XPS 15 9550 temperature observations (undervolt + repaste)

I cant seem to get xtu settings to stick after reboot. Even the profiles I save are completely cleared when I restart. Like I have 2 profiles aside from the default, and they show up in the list, but there's no changes from stock and it doesnt give me the option to apply them. Any ideas?

 

Mine took about 2 minutes plus on boot for xtu to apply my profile. So ya, xtu will auto apply settings on every boot, just have to give it some time...

Another thing I may have noted is, if you open xtu app before it managed to auto apply settings, then it will no longer auto apply...i hope that make sense here...

 

Use something like this:
https://www.reddit.com/r/Surface/comments/3xtq81/my_own_variation_of_xtu_undervolting_tasks_and/

 

It's overheat man my laptop is only 55 to 60 C

 

I recently replaced the wireless card in my XPS 15 with an Intel 7265 (highly recommended) and took the opportunity to try and repaste the cpu and gpu. I bought a tube of Gelid Extreme and used it on my i7-6700HQ. I had already undervolted by 150mv in intel XTU.

I tested the difference via running small FFTs in Prime95 for 15 mins and viewing the results in HWinfo64. The results were very positive.

This first image is with the stock paste and a 150mv undervolt. As you can see, the average temperature was 77.7C, with the average CPU frequency at 2.6GHz. Notice that there was thermal throttling and the CPU was having to clock itself down many times during the test to try and keep thermals in check. The CPU was only able to clock all 4 cores at the maximum 3.1GHz boost for small amounts of time.


The story was very different with the repaste. After applying the Gelid Extreme paste to the cpu and gpu and repeating the test the temperatures averaged 72.0C. More impressively, the CPU never throttled and maintained the max boost of 3.1GHz for the entire 15 min duration.


The GPU temperature didn't seem to change after the repaste. Another poster seemed to indicate that the gpu cooler wasn't making great contact with the gpu die itself, maybe the is the issue but I am unsure. I may investigate further in the future.

In summary, repasting the CPU is definitely worth it.

 

Seems like your first run, you have voltage a bit higher of 0.02v, maybe that also add 2-3 degrees to the temp?

What is the exact type of screwdrivers I need to open up the back panel and the heat sinks? Tiny parts they are and looks easily stripped

 

From the Service Manual:

Recommended tools

The procedures in this document may require the following tools:​

• Philips screwdriver
  
• Flat-head screwdriver
  
• Torx #5 (T5) screwdriver
  
• Plastic scribe

I found it almost impossible to not snap off the plastic retainers...

 

Hi, have you noticed any difference between the original wireless card and the Intel 7265?
By the way what did you do with the 3rd cable for antenna?

I'm awaiting my 9550 and considering changing the wireless card for an Intel 8260 (which is the wireless card to be found in the Precision 5510).

 

I don't wish to hijack this thread topic, but the differences I noticed between this Intel and the stock Dell card are slower wireless speeds in exchange for much better bluetooth stability. Since I am a heavy user of bluetooth mice, it was an easy choice for me. Range is about the same.

The third antenna is just laying on the card under the metal tab where it would go if there were a third antenna jack. I figure I will pick up a 3x3 Intel card when/if one is released.

 

I am planning on doing the repaste and was curious what thermal pads you are using? I was thinking something like this http://www.amazon.com/Fujipoly-mod-...qid=1452559198&sr=8-3&keywords=fujipoly++11.0 but I'm not sure how much of this stuff I need. I was gonna get 0.5mm and 1.0mm pads with the 11.0 w/mk rating (the 17.0 stuff seems overkill, and I don't know if the 6.0 stuff is good enough). It seems to either come in 60mmx50mm sheets, or 15mm wide strips. Is 15mm wide enough to cover the memory chips? It's hard to tell how long/wide the thermal pads are that you put on there.

 

I measured the stock thermal pads and they are 12.5x14.5 mm... The stuff I'm using is the 11w/mk. The blue pad I have that is 1mm is only 3w/mk but it shouldn't pose an issue. Anything is better than nothing in this case.

 

Awesome, thanks! Also, could you help clarify the bending process for getting the heatsink to sit a little flatter? You make it sound like you have to bend it to the point that it almost breaks. Is it necessary to bend it, or would repasting and reseating it be sufficient? Is there an easy/safe way to flex it a little to straighten it out?

 

The bending process is only difficult because I had to bend the heatpipes a tiny bit on my heatsink. The problem is they are amazingly easy to bend and pinch shut. The real trick is just taking your time and figuring out what needs to go where before you bend anything. In my case repasting and reseating wasn't enough for the last ram chip, but I could have just thrown a thicker thermal pad under it. Another thing you can try to do is use a flat surface to try to straighten the whole heatsink out on at once. It's not really hard, just don't want anyone to bend the heatpipe too much (did that years ago myself)...

 

Great. Sorry, but I had one last question. I have very little experience with thermal pads and I was wondering, is it ok to layer two of them? For instance if 1mm isn't sufficient for the last memory chip, would it be ok to throw a .5mm on top of it to ensure good contact with the heat pipe? I realize it isn't ideal since air gaps and such will probably reduce the effectiveness of the pads, but like you said before, anything is better than the complete lack of contact right now. I just don't want to have to buy a bunch of different thickness pads.

 

Don't be sorry. So you're not supposed to layer thermal pads, but I don't foresee it as being a huge detriment to cool ram in this scenario. Not ideal, but it's not going to hurt anything.

 

I did a stress test today and saw the DIMM jump to 75c
Tempted to try and put a pad on those to the case.
I am not sure how to go about it but I will probably mask off the DIMM's and put some soft play dough on it so I can see what thickness is needed.

Gelid do thermal pads but only in 1 and 0.5mm

 

It's amazing you're getting such measureable results from just a voltage tweak.

 

So I decided to pop off the cover plate last night just to peek around as I wait for some thermal paste and thermal pads to come in. I broke both retaining clips on the sides immediately. This was being practically as careful as I could be, using guitar pick style prying tools from the ifixit protech toolkit. These are essentially the flimsiest, least damaging tool you could possibly use to pop something open, and the clips still snapped. The only thing I think I could have done differently is where I started trying to pop the cover off. I tried to start from the front (which has two clips) and worked my way to the sides. I think if I started from the back, I would have gotten some leverage without cracking the clips. Still, even just lightly touching them with the cover off seems nearly enough to bend them and snap em off.

Overall, I think that's just a ****ty design on Dell's parts. Considering the amount of screws on the backplate, the laptop is never going to accidentally separate or move around at all, so there really wasn't much need for the clips. The microscopic plastic pins/dimples that stick up are enough to keep the plate aligned while trying to put it back together, so I really don't understand the purpose of the clips. I was pissed that a simple maintenance step so easily resulted in damage to a $2000 laptop, but ultimately the clips do not matter at all so Im not gonna lose sleep over it. If anything, it will make taking it apart easier from now on..

 

Would it br possible to have a picture of those retaining clips to see where to be extra careful?

 

I am going to take this one apart again when they replace it so I would love to know what clips I am looking for in case they are already missing, I was not gentle with this one as I treated it the same as my 9530.

 

I had a spare 30 minutes, one of my clips on the side had split and fell off when inspected (see if you can spot it on the flickr page I am linking to)
The clips are poor, just prising them seems enough to strip off the bit that latches preventing them fitting again. I had taken high res pics of the 4 clip locations. All I can think is using a stanley knife blade to slip in and push back the clip a little to assist in not buggering it up. The front 2 are needed to prevent the slight gap you see. I cleaned the front ones up with a small flatblade screwdriver to give them back the lip they need to latch but they only just catch.

Anyway here's my 9550 album showing the locations to help others who are about to rip the cover off. https://flic.kr/s/aHskpm4KzU

thinking about adding these to my tool kit.

http://www.ebay.co.uk/itm/iSesamo-g...038899?hash=item1a0b23b3f3:g:d6AAAOSwHjNV-SpT

 

I ended up deciding to re-paste myself just out of curiosity.

Stock Paste Idle Prime 95 (15 minutes Stress)
Core 0 29 75
Core 1 29 70
Core 2 30 74
Core 3 28 72

GC-Extreme Idle Prime 95 (15 minutes Stress)
Core 0 30 72
Core 1 30 70
Core 2 32 73
Core 3 29 69

During my stock paste test, there were a lot of throttling, fan was speed up at heavy use most of the time, and vid was being throttled.
I noticed after repasting, there was substantially less throttling, it maintained stable core VID throughout the 15 minutes.

While I could of tested this longer, just a few Celsius lower is good enough for me. Not much different in idle considering its usage.

Few days of burn in of the paste might let it settle even 1-2c lower. This isn't hard to do but I recommend this as well to gamers

I did repaste the GPU as well but, I don't have time to test this at work.

 

Are you guys using HWInfo to monitor GPU temperatures? If so, are you using the "DELL EC: DELL XPS 15 9550" sensor? HWInfo throws a warning that this sensor takes up more CPU time to read however it shows the GPU temperature, whereas I'm getting mainly empty readings under "GPU [#1]: NVIDIA GeForce GTX 960M:". I did a clean install with the latest mobile NVIDIA drivers.

EDIT: Nevermind, I realised that HWInfo finds active sensor during runtime, and that the dGPU was just sleeping due to Optimus. Running a GPU intensive application to wake it up showed the correct readings.

 

Hi,

What readings do you get from Dell EC: XPS 15 9950 GPU sensor during gaming? I am getting max 100degrees and don't trust this reading because all other GPU temp readings are in the low 80 (HW info, gpuz).

 

I'd actually like to know what each sensor is.

I was doing combined Prime95 (Blend setting) and Unigine Valley (1080 windowed, ultra settings) and red lining - over 100 on one sensor and high 90s on others according to HWiNFO. Both fans were over 4,000 RPM, but I really couldn't feel much air coming from above the exhaust ports near the hinge with my hand placed above it.

Personally, I like to do stress tests that focus on what is likely actual balanced usage (although I also do testing of just one aspect of a component). I think concurrent, balanced stress testing will create much more total heat inside the laptop to see if anything goes wrong over longer runs. Doing so has highlighted some issue with mine. Haven't had time to investigate yet...

 

Aida64 is quite good at sensor readouts, if you want a blended test instead of a try and blow the CPU test try the Asus realbench. http://rog.asus.com/file/?download=RealBench_v2.42.zip

 

Wow i just noticed my CPU temps going off the rockers when i played Fallout4.
So there is the limitation of cooling both CPU and GPU via 1 connected heatpipe sink.
I even had a BSOD in the process. I had to increase XTU offset to -175mv. The additional heat seems to require more vcore.

Just take note of this XPS15 users

 

Probably worth raising the back of the laptop during heavy gaming to assist in the cooling.

 

Suddenly my XTU settings are not applying on boot/reboot....
Any idea why? Halp!

 

Sometimes if XTU exits unexpectedly you have to relaunch the app and then reload your profile and apply it again. I've noticed mine does this every so often.

 

I think it may be due to me loading games. Before that xtu works great and always load up. Perhaps the gaming exe is confusing xtu?

Any further experience is appreciated. Hopefully we can get xtu work flawlessly for XPS15 over time, it is great app

 

Using the event manager to apply xtu values is more stable and it doesn't require keeping the xtu service running. I've been quite satisfied with this.

 

Good call, I'm going to run XTU that way, thanks for the tip

 

Wait, I'm confused. Does that mean by undervolting you will see higher temps while gaming?

 

No, this is what happens when you undervolt a stock CPU and thrash the tits off it, you cause system stability issues. While it may be fine for general use undervolting and gaming have never gone hand in hand

 

Nope, I mean by playing a game, the 960m is stresses and produced more heat than just stress testing the cpu only.

Therefore a good gauge of under volt stability is to test both cpu and gpu work load

 

On my XPS 15 i7 im getting -150mV easily with no problems. I'm testing -160mV and Stress testing on XTU looks stable

 

run a GPU stress test at the same time or it is pointless.

 

I tried -170mV and i used:
3 cinebench run
1 skydiver (with demo)
1 firestrike (with demo)
2 5 min xtu stress Test
1 xtu benchmark

The skydiver really pushed the GPU and the cpu! They got really hot!
EDIT: Tried xtu stress Test and some cinebench OpenGL run at the same time and it looks stable

 

Good stuff, not a bad undervolt to hold up at high temps.

 

What xps 15 i7 guys are getting on undervolt?

 

Hah, I'll let you know. My buddy just sold me his i7 version so I'll play with it tomorrow (if I get time) and see how she does. Now I gotta sell this i5, lol

 

well...I don't know about the performance difference, but i7s are i7s

EDIT: Doing -175mV stable. I wonder how far will this i7 go hahahah

 

Does anybody knows how to undervolt the 960m?

 

So my i7 is 12 hour stable on prime 95 at -175mv, but it fails LinX after 3 hours - I'm retesting at -165mv now to see how she does.

Sadly I don't know of a way to undervolt the GPU without a bios mod. Prema on these forums is the go to for that kind of stuff, but I have no idea if he does dell bioses as well...

 

But 12h is great! Keep the -175mV lol

 

Hah right, prime is a great test, not sure why LinX had an error so fast. I really like sable though, I normally test everything for 24 hours before I call it good. I'll have to open this guy up and repaste to see if that makes any changes...

 

Hello, I am new here.

I have not made any changes from the default, except updating the BIOS. Are these normal temperatures? I am considering doing under-volt + repaste. My fans kick in way too often. I want a quiet laptop.

Spec: i7/4k/512gb ssd/16gb ram

BIOS: 01.01.19

Room Temperature: 21 celcius

On battery:

CPU Utilization (idle)
3-7%, 5% most of the time

Core 1/2/3/4
35/39/33/35

Package Temperature
42

On power (battery 100%):

CPU Utilization (idle)
4-7%, 5% most of the time

Core 1/2/3/4
36/40/34/37

Package Temperature
43

On power (doing CPU stress test with Intel ETU):

CPU Utilization (stress)
100%

Core 1/2/3/4 (after 5 minutes of CPU stress, max temp)
75/77/67/72

Package Temperature
76

 

I'm trying -60mV on the iGPU now

 

Nobody wrote anything about the GPU up to now.

I am running at -150mV CPU and -100mV GPU. Works fine, but I did not try any harder.