XPS 15 9550 temperature observations (undervolt + repaste)

It doesn't work that way in my country. I would have to send my notebook to Dell... And then they would notice I have re-pasted and re-paded it by myself so my Waranty would be gone, I guess...

 

They don't have on-site service available to purchase?

How would they notice if you repasted it? Just take off your extra pads.

 

Actually it doesn't void the warranty, maybe try talking to them and tell them you opened it up to check the paste job due to it overheating and spotted the dent. You can then push them to ship a part as it is a manufacturing defect and why should you be without it whilst it is in for repair!

You never know your luck!

 

No, they don't have an on-site service available.

They would notice it because the paste looks different and also the thermal pads are different.

 

Good idea, I will try that. Thanks.

 

Hello everyone, I bought a 4K/i7/512GB XPS 15 model but sadly, mine is running hotter than most of the people in this thread have reported. I have undervolted the CPU by -160mV, the cache by -80mV and the GPU by -100mV and the temps while playing CS GO reach 90 degrees(the ambient temperature is probably around 26). I have decided to repaste it tomorrow and maybe replace the VRAM thermal pads if they don't make good contact with the heatsink. I read that some of you have put pads between the heatsink and the case and are reporting mixed results. Since my unit is running pretty hot I am thinking that this would be a good way to drop the temperatures a bit more. What are the negative side effects of doing so and would you recommend it?

Thank you for your time!

 

Bump, just did a quick "repasting guide" and added it to the first post. The Kryonaut thermal grizzly is freaking amazing paste, my ICD7 has been retired.

 

Elen,

Sorry to hear about your problems. All ultrabooks will get hot but start by carefully studying this thread as it has the most detailed responses on the internet. I am no expert but I did spend a lot of time studying this thread and asking some questions. . .

For me, Undervolting with Throttlestop (NOT Intel XTU as it did not work properly for me on the 9550) was most effective cooling process. But I (and apparently others) have knocked a few degrees off with each of these steps:

1. repasting the CPU & GPU

2. replacing the VRAM pads

3. Put a laptop fan under the laptop

Total drop might be 10*C, but more importantly this reduces the probability of throttling due to a defective assembly. So I see the benefit could be significantly greater from a practical perspective...

There have been plenty of reports that the pasting and VRAM pads are defective out of the factory so at the limits I personally consider they must be fixed. And if your change the VRAM pads, you need to repaste the CPU/GPU. Vice versa is likely true.
It seems one or two of the VRAM pads don't reach the heatsink; one poster wrote he saw some 10 9550s and all 10 had the VRAM defect, a sample large enough for me). The factory fluffy foam pads might not break when disassembling the laptop for a repaste but they might so I would expect to replace all four

Earlier this summer, I did a search on this site for people adding thermal pads between the heatsink and the metal case bottom; look for the posts. If I recall correctly 3 or 4 people tried it and all removed it (GoNz0, custom90gt, supermainedave...?). As I think GoNz0 (our resident laptop engineer) explained, heating up the base helps at first but then it tends to heat up the incoming air that is meant to cool the CPU/GPU so that the cooling system becomes increasingly less effective. And maybe the keyboard gets hot.

Anyways study the thread and ask questions and hopefully you get improved thermals...

 

Nice guide and photos!

What was the thermals improvement for you between ICD7 and Thermal Grizzly? Or is it just easier to use?

Also I would note to everyone - be careful when removing/replacing the battery cable as the contacts are impossibly small and delicate. Furthermore, use caution not to dislodge the monitor cable which is adjacent to the battery cable in the photo #3 (and covered by a silver covered plate with a black screw)

 

I will say again.

After testing pads added to contact the chassis it will eventually heat the chassis so much it will suck in hot air and basically stop cooling. following the throttling if you don't stop it will turn into a singularity and envelop earth. Yes I have been drinking.

 

On my desktop I noticed about a 4*C drop at idle and the laptop a 3*C idle drop. This is likely explained by just a better paste job, but for both of them to drop substantially like that I am happy.

 

Only drops can be counted as a success under controlled conditions such as 100% fan and the same room temp with a specific test being run. As a laptops got dynamic cooling most results can be tossed out the window at idle.

 

You can say that, but I check my temps multiple times a day every day and the idle drop is there. The fans for all intents and purposes never spin up on my XPS now, it's nice being in school and having a quiet laptop.
You can also read reviews on the paste, it's the best stuff out there aside from liquid metal pastes.

 

I read pretty much the whole thread today and found it really informative so thanks to everyone who contributed! I decided to do some tests with my i7/16G/4K/512Gb model. To begin with, the weather here in Bulgaria is extremely hot right now and the ambient temperature in my room is probably around 25 degrees.
1.When I ran prime with the stock paste and voltages the maximum temperatures were in the high 90s and the cpu throttled to 800Mhz.
2.I undervolted the core by -160mV, the cache by -80mV(even though I don't think that made any noticeable difference) and the iGPU by -100mV. The temps after 10 minutes of Prime95 were in the low 90s and the cpu only throttled to 2.6Ghz in order to cool down from time to time.
3.When I opened the laptop it turned out that the thermal pad which was meant to be on the ssd was on the battery . I just removed it and put in back on the ssd. After repasting with noctua NT H1 and replacing the thermal pads on the Vram the temperatures with prime are now around 80 degrees.
As a whole, I am happy with the results but one thing that bugs me is that the cpu still throtles down to 2.6Ghz sometimes. Is there anything else I can do to lower the temperatures even more? I have not tried pushing the undervolts further because I thought that there won't be a noticeable difference in the temperature.

 

Undervolting helps quite a bit, don't discount it. The trick is finding where you are still stable at the lowest voltage. Another thing you can do is turn off hyperthreading in the bios - on previous laptops I was able to decrease temps that way. Unfortunately there is only so much you can do because the cooling system just becomes saturated.

 

I wouldn't turn off hyper-threading. You're basically turning your i7 into an i5, and you probably paid an extra $150 for that. If anything, just disable turbo at 80C or something.

 

Hyperthreading is not often advantageous unless you're doing something that uses more than 4 cores... The extra ghz from lower temps will help far more than hyperthreading...

 

Thanks for the quick responses! I will try to lower the core voltage a bit more. Another question I have is about the thermal pads. I straight up replaced all Vram pads without even testing if they make good contact with the heatsink. The ones I put in are 1mm thick phobya pads. I am a bi worried that they might be too thick and might be pushing the heatsink off the gpu. Also how can I disable hyperthreading? I mainly care about gaming performance and I don't think that extra threads will make any difference in most games.

 

I can't answer the thermal pad thickness question, every heatsink is a little different. The best way is to do a test fit and see what works. For this newest laptop for me leaving the stock ones and then putting a 1.5mm on the heatpipe part worked the best by making the right amount of contact.

Disabling hyperthreading can be done in bios setup, I don't recall where it is and am in class otherwise I could reboot and tell you.

 

I had to double check to make sure I wasn't going mad but you are complaining the chip throttles to it's designated (none turbo) speed?

 

Well I saw some guys here claiming that theirs were constantly on 3.1Ghz during prime 95 so I was a bit jelly .

 

Ymmv but all yours is doing is dropping out of turbo

Sent from my SM-G920F

 

I am surprised you are reporting the undervolt did not provide much benefit. That is an aggressive i7 undervolt compared to other 9550 posts I have scanned here. It is tough to interpret but people seem to be reporting 5-10*C improvement and reduced throttling. My sense is that undervolt provides most people the biggest benefit (at least on benchmarks) when compared to the repaste & repad.

Are you sure that your undervolt is being applied? A couple of problems I noted:

1. Intel XTU seems buggy with the 9550 (it forced power throttling at about 79*C on my i5 and on the 9550s of a few others here). XTU has charts to show power throttling. I did not have throttling on Prime95 but did on RealBench stress test 15 minutes

2. Intel XTU interacts poorly with ThrottleStop so you can not have both on your system

3. ThrottleStop seems to work on my i5 as I would expect. If you use ThrottleStop post in that section of notebookreview.com as the programmer frequently drops in to help.

 

I used to use intel XTU but after reading some of the posts here uninstalled it and am now using ThrottleStop. Here is a screenshot of HWMonitor right after I booted the system. picture Looks like the undervolt is being applied.

 

Intel CPUs should be using as much Turbo Boost as possible, all of the time, unless they have reached the thermal throttling limit or the TDP limit. Even then, the CPU should not completely turn off Turbo Boost. It will simply modulate it, hundreds of times a second, so it can maintain maximum performance without going over one of these limits. Here is an example of a properly functioning 4700MQ while running Prime95 Small FFTs.



TDP Throttle is lit up in ThrottleStop which confirms that throttling is taking place due to the CPU reaching its 47 Watt TDP limit. Throttling is slowing the CPU down just enough to keep under that limit (46.9 Watts). The default multiplier is 24 so you can see that some Turbo Boost is still being used, even when it is throttling. That is how an Intel CPU is supposed to work.

The same thing happens when an Intel CPU hits the thermal throttling limit. It should only slow down just enough to keep the CPU from exceeding the maximum safe operating temperature which is typically 100°C for most Intel CPUs. During throttling, the CPU should not drop down to 800 MHz and sit there at that speed for any length of time. It should also not completely disable Turbo Boost. Throttling like this is excessive and it only happens when a laptop manufacturer decides to create their own limits and ignores the safety features that Intel builds into all of their processors.

I get around to quite a few forums. Ebay has some good prices on the similar Dell Precision M5510 but every time I read about a throttling problem like this, I think twice about what I am getting myself into.

 

I notice I only drop to 800 when I hit the thermal trip, it will slowly climb back up again, it doesn't stay at 800. When it hits 47w it only throttles back a bit. It quite happily sat at 3.1ghz for a while until prime started to run harder tests and went to 47w

 

UncleWebb,

It took a bit of work to get my 9550 to run correctly but I am really happy with it.

Love the screen, keyboard, and case. It is superlight and once tweaked, it is a good performing ultrabook. Runs cool and quiet.

If one is not willing to do a clean install of W10, fix the thermal paste and replace the defective VRAM pads, this is not the laptop for him or her. Warranty claims seem to have scaled down but are obviously not zero. Dell seems to have sorted out most of the bugs at this point but you can study the forums to get informed there. You can also see how responsive Dell service is as there are a few stories worth reading.

There is so much knowledge posted at NBR for the 9550 which makes ownership much easier; it mitigates risk as problems are well documented and there are some sharp people willing to lend a hand...

 

@unclewebb I'll be your personal xps/precision troubleshooter if you need one

 

Didn't you just sell your XPS 15 due to too excessive pain and suffering??? That is an important datapoint.

More importantly, Eason has been super helpful and we appreciate all his help and guidance!

 

Yeah but after I clean installed, it's been running flawlessly with the guy I sold it to. I was just ok with getting rid of it because I got a good price and I know that I'll be buying the very next quad core ultraportable with tb3 that comes out, so it was just a matter of time.

 

I did some testing with a friend's XPS 15 and when his i7 gets hot it doesn't reduce its clock speed straight to 2.6Ghz but to 2.8 or 2.9Ghz and maintains that. Mine seems to jump straight to 2.6 in order to bring the temperatures down and later goes back to 3.1 and gets hot again. Do you know how I can change that?

 

Read the thread m8, lots of information on what others have done/tried. You can decide on what course of action you want to take then.

 

I had mine in bits for a clean out today. Decided I was going to crack open the "Gelid GP Extreme Thermal Pad 1mm" Seems to be a match on the pads that were on as they looked the same and also I had resistance on the piece of paper I slipped between as a test (had the heatsink on and off about 15 times today)
I noticed that the GPU wasn't fully mating when I added a 1.5 or 2mm thickness pad (reused a 1.5 as a test) so I ended up tweaking the heatsink a little to bring down the metal plate and filled the gap with a cut off bit under the heatpipe (the part that doesn't contact on a stock build) so it all sits on 1mm pads now.
It still didn't fully mate but was as good as the original fit and did mate with no pads and spread the paste out thin.

I have stuck it all together and will have a peek in a few days once the pads have settled to make sure the heatsinks gone flat against the GPU.

initial tests were quite good so we will see. It will AIDA64 fpu test at 3.1ghz without throttling and the fan stays at 4k max

 

Sounds like some good results. Any pics of the work?

 

How did you know that the GPU and heatsink weren't fully mating despite spreading the paste out thin?

What size vram pads did you end up using? The same for all 4 chips?

 

Nope, I did this while my 2 year old was having a nap and during intense games of KerPlunk making sure I didn't beat my 6 year old!

1mm and added a bit in (triangle) where the dodgy one isn't a complete rectangle to fill the gap between VRAM and heatpipe.

The clue was the bit where I said I had taken the heatsink off about 15 times
Never has a laptop been repasted so many times in an hour!!!

The bare chips are small enough that the grain of rice method is more than enough to press out and cover the entire chip when you have good contact. the GPU was only pressing enough to form a circle within the boundaries of the chip and had not gone thin enough to resist pulling up into peaks when removed. A good spread will stay flat on both surfaces, a bad fit will pull peaks like whipped cream due to the thickness left on the surfaces.

I will try and remember to grab the camera when I have a look in a couple of days.

 

Hi could you tell me how wide the thermal pads need to be? is 50x15mm enough? thanks

 

Anyone have recommendation on the size of the thermal pads for the heatsink?

I see 15x15x1mm as well as 10x10x1mm pads on Amazon.ca

Which one would fit best? And is 1mm thick enough or should I go for the 2mm?

I forgot to check when I re-pasted earlier, lol.

 

I bought a 1mm thick pad and cut to shape.

 

Hi Bernie,

14x15 mm squares were just a bit larger than the 4 VRAM chips on my 9550; I cut just a bit larger. Someone measured the VRAM chips here and the actual dimensions were just a bit smaller (you can search the thread for those dimensions).

As to thickness, I guess that depends on how defective your heatsink is (perhaps the heatsink production is at a high level of precision with very low variance but I doubt that given that the overall production is defective). One member here opened nearly a dozen 9550s and said they all were defective.

Next, it seems that one OR two of the left VRAM pads are consequently too thin, but I don't know if that is the case for everyone. You can open up your system and test.

A. I tried using 2x1mm with 2x0.5mm fujipoly pads. But after clamping down the heatsink and unclamping, the pads barely had an indent from the chips.

B. So I tried 2x1.5mm and 2x1.0mm.

Testing (e.g. RealBench StressTest) results were virtually identical for scenarios A and B. It appears that most people are replacing defective pads with new 1.0mm pads (including GoNz0 above).

A couple of preliminary observations from my novice "research":

1. Higher rated pads seem to get very dense (the top-rated Fujipoly pads are like clay whist the stock pads are very foamy and compliant)

2. Even the best pads are not very good heat conductors. So use the thinnest ones you can get away with. That is, thinness is much more important than having top rated pads

3. The laptop clamping mechanism (a few screws) does not exert much downforce. So using overly thick/dense pads could unfortunately separate your heatsink from the CPU and/or GPU creating new thermal issues that thermal paste was not designed to address

4. You need to repaste the CPU & GPU every time you take off the heatsink. Also the pads are delicate and tend to rip; you don't want to reuse ripped pads

I figure you know all this but I like to keep personal notes here for future reference; and perhaps someone might find them a good place to start...

 

@pressing, I guess that is why I had trouble trying to pack another 1mm under that dodgy bit missing due to the heatpipe, I use Gelid 12W/mK and it was like clay!

 

I think that was the case!

I scanned some manufacturers' spec sheets for a few brands of thermal pads and also read a bunch of forum comments. As a general matter it seems the highest W/mk themal pads are much more dense than low rated ones are. That said, I am not sure there is a perceivable density difference between say the Fujipoly 5W/mK and 7W/mK pads. I also don't know which brands are "fluffier" at a given performance rating. Or how comparable those ratings are....

Interestingly, I never found a definitive answer as to whether to use thermal paste with thermal pads. I didn't do it because it is messy, expensive, lifespan risk (possible unknown chemical reaction between pad and paste), and lubricant properties of paste knocking pads out of alignment under pressure & changing heat.

Also never found a definitive answer regarding stacking thermal pads. I think a single pad of proper depth is optimal but if the contact between the pads is smooth and tight it probably does not matter too much. Due to the poor thermal conductivity of all pads, thickness really seems to be king.

 

The only time I have ever seen it done was on my old GTX280 when I put a Koolance block on it, they said it could benefit using paste with the pads on the mofsets, I assume this was due to a possible clearance issue otherwise I wouldn't bother as the slightest compression of the pad is more than enough for it to work

 

Do you guys think this would be sufficient for my VRAM chips? If not, can you guys link me to a quality thermal pad?

http://www.ebay.ca/itm/Phobya-1MM-P...981821?hash=item1a03ce003d:g:2wwAAOSwT6pVvgKZ

I'll probably cut them to size with a box cutter. Has anyone had any difficulties doing that?

Would going for thicker 1.5mm or 2mm pads cause any issues? I'm just worried that 1mm might not be sufficient contact with the heatsink.

I found other ones on eBay as well like these, but the thermal conductivity is much less at 3.2w/mk as opposed to the 7w/mk Phobya ones in the link above

http://www.ebay.ca/itm/30pcs-15mm-1...585450?hash=item3398204baa:g:NacAAOSwxH1UFslj

Then there are these ones from Arctic Cooling (a brand that I trust)

http://www.ebay.ca/itm/Arctic-Cooli...999291?hash=item1a14bbc37b:g:GY0AAOSwmtJXUpKF

 

harder 1mm pads (12w) caused issues for me so 1mm 7w pads should be fine, why use a box cutter? scissors work fine!

1.5mm may be of use where a bigger gap exists such as the vRam chip under the heatpipe.

Thicker pads stop the GPU part of the heatsink making contact as another user found out by not checking then wondering why the laptop keeps crashing.

 

Thanks, I ordered the Phobya pads.

 

Anyone know anything about undervolting the 960M?

With my 6700HQ capped at 2.6GHz and a -150mV undervolt, it's nice and cool when I run my Steam games. Roughly anywhere from 65 to 70 degrees.

On the other hand, my 960M can be seen anywhere from 75 to 80.

Would be nice to drop the temps down a little bit.

 

Impossible without flashing a vbios.

 

Hi Fellow Dell XPS 9550 owners.

To the folks who have reported their baby getting hot especially on the left side where the AC plugs I suggest to check out the condition of the actual heat sink. I recently got my hands on a spare Dell XPS 9550 and swapped out the heatsinks. I did this for two reasons. The first my laptop was also getting hot when playing. It was borderline uncomfortable and I know other people have reported the same thing. So second, when comparing my heatsink to the spare Dell XPS I noticed the left side of the heat sink was all banged up. Literally, it looked like someone stabbed it a few times and it just looked awful. A few dings here and there were definitely noticeable compared to the heatsink on the spare which looked better. I think this had an impact on airflow and cooling overall.

I quickly operated and after installing I played and guess what ya'll? no more heating up on the left side. I think this is how a Dell XPS 9550 should be. I haven't had time to repaste and now that I now how to access the heatsink I will go ahead and do this. I ran XTU and CPU while playing was at about 75-80 but I think this is because I just slapped on the heatsink from the spare. This might not be a good idea, but eh. Right now I have a bunch of stuff opened, mostly chrome, and is sitting at 43c which is acceptable and honestly, I think that was close to stock.

I'll report back after I repaste. For people thinking of doing the repaste method, just do it. It is really easy. A few screws hold the heatsink in place and boom.

By the way, what do all of you use to clean off the stock paste? A microfiber? A used diaper? what? Thanks.