Correct positioning of VRM thermal pads in XPS 9570

Hey guys,
I am new here, so forgive me if I get anything wrong. There has been a lot of discussion about the XPS 15 VRM temperatures. However, for the XPS 9570, I haven't seen any pictures yet on where to place the thermal pads to connect the VRMs to the back plate (to be used in conjunction with an external cooling pad). The question is not trivial, since the VRM placement in the 9550 and 9560 (for which suggestions exist) is different than in the 9570 (for which so far I have not found any suggestions on where to put the pads / where the VRMs are). So if this is my XPS 9570 motherboard:
https://www.dropbox.com/s/7pu7lt66bu4fav6/20190109_231401.jpg

Then this is where I would put the thermal pads:
https://www.dropbox.com/s/vohur2y5rdowvoz/20190109_233426.jpg

Would that be correct?


Cheers and thanks!

 

Hum, "correct"? I think it would be correct if Dell mated the VRM mosfets with a heatpipe to cooling fins. This pads are just ad-hoc patches that don't really work well but are presumably better than nothing*.

I haven't seen FLIR images or heatgun measurements for the open 9570 under load to tell which mosfets need to be padded most, but I guess you have left out a few. I assume you intended to leave some room for airflow, but mind there really isn't any, unless you divert some cooling air from the fins (check iunlock's mod).

Also some report that power limit throttling in the 9570 kicks in when the DIMM sensor reaches 63 (may depend on BIOS and DPTF versions... check using HWinfo64). If this is so, the VRM mosfets may be better "bridged" with a pad sheet to the heatpipes without touching the backplate, and you may consider padding RAM (incl DIMM sensor) to the backplate to offset throttling by a few degrees. This is still a cheap and simple surgery

*It appears that high-thermal-conductivity pads may be worse than nothing, because in the absence of a cooling pad the backplate may in the long run overheat and the hotter input air no longer cools the CPU and GPU well.

 

Hey,
thank you for this!

I am sure we can all agree that a better build would have been preferable. But in the absence of DELL doing its job, I guess we have to do with what we got. I also assume that we are all speaking about this solution in conjunction with an external cooling pad (it definitely makes no sense to just heat up the backplate). Now, I would consider myself as a technical intermediate, so putting pads on VRMs is fine with me, but bridging the VRMs to the heatpipe seems a bit too advanced for me. So if we stick around with the VRM-padding-to-backplate-plus-external-cooling-pad solution.

Then my stupid question no 1 would be: VRM = MOSFET? Then: what else would have to be padded? Anyone got any ideas?

I will also check for the DIMM sensor. But stupid question no 2: If I am padding the DIMM sensor itself, will that not be a problem, since I only prevent the heat from getting to the sensor, rather than letting it dissipate (to the backplate)? This would mean that I prevent the computer from recognizing that it actually does get too hot (so that it SHOULD throttle), rather than prevent the reason for the throttling itself (the heat buildup). Or would that not be a problem because the threshold of 63 is simply too low?

That being said, I am actually not that unhappy, I can play Battlefield 5 on 1920*1080 with low/medium settings and while there is power limit throttling, the computer manages to stay above 2,2 Ghz, so the game is still absolutely playable. Maybe this is the limit of the GTX 1050 ti anyways?

 

> in conjunction with an external cooling pad (it definitely makes no sense to just heat up the backplate).

On the 9550 & 9560, throttling from the first Ambient sensor near the VRMs, just shifting a little bit of heat to the backplate helps a lot, even if you don't even lift the rear of the laptop. An external cooling pad isn't essential; does help but not tremendously.

> putting pads on VRMs is fine with me, but bridging the VRMs to the heatpipe seems a bit too advanced for me

The other option isn't much more involved, folks seem to just get the Arctic 5x5cm pad (1mm?), cut it a little bit and slap on (not 100% sure if it sticks well enough). Light years from iunlock's thing.

> VRM = MOSFET?

I recall VRM probably stands for voltage regulator mosfet or module? Anyway the mosfets, thin black slices, are their power switching components that seem to produce most heat.

> If I am padding the DIMM sensor itself, will that not be a problem, since I only prevent the heat from getting to the sensor...

Best leave it alone before you are sure that temperatures are better because of VRM padding and you are sure that it triggers unnecessary throttling.
The idea is to cool down this location by few degrees by padding to the presumably cooler case (if you also pad VRMs to it it may well get warmer not cooler). It would still throttle if overheated, just later.
It is possible to kill throttling by removing Intel DPTF, but I wouldn't recommend it.

> while there is power limit throttling, the computer manages to stay above 2,2 Ghz, so the game is still absolutely playable. Maybe this is the limit of the GTX 1050 ti anyways?

Many gamers are okay if the CPU gets throttled but the GPU runs relatively free.
Have you undervolted both (Throttlestop and MSI Afterburner), repasted (if there is direct throttling)?

 

I thought I had posted this yesterday but maybe I didn't hit the post button.
Here is where I had the most luck sinking after several trials:
http://forum.notebookreview.com/threads/dell-xps-15-9570-benchmarks-temps.817970/#post-10741757

 

Hey custom90gt,
this is exactly what I (and I think many others) have been looking for. However, it is also interesting to note that, if you have a look at my photos, my motherboard - including the placement of the VRMs apparently looks different than yours (even though we both own an XPS 9570). Weird, you would think that Dell uses the same motherboard for all XPS 9570s. I have the i9, maybe that explains the difference?

And for _sem_ (also so that everyone else can get it): Your preferred solution would then be to just cut out a thermal pad that touches all VRMs and then either the backplate or the heatpipe?

As a side note: thanks for this awesome help and I will try undervolting the GPU tonight, regardless from anything else.

 

Ah I didn't look at your picture. You have the newer revision motherboard with extra VRMs. Based on the one person on Reddit that had the same board, you likely won't have to sink the VRM at all (just a guess on my part based on his scores). Sadly without a lot of testing, I am not sure what the best advice would be for you. My hope is with the redesign and change to the VRM, temps will be lower VRM wise.

 

Different from what one might hope, I suppose the padding of the VRMs still seems necessary, as the newer version does not appear to be that much cooler. I can say this because I happened to have had an earlier version of the computer (also with i9) and I sent it back because of crashes and high temperatures. I have then reordered the computer for a second try (same configuration, but apparently revamped board).

On the old motherboard (unevised board, same config) "temp 1" in hwinfo (I think this is the VRM sensor) went up to about 125 degrees Celsius when simultaneous stressing the GPU and CPU. After padding the VRMs, they were comfortably in the 90s. I then sent the computer back for various, unrelated reasons however.

I then got my new XPS 9570 (which I am using now): Without padding, it stayed below 120 degrees, but not by much. So an advantage over the old configuration, but by only 5-10 degrees. Now with the pads applied as shown, it did not go above 88 degrees (but some power throttling occured nonetheless). So it seems the new VRMs are a bit cooler, but not by much.

By the way, for all my XPS VRM temperatures never went through the roof as they did UNLESS the computer was a couple of centimeters above the ground. Just placing the computer on a hard surface meant GPU and CPU would get hotter, but not the VRMs. So if I used a cooling pad, the CPU and GPU would stay relatively cool, but the VRMs would get super hot. With the current padding, everything seems fine now (power throttling still occurs, but maybe unavoidable, right)?

Stupid question no 3: Is it a problem if I pad some but not all VRMs as I might have done?

 

Yeah padding those are totally fine, you can always play with which ones to pad and see if that makes a difference. Sad that the new revision doesn't really help much.

The power limit throttling is set by Dell but if you're comfortable with it, you may be able to remove the limit with throttlestop:
http://forum.notebookreview.com/thr...ongfang-gk5cn6z.815943/page-274#post-10843849

Just watch those temps. With the increased power limits comes hotter cpu and VRM temps...

 

> Stupid question no 3: Is it a problem if I pad some but not all VRMs as I might have done?

As already written, this is difficult to tell without measuring temperatures of the insides of the open laptop under full load with a FLIR camera or by moving a beam thermometer over individual components, opposed to the few measurements in this area available online (1st Ambient, DIMM... not sure if any other is near). Considering that we don't know which VRM feeds what, at what power...
I recall such measurements for the 9560 are on a pic in the iunlock's thread http://forum.notebookreview.com/thr...ures-benchmarks-xps-15-9560-kaby-lake.802345/
Mind there are two issues here:
1st, to fix local overheating of the mosfets and their immediate surroundings
2nd, to actually get the heat accumulating during prolonged heavy loads from the VRM area out of the laptop.
The backplate padding approaches mainly address the 1st, while the 2nd only to a low degree, because the backplate can't dissipate much heat, even with a cooler stand. The bridge padding variant might do this better, but not sure if the pad can conduct much heat side-wise, and the CPU+GPU temps might rise a couple of degrees (perhaps not good with the GPU limit lowered to 74).
The iunlock approach tackles 2nd too, by creating more internal "fin surface" and redirecting some of the cooling air from the fans over there and then out via the narrow central grille. But it is much more involved.

 

@custom90gt I didn't know throttlestop could do that! Is it a new feature?
Have you tried it? Can you keep the power limit above 56w? Can you avoid the dips to 42w during stress tests? In short, is it able to bypass DELL power management?

 

Hey,
it is a new feature. Sadly I haven't tried it on my XPS since it's boxed up and listed on Ebay, but I did try it on my Overpowered15 laptop and it worked. I may even try it on the X1 Extreme too.

 

@Mulgul the real question is: does your computer throttle because the CPU/GPU get hot or does it throttle because the VRMs get hot?
If it doesn't throttle for VRMs, why bother padding?

 

@custom90gt It works! I have been able to set custom PL1 & PL2 limits and also set the duration of PL2. Did a few cinebench runs to verify the behavior.
With this trick, you could set say 30w for the CPU during gaming and avoid disabling the turbo.

I have to do further tests to see if under certain conditions dell power management will kick in and change those values.

 

The running assumption was that it was throttling in games, though Mulgul didn't specify what he checked. I guess it is natural that one should check for direct throttling first, undervolt, repaste if required.

This PL adjustments look cool indeed... I guess with a properly prepared XPS it should be safe to run the CPU alone (GPU idle) long term at the default short-term turbo limit as long as the temperatures look reasonable... While for the gaming scenario restricting to 30W you suggest is likely a better idea.

> I have to do further tests to see if under certain conditions dell power management will kicks in change those values.

I recall one could change power limits in XTU but they didn't stick, maybe TS keeps reapplying them?
I recall there was a suggestion a while ago in Lenovo hack circles to script XTU settings, maybe this https://medium.com/@n4ru/45w-performance-from-15w-kaby-lake-r-d8d5e7ea4fad

 

Hey guys, so just an update on the results: Whereas before I would have to reduce clockspeed to about 2.2 Ghz to avoid power limit throttling, I can now increase the FPS by an additional 5 percent and there is absolutely no power limit throttling any more (constant at 2.9 Ghz, undervolt 130). Could also be because I now also undervolted the 1050 ti (everything flat after 850 mv). BF 5 with 1440p at 75 % resolution absolutely no problem any more, which I am pretty sure is all you can expect from the GPU. So thank you for your tipps. Some tweaks and this XPS 9570 can be turned into a great gaming machine.

 

Howdy, someone with a 9560 4K max spec machine here.. I just so happen to have the equipment that you all seek.
Pretty sure you missed one of the VRM's there, I've highlighted it in red in the attached image.
Outside of the three VRM phases, there's nothing else to pad as there isn't really anything that gets hot outside of the heat soak those three IC's disperse.

 

Hi guys,
so can anyone confirm @republicofsam, that it would make sense to apply a fourth padding where he indicated? I am still a bit worried that only padding some but not all VRMs will let the non-padded VMRs overheat, while the sensor fails to trigger an alarm (and fails to downthrottle) because the many VRMs that are padded do not generate heat. In this case, one VRM (the non padded) could overheat unnoticed, or have I gotten something wrong? However, for now I just trust @custom90gt who suggests that only padding some VRMs is ok, so I think I will stick WITHOUT padding the last sensor indicated by @republicofsam.

This is because with the indicated padding (not the VRM in red), I am very happy. I can confirm that even if I do not undervolt the GPU, power limit throttling has become a thing of the past. It must have been triggered by the VRMs, which used to reach close to 120 degrees before padding and are now below 90. Config all in all: i9 undervolted by 130 mv, VRMs padded as indicated and GPU forced to reach 1733 MHz at 862 mv (not necessary to avoid power limit throttling, but still nice and 1733 Mhz should be safe as long as the voltage is that low, right?). I have never repasted anything so far.

Anyways, maybe others who have power limit throttling can try the same and see if it works for them as well. I can confirm so far that this is a "solve all problems" solutions for me so far. Maybe others want to start with the padding, plus CPU undervolt 100, plus forcing the GPU to operate at about 1620 Mhz with 900 mv, just to be safe. This should work on most systems, then you could work your way up the silicone lottery. Glad to see that this machine can be made to work after all and thank you so much for your help, @custom90gt, @republicofsam and @_sem_ ! Would be very interested in results from others!

 

Oh, and one more thing: all of this only works for me in conjunction with an external cooling pad, but then it's perfect.

 

I've recorded a thermal analysis for you, I have to edit it some parts of it tonight a convert over the video formats for sanity purposes when I get time. But you can see lot's of valuable information. Please expect something tonight or tomorrow morning... Recorded on FLIR E8 @ 15FPS.

 

@Mulgul, I think the main issue with sinking all of the VRM is that the backplate cannot handle all of it. If you're on a cooler than perhaps it would be best to just sink them all. All of my conclusions came from days of testing and hundreds of HWinfo logs. Also remember you have a different VRM setup than me and @republicofsam, so any of our results don't totally apply to you.

 

Ok, let's see what comes out from @republicofsam 's analysis (cool stuff dude). However, as long as only padding some VRMs is not dangerous, as I said I could not be happier so far. But just as an aside, with my external cooler, my back does not get super hot, so it seems to be able to handle it. Really curious what @republicofsam is going to find!

 

@republicofsam Thanks for your work!
EDITED:
Looking closely at your pictures. It seems that you have the first revision of the motherboard.
In your pictures, only the MOSFETs from the CPU side are getting.
What stress test are you using?

 

Here's a better idea of what my power regulation circuitry looks like on my machine. Again, I'm aware it's slightly different layout.. That said I'm positive it's a good starting point as the 9570 is going to be heavily related.
I haven't even started pulling part numbers yet to see what's doing what. Lots of footage to go through. My exam was simply a CPU only test with wPrime 2.10 1024M iterations. A combined test would be in the works.

PU1202, PU1200, PU1201 are the item's we are interested in here.

 

if you run heaven you will see the MOSFETs gpu side get hot.

 

I guess I never realized how terrible the combined load VRM situation really was. Holy smokes, that's borderline dangerous from my results. I actually got a reading of 118*C on one of my other tests.. Geepers. Those GPU MOSFET's are outrageously hot (so hot in fact they nearly overheat the CPU MOSFET's!), can't believe Dell engineers would leave them like that.

 

Hey @republicofsam,
many thanks for this! There are even stories (which I can confirm) about the VRMs reaching 125 degrees. Seems pretty unsafe, right? Anyways, I am sure this test was long overdue for many XPS 9560 owners. Let's hope someone gets around to doing the same thing with the new XPS 9570 motherboard!

 

@Mulgul Anytime bud! Let me know if there's anything else I can do with my equipment. Nice to find new uses for it.
I think I'm going to open a new thread and cross post this on the XPS forum directly so people can find this a bit easier than digging through a thread of a different model machine.

 

@republicofsam there is a 9560 thread already http://forum.notebookreview.com/threads/xps-15-9560-owners-thread.800611

 

I don't see any reason that @republicofsam shouldn't create his own post and not get lost in the sea so to speak.

 

So, just a quick question: In the image that is attached, it would make sense to also pad the 5 additional black chips indicated in red (hoping that an externaler cooling pad can cool all of this)?

 

Yes

 

How much Thermal Padding did you end up using?

I want to pad my XPS as well, but not sure how big should the pad be for when I cut it been able to cover everything.
Could please share what pads did you used and the size?

 

Hey @Asiier
attached you find a photo from my padding:
https://imgur.com/a/iEYGKoV

Because of what @_sem_ and @Dialup David said, I padded the remaining VRMs. I use 3 x stacked Arctic Thermal Pad 6.0 W/mK 1.5 mm. Works for me so far!

 

And how big was the thermal pad that you bought? This ones 145 mm x 145 mm x 1.5 mm?
Or did you needed to buy more than one fore cover up everything? Since you stacked them

And are you planning on repasting?

Thanks you for the picture!

 

Hey @Asiier
one pad as you describe is more than sufficient, actually, you will probably need less than about 10 percent of it. I did not repaste since I have the feeling that I have pretty low temperatures when idle (lower 40s, upper 30s) and since I can clock to over 4 Ghz without my processor getting super hot. I might do so in the future, but maybe Dell got one thing right once

 

Hey @Mulgul

I have exactly the same DELL XPS 15 (i7-8750H) motherboard design as you. I basically followed your steps and ended up having the same VRM pads as you (5 areas padded). I also repasted the CPU and GPU with Kyronaut and undervolted my CPU (-250mV) and Cache (-125mV).

However, you say your idling temperatures are in the low 40s to 30s. Mine are basically at 50°C (or 49°C). Do you guys think I did a bad job at repasting? I followed the 8auer tutorial on the Kyronaut, using the spatula-thingy.

@Mulgul what are your clock speeds under load? ADIA64 (with the undervolting etc.) shows me that my clockspeed is about 3.5 GHz, and I didn't manage to achieve higher speeds as I run into "Power Limit Throttling" (seen in Intel XTU). I do no longer have real thermal throttling.

Furthermore, before the 1.81 Bios Update I felt as if the Fans ramped up quicker and went faster. Now half the time my laptop stays in quiet during even the stress test..

Looking forward to hear from you!

 

Hi @YZIOER ,
I guess what went wrong ist that your undervolt does not stick. I cannot imagine that an undervolt of -250mV on your core actually sticks, typically people manage an undervolt of maybe 120 or 140, but definitely not of -250 mV. So I think this explains your higher idle temperatures. However, I guess the idle temperatures are not that important anyways (you might simply live in a warmer country for example). As long as your computer is fine under load, you should be ok. I would suggest you use e.g. Hwinfo to check whether you have really undervolted your CPU by as much as you say and then possibly redo the undervolt.

 

@Mulgul, HWInfo and Throttlestop show me -250mV, and -125mV on the cache. It seems to work. So far no instability issues, I ran the TS Benchmark + AIDA64 (for about 30mins), and gave Subnautica a test run (20mins).

Strangely now I saw the first time core temperatures of around 43°C.

I still wonder how you get such low idle temperatures, what is your room temperature? I think mine is about 22-24°C, so pretty standard.

Did you undervolt the Nvidia GPU, or the Intel GPU (Throttlestop allows you to do this!).

Cheers
yzioer

 

As far as I know, undervolts of the cache and CPU are linked in modern CPUs. Essentially, your lower undervolt on the cache causes the CPU undervolt also to be -125mV.

 

Hello @YZIOER,
I am not sure, maybe I am just lucky. My room temperature is pretty similar to yours. However, my impression is that idle temperature is not important as long as your temperature under load is ok and 3,5 Ghz seems like a good constant clock to reach under load. If your temperature goes down to 43, that seems perfect to me.