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

How exactly did you guys figure out that there was a throttling happening because of the VRM temp?
Where and how does it appear?
Is it flagged somewhere in the logs of XTU or HWINFO 64 as VRM throttling?

I want to be able to distinguish Temp Throttling, Power Limit Throttling, and VRM Throttling.

 

@GoNz0 and several of the other guys here nailed down the thermal issues with the 9550 & 9560. Aida64 mislabeled that temp sensor when the 9550 was released so it was a bit of extra sluthing. GonZ0 is a professional laptop engineer and a lot of the people here are engineers or extreme overclockers.

The VRM throttling is misleading as power limit flag; best we can tell, it appears Dell programmed flag when Ambient sensor (1) goes to about 77C.

The issue here is that the XPS has very little space for the VRM components and for coolng the VRM. If Dell had space for better cooling or more phases, the VRM would run more power and cooler. As the mosfets get inefficient at high temps, things start to spiral out of control; getting hotter but producing less power. So Dell programmed the low 77C temp gate for safety. So all the VRM mods you see are principally to work around Dell's 77C temp gate (the mosfets produce more power at low temps as added benefit).

Not sure if that is an issue on the 9570; the VRM looks similar but I would guess mosfet technology has moved forward over the past 4 years...

 

My guess is that it's moved forward as much as the cooling that they used in the 9570 has... I bet the only reason they don't have VRM throttling is because there doesn't seem to be a huge enthusiast following where people are taking logs to see why their cpu/gpu is slowing down.

 

How similar is the 9560 to the 9550? and could i do this on a 9550?

 

Virtually identical if not identical. Yes you can.

 

thank you

 

Trying to exploit your knowledge guys for the new 9570.

1) What is the exact clearance between the mosfets and backplate?
2) What is the exact clearance between the chokes and backplate?

Did the chokes need cooling as well in the 9560?

I have seen @Eason mod staking 3 1.5mm pads between mosfets and the backplate. That implies that clearance at most 4.5mm.

Also, isn't it a bad thing in general to stack thermal pads?

 

hi @iunlock

I just bought an xps 9570 and it has the EXACT same issue as your xps 9560 had. The Ambient 1 shoots upto 103, sometimes the laptop shuts down when gaming, im assuming its because the FET reaches its ultimate point above 110 or something. I plan on doing your mod, could you please tell me the size of those aluminum heat sinks you used?

Dont want to order something too big or too small and waste money

Cheers

 

I realize this is a large thread, but that question has been asked numerous times. In fact it was asked around 10 posts ago...

https://www.ebay.com/itm/20pcs-Alum...MOS-IC-Chipest-Cooling-Heatsink-/221723027375

That's what I used.

 

In fact this was not typically going on with the 9560, it would instead start downthrottling the clocks with Ambient 1 in the seventies. But lately there have been reports of freezes and restarts also with the older versions, maybe Intel DPTF is reconfigured or fails?

 

Well... Interesting development for my 9560. Over a year ago, back in March 2017 I put coollaboratory liquid ultra on the cpu and gpu... Just within the past week I've seen a major increase in temperatures. I'm now hitting 79C on an undervolted gpu... Very disappointed. Somewhat not shocked by this since my other previous experiments using clu ended up with a dried out clu. I'm guessing the clu is completely dried up and somehow moving the laptop caused it to break the seal... So now I'm kinda back to square 1.

I'm going to end up taking the clu out and using mx4 that I have some lying around. I have some extra thermal pads adding pressure to the cpu and gpu right at the heatsink so I'm hoping this helps additional cooling without the clu. Will report back later. I'm also in the process of moving...so timing couldn't be worse

 

I guess LM drying up isn't unexpected considering that gallium is known to migrate into copper. I understand the process slows down with subsequent LM applications and isn't as fatal as with aluminium heatsinks

http://forum.notebookreview.com/thr...-compatibility-with-copper-heat-sinks.800890/
https://en.wikipedia.org/wiki/Galinstan

http://forum.notebookreview.com/thr...-works-why-it-fails-and-how-to-use-it.809332/
http://forum.notebookreview.com/thr...t-vs-cool-laboratory-liquid-ultra-pro.791489/
http://www.overclock.net/forum/63-o...-thermal-grizzly-conductonaut-73-w-mk-12.html

 

I have some questions about PL1 power throttling on the 9560. I've repasted and undervolted the CPU and GPU and padded the VRMs, though I haven't done the whole iunlock mod. I'm still getting throttling after some gaming despite apparently decent temps.When playing a game, my temperatures seem stable and acceptable - GPU, CPU, VRM all at about 60C, but then the PL1 power limit starts dropping. Here's a HWiNFO screenshot: https://imgur.com/a/ejRr2YJ I closed the game when I saw the PL1 limit start dropping.

Why is my laptop throttling in this case? I thought the PL1 limit was driven by temperatures; am I mistaken? With these temps, am I safe just throwing in the hotel and disabling the Intel Dynamic Platform and Thermal Framework?

Any help would be appreciated.

 

Long term max power is 45W.

 

I recently find some posts or information about this Intel Dynamic Platform & Thermal Framework as being problematic with Throttling issues.

Can we consider it safe to disable them when an almost complete thermal mod has been made on the laptop ?
In order to completely avoid the throttling.

 

No

 

edit wrong thread, pls delete

 

I'm pretty sure I'm using the same type/brand, but the bottom shell of the 9560 doesn't have a tight seal where those heat sinks are placed. If I look at the laptop from behind and with my eyes level to it, I can see that there's a gap. Did you have to sand/grind down the mini heatsinks for the bottom shell/cover to fit flush with the base of the laptop?

 

No, in fact I had to put a 1mm thermalpad just so they would barely touch the bottom case.

 

Hi all,
After some testing I'm observing some improvements after padding the PCH as well... Did someone tested that too?

Not that the max temperatures are lower, but throttling seems lighter and and fan isn't coming on too early.

Hope to hear from you!

Here is my build as of today:

 

I just re-did iunlock's mod on this new 9560 and I put heatsinks/thermal pad over the PCH because it seems to help the fan not kick in as much. So far I'm pretty impressed with this little i5 in terms of temps.

 

Finally found this thread after using my Dell XPS 15 for a year and living with its limitations. I have several questions:
1. How did you make the aluminium heat spreaders stay in place?
2. After cutting the stock copper liner, what did you do with the rest?

 

1. The heatsinks have thermal tape on the back that helps them stay in place. I also put thermal pad between the copper sheet and the heatsink to help them stay in place.
2. The rest of the copper liner stays attached to the bottom of the case.

If you have an extended warranty, I suggest picking up a used bottom cover instead of cutting the copper on yours, but that's totally up to you.

 

Is it feasible to leave the original copper sheet in place and use another (ebay...)?

 

Sure that would work too.

 

I have a 9570 on the way and want to attach heatsinks as well. Why have some of you chosen aluminum and not copper? I would think copper would be better?

 

Copper may be better but good luck finding ones that small. It is also heavier and more likely to detach when moving the system.

 

Good points, thanks!

 

Probably not in this case though. Copper has a high heat conductivity, which means it can move heat quickly from point A to point B (like from a CPU to the heatsink fins). In this case, you are not really moving the heat anywhere though (within the copper material). The difference between copper and aluminium's heat dissipation in natural convection is also fairly similar (as opposed to forced convection (a fan blowing directly on the heat sink) where copper would win).

The only other advantage copper has is more heat capacity, which means it takes longer for the material to get hot. But once it is 'saturated', it's cooling efficiency would be exactly the same as the aluminium heatsink. It just takes a bit longer to get to that point.

In other words: without proper airflow, you'd be paying extra for a heavier heatsink and nothing more.

 

I think I might actually have some time this weekend to take the plunge and disassemble my 9560 to remove the old CLU thermal paste and replace it with mx4... Ive seen close to 80c on the gpu and I'm beginning to think now is the best time, especially before my grad school starts back up again...

It's at least 10c higher then when I originally put the clu on the chips...

 

You should take pics on how the LM looks after being on there. How long since you repasted it?

 

I'll charge up my dslr and gopro for the taredown. Clu went on back on the end of February 2017. So it's almost a year and a half old.

 

Oh dear goawd..... this CLU is AWFULLL!!! I'm actually amazed that my laptop didn't die this entire year and a half I had CLU on the GPU and CPU...


Full size pictures here ----> https://flic.kr/s/aHskUTACyG




The cooler has some gallium migration into the copper and I cannot remove it even after sanding the copper base. It is what it is...

I changed out the thermal paste to MX-2 and I have very similar results. I'm guessing this is probably the case because most of the CLU has completely squeeze away from the die this entire time. What a PAIN! It took probably 30 Q-tips and cotton balls to clean off the CLU... Quite possibly the worst thing for a laptop... Yes, the cooling performance is good, however, it seems that even with me doing the application perfect back in February 2017, the CLU somehow left the die partially exposed... oh btw, I'll get pictures later but the CPU seems to be permanently etched with the gallium/indium combination that's in CLU. GPU fared great but CPU seems to have taken a hit. No ill effects from it.

 

That CLU dripped everywhere. Thanks for the follow up here.

 

So after some more testing the mx-2 paste I just used is definitely better that the clu I had on previously. I'm guessing the better temperatures is attributed to the 4 thermal pads that constantly push down the heatsink via the bottom cover. The elasticity of the thermal pads is actually really good and I've been able to keep constant pressure with no ill effects from flattening the thermal pads. They're basically acting as a spring with thermal properties. The silicone pads are actually similar texture and softness as sorbothane which you can pick up from McMaster-Carr. I've use those in the past for different applications because of their great ability to maintain shape under pressure.

If Im not doing something very demanding I'll usually keep my gpu at underclock status with 0.900v on the core and 1680mhz clock. Valley benchmark tips out at 65c max and is typically around 62-63c at those clocks. For overclocks I keep the core voltage at 1.000V and can go upto 1860mhz. That tops out at 73c right now with the new paste.

 

Well the thermal conductivity of liquid metal is probably 4-8x higher than that of a good non-conductive paste. But if the gap is as thin as possible and even, normal paste should be good enough. The thickness and elasticity of the VRAM pads matters regarding heatsink alignment and whether the gap is as narrow as it can be all over the surfaces.

Migration of gallium into the heatsing is a known phenomenon, hence it is known that LM may have to reapplied after a while and that it sometimes "hardens" due to the change of the composition.

 

Thanks @Philaphlous . Interesting results. Do you have a photo of the four pads you are using? Do they make the cover really hot?

One item I might worry about with higher pressure on BGA chips is that the soldier balls prematurely fail. That might be more of an issue if you are putting the laptop into a bag frequently and the case bottom flexes a lot (putting pressure on the chips).

Those type of soldier ball failures are very rare on the 9550 and 9560. I don't remember any post here but maybe a handful of motherboards that died could be from packaging. . .

In any case, if you face some catastrophic failure, you could just reball a chip following Louis Rossman's great videos on YouTube. He sells good flux and paste for this job:

https://store.rossmanngroup.com/ind...elpful-tools-extras/flux-soldering-tools.html

Hot air stations:

https://store.rossmanngroup.com/index.php/soldering-diy/soldering-stations-hot-air-rework.html/

 

I'm not worried too much about movement of the base when carrying it that'll effect the thermal pads and push against the dies even more and cause there to be any possible physical movement of the soldered joint. I've got pictures of the pads in my flickr album. I cut 2 pieces in half so a total of 4 so I can have 2 over each chip and on both heat pipes. I would be more concerned if I ended up doing just one big pad, trust me that crossed my mind. But since the pads are relatively small. There's a great chance that the pads are more so using their elasticity to keep light to moderate pressure on the heatsink from the case bottom.. Plus, with everything screwed down, the case doesn't flex much at all so it's not like you can push it right in those 2 spots and theres a good bit of deflection.

Heat on the bottom of the case hasn't really been much of an issue.. I know weve talked alot about how the heat can migrate and stay on the case and cause there to be poorer cooling performance over time but I haven't really noticed any degradation in performance. I gamed a bit on my lap with it lastnight and whereas the bottom center of the laptop case gets pretty hot it's not unbearable. It'd be better suited on a cooling pad which will actually work like it's supposed to. Overall I'm more pleased with the standard paste on the chips. There's less chance of a short, much easier cleanup, lasts longer, and ends up having better cooling performance even with a way more superior paste.

I think I'm going to end up keeping the gpu at 1.000v which gets around 1840-1860mhz. I also oc the memory +240 which gives me a little boost in performance. Overall I get about a 10% improvement in performance with my overclocks over stock. I'll take that but wish it was closer to a 1050ti... Oh well. Im sure these new 11** series chips coming out are really going to age the mobile gpus we see today. I bet 3-4 years from now they will be laughable...

 

Strange. I guess mine are too tall

 

I understand iunlock's design mates the heatsinks to the copper foil which is mostly airgapped to the backplate. I think if the 9570 PL1 throttles with the DIMM sensor at 63 and one pads RAMs and the sensor to the backplate in is important not to take much VRM heat to the backplate.

 

Hi everyone so i began testing how my xps behave when playing certain games and so far i come to this:

Before anything i have:

- Undervolted CPU via Throttlestop -0.125
- Disabled Turbo Boost in BIOS
- Updated to the newest BIOS ver. via Dell Drivers website
- Enabled Speed Shift with value 128
- Downloaded Dell Power Manager and set it to Ultra performance when playing games
- Updated DPTF to the newest version
- Used cooling pad on 100%
- In Nvidia Control Panel changed the “Maximum pre-rendered frames” value to “2”
- Removed stock copper liner on backplate

How does it behave when playing BF3 for 1 hour?

- At the beginning the performance was flawless
- After several minutes (3 or 4 minutes) performance dropped but still playable at least
- MAX CPU temp was : 75C
- MAX GPU temp was: 78C
- First Ambient Temp: 89C <------ ou ****!


I think this is throttling due to VRM area not cooled properly as discussed here before. The first ambient sensor heats up very quickly. I think i will try thermal pad the VRM area and test it after that.
Just want to ask you where to place them please?

- Should i pad the VRM to the case or to the heatsink?

Thank you, any tips are welcomed.

 

Sounds like bad VRM heat. I renamed the ambient sensor to VRM because that's what it is. Add some thermal pads and cal it a day... they won't get over 80C with pads.

 

Try both options. We need test data to see if the heatsink one is effective enough or not, since the reports I've gotten vary.

 

Hi folks !

After several attempts, I finally made the all mod for the temperature with rads, thermal pads, and the copper slice.

Additionnally I repasted both the CPU and GPU with Kryonaut, and I'm applying settings with ThrottleStop with -120mv on CPU, and -70mv on the dGPU.
I've also disable the Intel DTPF in the device manager.

Plus I'm applyling a profile in Afterburner to ramp in a straigth line the gpu clock.



The temperature looks very good, I just made the Fire Strike test

However, as I experience it while gaming on Fortnite for example, I'm still having GPU Throttling, even during the test as I saw it in afterburner.

Why it happend and how can I prevent it ?

Thanks !

 

I change some things but no clear effects :

I change the curve for afterburner



But during a game session, without any apparent reasons, this happens :



Any ideas ?

 

You seem to have max frequency well over 1800 MHz. try to limit it to 1797 or 1809 MHz

 

VRM throttling...only way to explain it. What's your ambient 1 temp in HWINFO? THat should give you some indication of what's going on with the throttling...

 

Must be one of those ambient temperature that gets to 79C and since then it starts throttling... What is the Max temperature of those vrm please? Is it okay to uninstall DPTF and watch the ambient temps so they dont get more high than lets say 95?

 

Hi everyone,

I am wondering how is conductonaut holding up for all you guys that re-pasted using liquid metal?

Is re-application needed after 1 year of use?

Thanks

 

Hello all again, Ive tried repadding VRM area to the backplate of xps…. The temps looks good when playing Battlefield but after playing for 5 minutes.... I get lot of lags again... So I looked at the HWinfo if its the ambient sensor again but no its not. I am not sure what is causing this lag? Does anyone know pls? Here are temps monitored https://imgur.com/a/kMyA1l4