[CPU + GPU Temperatures + Benchmarks] - XPS 15 [9560] Kaby Lake

Good to know, always thought the limit was made up by the chip manufacture. But if I recall correctly iunlock mentioned that his FET got up to 105°C before throttling.

 

I think this cpu thermal limits 97 not 105

Do what I do, stop giving a toss and leave it alone after a repaste and enabling speed shift



Sent from my SM-G920F using Tapatalk

 

I don't know about XPS. But it's unlikely you will find temp sensors for other than cpu, gpu, pch and ssd's in most laptops. Some even with no temp reading for pch and ssd's. It's the ODM/OEMs who put the limits in their cancer firmware!!!

FYI T junction temp is 100C for newer Intel processors.

 

For power limit throttling, the limit is a bit over 70 at the sensor in the VRM area. That 105 was measured manually on a mosfet.

 

I have vacation right now and might test the mod with the heatsinks and the modded ventilation, iunlock published.
But you´ve got a point there.
I do have to manually enable speed shift? Thought MS did implement it in Win10 last December and it would be enabled by default.

I just reread iunlocks post and he had 105°C on the first ambient sensor (HWiNFO64), which should be near a mosfet.
Did I get it wrong? I thought when this sensor reachs 105°C the CPU or GPU would throttle, because the mosfets are too hot?

Thank you guys, learning lots here.

 

This laptop by default has speedshift disabled and all option hidden, you can either use throttlestop at startup to enable it or as I have done via the EFIshell to flick the switch in the BIOS, then I added the reg entries to show the options needed in the power options, it's all in the 1st post here.

If you enable speedshift and change nothing turbo is technically disabled as it never reaches it, I changed "Processor energy performance preference policy" to 40 so the max turbo multiplier is reached.

 

There is a full thread here dedicated to SpeedShift on the XPS which will get you started. I think SpeedShift is an excellent Intel feature for energy savings and can provide some performance enhancements for laptops

http://forum.notebookreview.com/threads/dell-xps-speed-shift.796891/

 

Energy savings giving Performance enhancement?

 

hahaha. Unfortunately with the XPS, if we disable c-states, we lose turbo speeds.

SpeedShift is snappier than SpeedStep (or disabling c-states) so you can have your cake and eat it too!

 

@GoNz0 , @pressing Thanks guys, I´ll enable speed shift as soon as I get the XPS.
Really thought something like that would be enabled on default, cause it´s an advertised feature of Skylake, what was I thinking

 

That's what a lot of people were thinking. Dell probably saved some testing time by disabling SpeedShift.

For my huge virtual instruments, SpeedShift significantly lowers latency so is a helpful feature...

 

If as you say so, the machine is crippled It's your machine. Not the ODMs. Even Micro$hity claims your pc is theirs

 

It is crippled. That is big brother "protecting" us...

 

Hey guys, I'm about to try cooling my VRMs with thermal pads, but I can't find those mini heatsinks anywhere (Philippines). I may be able to find thermal pads though. What if I just straight up connect the VRMs to the bottom panel using the thermal pads? I have my laptop on a cooling pad when gaming anyway (yet it still VRM throttles, hence why I need to try this.).

I read some stuff elsewhere that connecting it to the bottom panel "will superheat the air around it and compromise the entire cooling system." But basic physics knowledge tells me whatever heat isn't inside the case means it's good. What do you guys think?

Can I get away with just connecting all these hot VRM components to the bottom panel via Thermal Pads?

 

this was somewhere in the Speed Shift threat
I read your post and thought I am exactly this kind of idiot
But thanks for giving me a link to that threat, I never imagined it to be disabled on default.
GonZ0 you, pressing, unclewebb, Rockstar75 and goodwin_C really did a awesome job there, thanks a lot!

 

There are other coolers available normally for desktop Pc´s, there is a kind that has like four straight rods, these could be easily shorten with a saw, just keep them under 3.5 to 4mm. But google or search on ebay for mosfet heatsink and should find some. You have to have airflow for them to work.

I would not attach thermal pads to your bottom planel, first of all most people trying it, removed them, because there wasn´t really a benefit.
Second if you use good thermal pads, the panel will get hot, doesn´t matter whether you have a cooler or not because to dispatch such an amount of heat you need a real cooling solution. Just watch this video:
And if your bottom plate is hot, your whole pc will heat up, so almost every component. So for the sake of just one or two degrees of the vram, which still will be to hot, you are heating up the entire Laptop, which could have other disadvantages as well.

 

So there really is no way other than adding heatsinks and creating the sort of air channels iUnlock made?

Why did he still put thermal pads on top of the heatsinks to touch the bottom panel?

 

He used the thermal pads to connect the copper sheet with the thermal pipes, he did that so no air could escape that way. And he used the thermal pads on the heatsinks in the middle to separate both airstreams (from the right and the left fan) in order to channel the flowing air in the right direction (out of the housing). That´s also the reason why only the sinks in the middle had thermal pads attached.

(first page)

But as GonZ0 said, it isn´t really an issue, in normal use you won´t get TDP throttling and it´s best to leave it alone. But if you want a solution that actually helps, it seems the only way is iunlock´s way of doing it.

 

If you can put thermal pads against the lid (use copper foil on the underside of lid Who touch pads). And use mod U3 cooler with proper fans for max airflow.

 

To be honest I'm not really sure what exactly is going on. I undervolted my GPU to run at 1544 MHz @ 0.812V, and it still throttles down to ~1100 MHZ @ ~0.7V. On GPU-Z it says "PerfCap Reason: Pwr."

It's plugged into AC power. I also have ThrottleStop set to turn off Turbo boost when the GPU is running. I'm guessing the VRMs really are overheating and not being able to deliver enough power?

Or is this just TDP throttling? My CPU stays at a constant 2.8 GHz while gaming (Turbo off), it's really just my GPU that throttles.

Edit:

What would happen if I make a thermal pad "bridge" to connect the VRMs to the heatpipe? (w/o touching the lid)

 

The heatsink will not be able to shift the heat obviously as it already struggles without people thinking they know better than the development team

 

Ahhh I see. Okay, thanks. Guess there really isn't anything other than iunlock's method once I get my hands on heatsinks and thermal pads.

 

I strongly advise against putting lumps of metal inside the laptop as it's going to void your warranty when it falls off, blows the mainboard and leaves short marks.
But it isn't my laptop so I don't care

 

Update: I opened up the laptop, cut off the copper thing blocking the middle air intake grill, and put it back together. I ran games with it on my usual laptop cooling pad, and the VRM Ambient (The first HWiNFO ambient reading) doesn't go above 67C. That together with my 1544 MHz @ 0.812 V GPU underclock, it didn't VRM throttle anymore.

Without the cooling pad, it just rises until 78C after which it VRM throttles. I suppose therefore the cooling pad's air reaches the VRM area.

It's a compromise I suppose. I don't game on this thing away from home anyway (away from my cooling pad), so for everything else I never run into VRM throttling anyway.

Also, I share your sentiments with extreme modding :/ I read about the PCB pressure/MOSFETs falling off and stuff. Glad I solved it with this slightly less invasive method. Should I need to RMA this thing, I'll just buy another bottom panel and jack the copper part I guess.

 

Sorry I think I was to late.

I wrote this before you answered @Carlos Resurreccion but you posted first so now it seems to be obsolet.

Are you shure it throttles because of the vrm? I am not shure whether I remember correctly, but in the the temperature observation thread for the 9550 somebody undervolted is CPU and had similar throttling with his CPU, maybe it´s better if you try it without GPU undervolting?
Have you tried a Benchmark without the bottom plate mounted ? Just to look whether cooler vrm´s solves your problem?
Some people had bad contact between heatpipe and the GPU, CPU or Vram and after bending it or using other thermal pads on the vram, the problems were mostly gone.

 

I doubt a Dell tech would notice the copper sheet tbh

Sent from my SM-G920F using Tapatalk

 

Yeah I think so because of two reasons:
1. GPU-Z shows "PerfCap Reason: Pwr" meaning it isn't getting enough power. "Pwr" is different from the other code (forgot what) that indicates it's reached its TDP limit. I'm assuming not enough power = VRM is throttling itself to avoid burning up into smoke.
2. When I removed the copper part blocking the middle air intake part, and put a fan there, the first Ambient reading (HWiNFO) dropped to a max of 67C (thus no GPU throttling w/ my GPU undervolt). Without the fan it reaches 78C then throttles.

I don't have a heat sensor gun thing :/

 

I played an hour of World of Tanks with and without pads. Thermal pads on CPU and GPU max temp was 77C on the hottest core. Using RealTemp, the core temp was 74-74-77-76. Without the pads, 79-79-80-79. Undervolting the CPU dropped the padless CPU by 1C across the board. I haven't tried undervolting the GPU yet.

I was using 1.5mm Arctic thermal pads rated for 6.0W/mK. I also put the laptop on top of a 140mm USB fan here: https://www.acinfinity.com/component-usb-fans/multifan-s4-quiet-usb-cooling-fan-140mm/

Putting the pads on without an external cooler will do nothing at all. Even with the 140mm cooler, the padded bottom cover gets pretty hot. It would rapidly become heat soaked without an external fan, thus removing any cooling benefit that thermal pads between the CPU and GPU and the bottom cover would provide.

I don't do any real gaming with mine, but if I did, I'd buy a spare bottom plate off of Ebay, and modify it for larger intakes and large holes and slots for the VRM area AND I'd build build an inch high cooler stand for it using slim 140mm fans wired to a decent power supply (USB ones are too reedy, even the good ones).

Even that's probably overkill. A CPU/GPU repaste along with undervolting the CPU and GPU probably covers the most ground in terms of cooling.

 

On a side note, would you know exactly why this thing has a copper sheet to begin with? Is it for radioshielding or is it just to prevent any part of the internals from touching the electrically conductive bottom panel?

Having done my slight mod by cutting out said copper sheet, I'm thinking of putting electrical tape on the bare aluminum it used to cover, but I'll still leave the middle air intake uncovered.

Though if the copper is essential, I'll just cut off the half blocking the air then put back the other half that merely covers the bare aluminum.

Lastly, if the copper was in fact used to electrically insulate the internals, wouldn't it then be a bad idea to turn it over like what @iunlock did?

What do you guys think?

 

It distributes the heat better than the aluminium does. A lot of laptops have this to move radiated heat away from the source. None of it is done for the fun of it and the black coatings to prevent shorts should it fall off over time.

Sent from my SM-G920F using Tapatalk

 

Ahh, that makes sense. I'll put most of it back then Thank you!

 

If I remember correctly you got rid of the copper as well?
Did you just remove the part between the fans or did you remove the whole thing?
Is it sticky on the copper side? Can I remove it with a hairdryer and put back later?
Just out of pure curiosity, I might try it without, to see how temperatures change.

 

I only removed the exact same piece iunlock removed (basically all the copper from the fan grill onwards towards the back hinge. I left the copper from the front up until the start of the fan grill.
Yeah, the copper has some adhesive that kept it stuck. I just peeled the copper off the aluminum. I'm not sure if it'll go back on with a hairdryer though.
It won't make a difference unless you have a fan underneath

 

This looks like a simple and effective compromise for your situation.

You could make a clipped or sliding plastic plate for the center grille when traveling to get close to "factory" airflow when away from your desk. You lose the heat transferring benefit of the copper foil but also don't have the risk of DIY metal floating inside the laptop.

 

You remembered wrong.

 

OK thanks

 

I recall someone wrote about doing this with some success, but without any elaboration of the "bridge" and the outcome... so you can consider noone really tested this. Some even higher-specced thin gaming laptops use this approach with additional heatpipes and likely more fin surface. Mind the XPS 15 doesn't have much cooling fin surface, so this would likely have an adverse impact on the CPU and GPU temperatures (just like diverting a part of the cooling air with iunlock's approach), so firstly you'd need enough headroom on those. And of course you'd need to be extra careful to avoid shorting something.

For a moment I though you were praising Dell all of a sudden but reading more carefully I see this need not necessarily be the case in fact, hehe, merely the others are claimed to be even more ridiculous...

I think the development team should have known better than leaving the VRM area without proper cooling, and than designing a cooling system so likely to end up thermal throttling directly when mounting and pasting the way they regularly do it at the factory.

 

I'm preparing to implement iunlock's mod following the instructions as closely as possible but i have a few questions:

1. are there any pros/cons to extending the cut for the center portion of the copper sheet so that it covers the entire heat pipe assembly?
2. does it matter what type of tape is used to seal off the airflow (i plan to use clear scotch tape)?
3. has anyone figured out an effective way to use the aluminum case for heat dissipation?
4. are there any other improvements to this mod or other similar hardware mods that i should consider to improve heat dissipation?

 

You might research some of the high-temperature & electrically insulated tapes like the 105*C stuff from 3M or something higher performing.

No.

If you research the threads you will a couple dozen people who tried and removed alu case sink mods. Maybe 2-3 people used weak thermal pads to just slightly heatsink the 9550 for specific games with some success.

The case bottom has very little capacity to absorb then transfer heat. It can help delay throttling a few minutes.

Also by heating the case bottom, GoNz0 noted you will superheat the intake air for the fans, killing the weak radiators.

 

1, you are removing something designed to wick away the heat further into the chassis.
2, you do not need tape as it already has foam to stop air going back into the chassis
3, yeah strap a 300 cfm cooling fan to the bottom of the case or it just doesn't work as the heat builds up until it suck in hot air and throttles, and the laptop is so hot you can't rest your palms in it.
4, buy a gaming laptop or accept it for what it is once you repasted.

 

Hi folks, last week my XPS 15 came and thanks to this forum I knew everything about it`s flaws, but I also learned how to fix most of them, now I want to share how everything played out.

First after checking for hardware fails I made a the BIOS update 1.3.4 and enabled speed shift via UEFI Shell.

After that I Benchmarked. Everything was tested in my basement, because temperature and humidity stay at 20°C and 55% the whole summer (I measured). Windows was untouched everything was the way Dell set it up, not even Internet was set up, so it won´t run updates in the background.I ran Heaven, Fire Strike and Time Spy.

After I tested with everything stock, I repasted with Thermal Grizzly Kyronaut and K5 Pro for the V-RAM for better contact. I also applied Kapton tape (see picture), so that no air could escape in to the housing, except where the vram sit´s in order to give it some airflow. I put another layer of Kapton tape on to the sticky side. I did that so that no dust would tstick there and block the airflow. (It was a mess took me around 3h to get the tape right) (Kapton tape is for 3D-printers and can be used up to 350°C and is non conductive)



After the modification I ran the Benchmarks again. This time in different ways.
1.without bottom plate
2.without bottom plate and fan under it
3.with the bottom plate
4.with the bottom plate and dust filters on the grill
5.with the bottom plate and the full grill uncovert (without the copper thing)
6.with the bottom plate and the full grill covert with a dust filter
7.with the bottom plate and the full grill covert with a dust filter and undervolting (CPU -135mV; GPU -150mV)

I did a lot more test, but undervolting is different with each CPU and GPU so in my table there´s just the final result. I did the "GPU undervolting" with MSI Afterburner, I changed the clock speed/voltage curve.



And here are my results:


So I hope this answers some questions!!!
Special thanks to @pressing, @GoNz0, @Rockstar75, @iunlock

 

The moment I "undervolted the GPU" via MSI AB, HWiNFO64 shows Performance Limit - Power = Yes, Performance Limit - Thermal = yes. But temperatures were lower, while benchmark scores were higher. Anyone an idea why?

 

because you undervolted springs to mind so it could obviously hold higher clocks?

Sorry but I didn't read your post above, I lost interest in the modding side as it brings nothing new to the table nowadays.

 

I meant, it wasn´t throttling (power and thermal) while I hadn´t undervolted the GPU. It had 73°C in Time Spy, with undervolted GPU it had 65°C and HWiNFO64 shows that it is throttling. But the GPU doesn´t throttle thermaly with 65°C. But HWiNFO shows that the moment I use the dedicated GPU, while it is being undervolted. (I tested this with just -25mV and only light use)

What I was asking is why HWiNFO64 is showing that is thermaly throttles although it obviously doesn´t ?
Is it because MSI AB is in charge?

 

I don't think afterburner will be doing anything other than taking some wattage off the GPU allowing the CPU to use more as the laptop has an upper limit it has to share out. You will still be hitting the wall and throwing the cap warnings if the application is demanding enough, and benchmarks are usually the reason you are seeing it as it is a synthetic benchmark not real world and is designed to punish gaming grade hardware. This is why I gave up on all that, repasted and got on with it

 

Do you have any proof that MSI Afterburner is actually undervolting the dedicated GPU? The GTX1050 right? Not the iGPU? Post a pic of HWINFO with the undervolt and your GPU voltage... I believe it runs stock around 1.050-1.075V... I can do 1890MHz on the GPU on stock volts! hehe!

 

Here without the undervolt:


And here with the undervolt:


Hope that´s proof enough.
I did it with MSI AB. Just press <Ctrl> + <F> and you get the Voltage/frequency curve editor and there you edit the frequency you want to have at a certain Voltage. I gave it the max frequency at 925mV and not at the stock 1075mV.

 

HOLY..... dude that's like the holy grail with what i've been trying to find... That is freaking awesome! I'm undervolted to 1.000V right now at 1797MHz and I barely hit 70C now in BF1. You're getting an even lower undervolt at 1797MHz??? That's crazy that Nvidia gives that much headroom...no wonder why the CPU/GPU combined run so hot..

 

I see on your HWiNFO64 report your PL Limits are stock. Search my post as XTU will allow you to boost those numbers say 5 or 10 watts.

That trick might knock off the limit flags. It made a huge difference in my 9550, particularly throttling for the ROG StressTest (which is not so stressful but really tough on the XPS).

A few notes:

- if you boost those numbers by a large amount your computer will ignore the changes
- XTU is buggy and resets values frequently so you need to recheck constantly (more than weekly)
- you must carefully monitor that max thermals stay below say 77*C on the vrm and well below 85*C for everything else
- this is quite dangerous in all cases

 

How do you "set" this undervolt? I just tried similar settings in my Afterburner, but whenever I benchmark or game, the voltages don't seem to set. No matter what, it always runs at the stock 1.075v.

Edit: Nevermind, I am an idiot and apparently didn't press the apply button hard enough.

Edit 2: Seems to work pretty well and help significantly against throttling. I wasn't able to get -150mv (with overclock) without freezes, but -100mv so far works. Have to finetune it some more.

Some statistics (didn't screenshot)

Valley Benchmark Stock Voltages:
GPU Temp: 77 degrees C
GPU max voltage: 1.063v
Core clock: 1.683 mhz
Memory clock: 1.752 mhz
Valley FPS: 50.7 fps
Valley score: 2121
Valley Min FPS: 21 fps
Valley Max FPS: 103.6 fps

Valley Benchmark Undervolt (-100mv)/Overlock (+100mhz):
GPU Temp: 74 degrees C
GPU max voltage: 0.963v
Core clock: 1.784 mhz
Memory clock: 1.752 mhz
Valley FPS: 58.5 fps
Valley score: 2449
Valley Min FPS: 19.2 fps
Valley Max FPS: 107.5 fps