550 GPU vram cooling pad issue limited or a problem with a high % of 9550s?

2. Can you eyeball the gap easily or need a camera? I have a tough time seeing it with the above photo zoomed 400%, but it is obvious.

3. Does the problem seem to be the same chip for everyone (e.g. next to the left fan towards the back of the laptop - the one directly below the heatpipe)

4. How can you be sure all 4 pads are contacting the heatsink? Just put some extra thick ones on?

5. Custom90gt mentioned, "From the heatpipes to the bottom cover I used 1.5mm."
a. Does that mean you used case bottom to push the heatpipe into the vram pad?
b. Does that mean you used the case bottom as an additional heatsink?

5. I see some people used heatpipe-pads-casebottom to make the case a more effective heatsink. But some have removed the pads (superdavemaine & Gonz0). So maybe limited pad might be a compromise.

Thanks and apologies for the detailed basic questions.

 

Lots of questions so I'll post below in the quote...

 

Thanks for the detailed response custom90gt.

Knowing that all 8 (!) of the 9550s you owned had defective GPU VRAM cooling makes this inspection (and likely repair) an easy no brainer.

Based on several comments and photos on the site, and my experience with every laptop I have owned, I will assume that the CPU & GPU will need repasting.

So I will just pick up all the necessary parts and in one swoop: add new SSDs, repaste the CPU/GPU, and fix the VRAM cooling.

 

Sounds like a good plan. I wouldn't bother trying to re-bend it, just put appropriate thermal pads on it. I recommend getting different thickness thermal pad (I have one that has .5mm - 2mm

 

My XPS15 will arrive Thursday. What's the recommended procedure and parts? I have a little bit of pad left over from a previous mod, but it probably won't be enough.

Thanks.

 

Hi custom90gt,

I saw your posts with thermal mods and they've been very helpful. I have a large Arctic 1.5mm pad (6w/mk) for heatsink-case and a small Fujipoly Ultra Extreme 0.5mm pad (17w/mk) for chips-heatsink, and I'm wondering if it's utterly necessary for me to get 1.0mm pads because Fujipoly's real expensive. Is it possible to do 0.5mm for 3 VRAM chips and 1.5mm instead of 1mm for the loner VRAM and still get perfect contact? (all chips in contact without excessive compression of any thermal pad)

Thanks in advance!

 

Nothing stopping you doubling up a 0,5mm pad, just do your best to get a good contact laying it from one side to push air out.

 

Like Gonz0 said, you can stack them. The key is really just mounting it (without cpu/gpu paste) and then taking it back off to see if the thermal pads made contact (they should squish down a little bit)