XPS 15 9550 temperature observations (undervolt + repaste)

Interesting point. Maybe it would make sense to use a special plastic or maybe cover the sticky side (with reversed tape) where air goes...

 

LOL I saw someone on reddit cool the wifi card...why not? lol

and I put just some regular paper over the open area of the electrical tape...forgot to mention that part...the only stickiness is where its touching the heatpipes and frame...lol

 

This closes just the path of the vent air over the non-finned side of the heatpipes.
There are pads to direct air, but I guess the sealing isn't tight, and with increased pressure likely more cool air gets past the rubber pads to the VRM area, likely some towards the edge too.
Yes dust is likely to stick more, but I've noticed it already does quite fast with no mod despite not working in any particularly dusty place.
Dust accumulation before the fins might block the fan exhaust more tightly without this bypass, not sure if this might harm the vents.
The tape is likely to unstick, shrink, twist... Maybe the fluffy cheapo thermal pads instead?

 

TBH unless I see a reasonably controlled before and after I think it has enough ducting in place.

Reminds me of the famous quote "if he stuck his head in a fire would you?"

 

Yeah, was thinking you could line up thermal pads there and they could touch the rear cover

 

So I tried this with 1.5mm thermal pads sealing the gap instead of tape--I think it lowered my peak temps by 2-3 degrees. Give it a try instead of the tape and let me know.

 

My concern with adding thermal pads above the heatsink fins is that it'll push down the heatsink and cause uneven contact on the CPU or GPU... I figured the tape was good and to keep the heatsink fins floating since there appears to be a gap above and below... Bottom of the laptop definitely gets much warmer than before with the new thermal pads...not uncomfortably hot though which is great!

 

I'd venture to say I can bench the gpu at 1900mhz...lol might blow up though


Sent from my iPhone using Tapatalk

 

I'd venture to say I can bench the gpu at 1900mhz...lol might blow up though


Sent from my iPhone using Tapatalk

 

Are you sure about those temps? Because at this point if everything we thought worked actually worked we would have i7s running at 70° during stress tests.

Just to check what the actual gain was (and to see if it was easily removable) i removed the liquid metal and replaced it with kryonaut and temps increased by 4+ degrees, i'll run it for a couple of days (waiting for dell's response on a replacement and if they send a technician i'd arther not have liquid metal around).

By the way removing conductonaut is kind of a nightmare because it stays liquid, it doesn't stick to anything and it doesn't get absorbed by anything (plus trying to be careful to not make it touch anything else makes it even worse) but it's doable.

 

Are the pads putting exhaust heat into the chassis now?

How is it after several hours of mild use, that was the end of my testing after adding pads to put heat into the chassis as the keyboard and palmrest got hot after a few hours surfing so I gave it up as a bad idea. Seems to be ok while it is working hard as the fans get shut of the heat. As soon as the fans shut off (and the mods helped keep them off longer) things warmed up and ended up after a while sitting at 3000rpm and it got annoying (and hot)

 

I'll use it more today and let you know. I just played overwatch for 30 minutes or so

 

Ok, after about an hour of OW my CPU went down to 800mhz and my fps got completely screwed. But my highest temp was 82 on the VRM, my highest CPU temp was 80, and my highest GPU temp was 78. What gives?

 

Did you have pads all over the heatpipes near the vents or just to seal that gap at the fins exhaust?
Did the VRM area temperatures show any decrease initially?
I guess that heat conduction from the heatpipes to the case bottom may show initial decrease of CPU & GPU temperatures, but lead to VRM heatup after a while.

 

Are using the case bottom to heatsink the vrm and the two radiators?

If so, I think the case bottom is overheating then not providing any heatsinking to the vrms; it also superheats the fan intake air as GoNz0 suggested a few months back. I had that issue with 17 W/mk pads to the vrm chokes- thermals spiraled out of control pretty quickly...

 

I'm using a very thin strip to seal the fan only. I have two strips on top of the GPU sink.

The case bottom does feel very hot, though. I have electrical tape, but it's only rated up to 80C. I'm concerned about using it on the pipe gap. The CPU/GPU peak at 80C, but I'm concerned the end of the heatpipes will be close enough to that.

 

I recently just repasted my 9560, It has decreased temps by around 5-10c (Waiting to to proper benchmarks etc) but it has brought on that horrible coil whine, I was wondering if there would be anyway to fix this? Could it be that I've tighten the screws ?

Thanks

 

hmm... so I get CPU throttling in BF1... temps only get upto 74C... VRM 79C...not sure what it could be. I've increased my wattages and amp limits... Fortunately it doesn't really reduce performance that much. I can still hold 60fps in BF1 on medium... GPU doesn't throttle at all which is nice. Still haven't reached the max GPU clock... +160 on the core, hitting 1873MHz in Heaven 4.0 benchmark..

 

Yeah this is my boat as well. The GPU will hold clocks but the CPU goes down. None of the ambient or other sensors read over 80C.

 

The only way I can stop it from throttling is by uninstalling then disabling the Intel Dynamic Thermal Platform stuff through device manager... which works until it reinstalls itself.

 

On my 9550, when the vrm hit ~78*C, PL throttling began.

Some users insisted that PL throttling was not temp related but that was not my experience

 

Bah. I've got the main VRM area covered and sinked to the rear cover. It isn't highly conductive stuff, maybe I need a more conductive material?

 

Balancing act. The bottom cover is a lousy heatsink. So remove any thermal pads not related to vrm and that might give you more headroom for vrm without using different vrm pads. I think you still need to heatsink nvme ssd.

I used 17W/mk pads on chokes and that was too much heat to bottom case as you know...

 

Nothing at all is touching the cover except the VRM. I'm using a 3mm pad originally for a heatsink on the VRM/chokes.

 

I'm not clear on what you mean by two strips on top of the GPU sink?
You do seal the fin exhaust to the case edge as the tape in the photo did, don't you?
The fan exhaust (still cool) to the fin entry shouldn't be sealed I think, for cleaning dust which mostly gathers there, and a little bit of leakage towards the VRM area might help cool that.

 

If you drop the prerendered frames in NVCP to 2 instead of "game selected" or 4 it should stop your CPU throttling issue... it does help reduce the load on the CPU...FYI!

Went in for another game this morning... Can easily do 1860MHz in BF1 without throttling... gets upto 72C without a cooling pad and VRM top out at 79C which seems to be right around the right number... That's about as much as the laptop can take in all honestly... unless I can get a cooling pad to really reduce temps and push air into the heatsink on its own without the blower fans doing all the work...maybe I can squeak out a little more...I can probably do 1900MHz in heaven or valley benchmark... Stock valley is I think around 31-32fps... overclocked to 1860MHz and memory overclocked I get 36fps which is a nice 10% boost in performance... Got a nice firestrike score to go with it too... http://www.3dmark.com/fs/11907972
It's pretty solid... think I did about an hour on BF1 last night and had no issues whatsoever....

 

Thanks!

 

Judging by reviews of the GTX 1050 desktop...the GPU core can handle a ton of power....what's interesting though is desktop versions are clocked at the same numbers...same voltage...and essentially same performance...yet they pull around 75w in total... this GPU should be more around 50w if my research is correct... quite surprising the difference...

I'm pretty happy now that I can use my laptop for a good 5-6hours on battery and have absolutely 0 noise coming from it on battery...fans never turn on... then go to gaming and play every game I have no problem... I'm really glad I got rid of my big huge desktop for a laptop... heck even my ITX build was too big... definitely diggin the laptop life

 

Maybe you need lower rated thermal pads on the vrm. Or just less padding. A very hot case bottom is terrible for performance in my experience.

Why not use 105*C electrical tape? I have a 3M link a few pages back...

 

Apparently you can't buy electrical tape in Bangkok lol

 

Look for "Kapton" tape. It is specifically used for high temp applications. 3D printers users are using it on the heated bed, etc.

 

Picked up a crappy Deepcool aluminum cooling pad. Will likely mod the USB controller for an outlet DC wired power connection but should help with overall temps...

 

What mod?

 

Cutting off the stupid USB power for the fans and wiring up a DC power controller for the fans power instead...


And yess... Thank you free subscription to Dell Premier since I'm a student... 1 day free shipping. It's nice living close to the warehouse.... Should be able to do some mods on it this weekend! Will report back my greatness!

 

hehehe - typo or translation error.

I think you meant to say something like "Look forward to reporting back great results"

 

Educational post only

Uncle Webb (creator of ThrottleStop) had an interesting post on "adjusting" PL2 and PL2 for his Kaby Lake system. Intel XTU lets us make these adjustments easily.

But can we bypass XTU on the XPS?

On my XPS 6300HQ, one can adjust PL1 wattage, PL2 wattage & turbo boost time window via Intel XTU.

Those adjustments helped improve performance in my 6300HQ for certain applications. Problem is XTU is a bit quirky and randomly drops those settings. Thus, I need to open XTU every few weeks to make sure it has not messed up these settings or undervolt settings.

It looks like you may be able to change those settings manually on the XPS without using XTU. But certainly not certain. Perhaps the CPU looks in three (?) places and chooses most conservative value:

1- Memory via RW-Everything ==> FED159A0



See two registers at FED159A0:

00DD 8 1E0
0042 8 210

1E0 = 408/8 = 60 watt turbo boost power max **
210 = 528/8 = 66 watt turbo short power max **

The 8 indicates the register is active

00DD = 221/8 ~28 seconds turbo **
0042 = 42/8 ~5 seconds turbo short

** = I set via Intel XTU. Factory settings are less aggressive

2- MSR610 (?)



Note the turbo time here is a little higher than value in FED159A0 (probably screwed up from my testing).

00FE = 254/8 ~32 seconds turbo **

3- Intel Dynamic Platform & Thermal Framework
Not sure what to do with this. I know some people deleted the driver but that will kill some thermal protection. And the XPS does not have a lot of thermal headroom...

Note - TDP seems to be locked down
Note - Not responsible for your damage from trying this stuff out. It can easily brick your laptop and destroy all your data...

 

Interesting...

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-563

I read in the Lenovo T460s review that they allow a higher continuous TDP than 15W for U processors in some (not all) models, presumably because they are confident that the cooling system can dissipate the heat. Which mainly allows the highest i7 models show more advantage in comparison to the lower ones, because in most laptops they can do turbo for just a few seconds, and their power draw in turbo actually is above 15W.

I understand correctly, with the 9550/9560 you can raise turbo TDP a bit using XTU, but not continuously? What's the max turbo time? Or you can go continuously?
And that this would mostly make sense with a decently-pasted heatsink so that there is some headroom in the CPU temperatures, and when the GPU is idle (unless maybe using extreme mods).

http://www.notebookcheck.net/Lenovo-ThinkPad-T460s-Core-i7-WQHD-Ultrabook-Review.161028.0.html

 

AFIK Lenovo has been using 28W TDPs in their T4-- series.

 

Thanks _sem_

TDP is locked and I don't think anybody has changed it on recent HQ processors. Not sure it matters as there are plenty of gating items preventing you from going much above 45 watts for any duration. For example:

1 - The 130 watt power brick is provides some external cap on total power available for screen CPU, GPU, SSD, network card, etc. (assuming no battery boost).

2 - The vrm caps CPU power. For example, there is a controI chip PU1000 and the components have physical limits. I doubt the vrm could even reach 50 watts for any duration (e.g. 9550 mosfets (AOZ5019QI) are rated at 30A max 1.8v):

* You can probably undervolt CPU to approx 1v but...
* The mosfets' 30A rating falls (a lot) as the mosfet is pushed to power and thermal limits
* Assume additional 10%-20% efficency loss in the vrm
* Assume mosfet performs below spec
* vrm needs margin of error

3 - Gates created by Intel and Dell in the CPU, via BIOS, via drivers (thermal, voltage, timing, etc). Very little public documentation available here so lots of assumptions

You can not change TDP. But you can increase turbo power limits (PL1 & PL2) for a limited amount of time with Intel XTU. Noted above are some of the gating items which may cap your XTU settings.

Don't know. Some ideas to start

0. You can disable c-states to keep processor running "flat-out" but that also disables turbo and creates a lot of heat which may drive premature throttling / component failure.

1 - With ThrottleStop, I can enable SpeedShift (EPP=0) and my laptop can run my CPU at turbo speeds ~100% of the time (3.0Ghz is max 2-core turbo speed for i5 6300HQ). At the same time c-states are mostly c7 and below so consuming very little power and creating very little heat. That is a nice trick for snappy but efficient performance.

2. Intel XTU has a 96 seconds max "Turbo Boost Power Time Window". I would guess Dell and Intel have hidden max settings that override any extreme manual changes. For example, in my previous post you see a 5 second max short turbo in MSR610 & FED159A0 which was not accessable to XTU. There are likely other hidden factors in the BIOS, Intel Dynamic Platform & Thermal Framework, etc...

3. Since there is so little transparency from Dell & Intel, I "roughly" optimized with a few benchmarks. Then did some trial and error expirments with my favorite software. You can see the SpeedShift thread to see what worked for me. Your favorite software or games might perform better with different settings.

As you have seen, the 9550 & 9560 genarlly run poorly as temps reach 80*C. There are plenty of cases showing the typical quality of Dell factory thermal paste and how a good repaste will increase thermal performance. Moderate undervolting can also provide a good performance boost.

I don't think increasing turbo watts and time is a particulary good idea in a laptop. In fact use extreme caution. The vrm is at its limits and can fail with more heat-wattage. How would you repair that? The Kaby Lake processors are small and not especially robust so less tolerant to high wattage than older CPUs were. On my laptop these turbo changes help some software run smoothly in some low power situations but I am not sure why. Perhaps just knocking out some mistakes or poorly set limits by Dell-Intel.

 

I'm not afraid that something would overheat. With stock paste, my CPU was very smoothly limited at 100oC (my ancient Lenovo T would say Pk! and stop immediately). Also the power limit throttling seems to take good care of the VRMs. I guess raising the power limits doesn't have any influence if the temperatures are too high. I wouldn't go higher than the original short-term power limit anyway. I think the heatpipe should be able to take care of that continuously with the GPU idle. The VRM wattage might be an issue.

In that Lenovo review they write about configurable TDP, I understand this is actually the power limit you discuss. For the U processor, both the short-term and the long-term power limit are at 25W, so the max turbo time doesn't matter (usually the long term for U would be 15W). Surely Dell wouldn't want to stretch that with their sloppy pasting, Lenovo don't in their thinnest laptops.

 

I really try to keep all thermals below 80*C.

As a general matter, as electronics get hotter, failure rates increase exponentially. We can't get around the basic physics...

 

Uhhh guys be careful with the shielded VRM. I was putting the little metal shield back when there were some sparks.... the laptop didn't boot at first, but after unplugging the battery and resetting, it works now. Just remember to unplug your battery before you do anything

 

Thanks Eason - definitely want laptop unplugged, battery disconnected, and proper personal grounding while messing inside laptop.

Could you highlight on a picture what shield you are noting sparks from?

 

It was when I was trying to put the shield back into the metal clips. It touched the two clips on the right side and then I got sparks. Quite freaked me out

 

I also am stupid enough to work on my laptop without pulling the battery

Did you get that sick to the stomach feeling?

 

it's actually the first time in years I ever got something like that. My stomach *would* have flipped over was I not already expecting a technician with a fresh mainboard 30 minutes from then. The fact that I got it to turn on again was just a bonus since I wouldn't have to make up some story for the technician, lol

 

My opening line would have been

"so anyway, the magic smoke leaked out earlier"

 

In your album I see that you have thermal pads directly over the GPU/CPU.
This is a bad idea as it compromises the cooling concept in some way.

The heat that would now normally be transported to the fins and dissipated through air directly outside is now in part transported to
the case lid which in turn radiates heat inward and outward putting more thermal stress on the other inner components like VRM/RAM/Chipset/SSD.

Source: I tested this myself and got way worse results.

 

Yeah we did mention this but he carried on anyway

 

Oh, sorry. I just now noticed I landed on page 110. Oops