Hi. We use cookies to ensure that we give you the best experience on our website. We do not collect your personal data.  I agree  Find out more  About cookies

How to Fix Throttling on the Dell XPS 15 9570 / 9560

By Douglas Black , last updated on August 10, 2018

Note: title has been changed to reflect that the XPS 9570 has the same problems and the same solutions as the XPS 9560. I’ll update the contents if needed when I get a 9570 to test (already did, here’s the review), but for now these tweaks will work on both builds.

So, you’ve already maxed out your Dell XPS 15 with 32GB of DDR4 RAM and the best 1TB NVMe SSD for the money—now your XPS 15 is as good as it gets, right? Nope, not yet!

Without some hardware and software tweaks, your brand-new XPS 15 won’t be able to perform anywhere near its potential due to throttling.

In this article we’ll tell you how address this issue, but be aware that while this procedure works and is safe when done correctly, it may void your warranty or damage your computer if you don’t know what you’re doing. We’re just providing a guide and cannot be hold accountable for your actions, so proceed at your own risk.

Before you begin the journey of optimizing your XPS 15, you’ll want the following: 1.5mm 6.0W/mK thermal pads, Thermal Grizzly Kryonaut (best thermal paste at the moment without resorting to the liquid metal variants), some highly pure rubbing alcohol and some microfiber cloths.

As mentioned in our detailed review, the XPS 15 9560 with the Core i7-7700HQ processor (as well as the 9570 with the i7-8750H update) is prone to two types of throttling:

  1. Thermal throttling of the CPU or GPU (generally the CPU) when temperatures get too high
  2. Voltage Regulator Module (VRM) throttling caused by it getting too hot and being unable to deliver enough power

The first type of throttling is quite straightforward to all but the most uninitiated: the CPU and GPU have temperature limits to prevent damage. Exceeding the temperature max (100C for CPU, 97C for GPU) will cause an instantaneous shutdown—but you should never see temperatures anywhere near that as the components will throttle their performance long before they get to that point.

There are softer limits than 100C and 97C for the CPU and GPU respectively, however. The GPU will reduce its clocks to keep its temperature below 78C, and the CPU will dynamically reduce its turbo-clocks based on temperatures and power consumption.

The second type of throttling is also the result of heat, but because people generally look at CPU and GPU temperatures alone when benchmarking, it went undetected for a long time.

The VRM’s job is to convert 5 or 12V power from the adapter or battery into much smaller voltages to feed the CPU and GPU (generally 1.5v or less). A VRM has several components, but we are primarily interested in only two: MOSFETS (short for metal-oxide semiconductor field-effect transistor, but you don’t need to know that for any reason except trivia) and chokes. The MOSFET is responsible for switching the voltage down to a manageable level to feed the CPU and GPU; the chokes stabilize the current.

In this image (courtesy of iunlock of Notebookreview), the MOSFETs and chokes have been identified with their temperatures under load. Throttling occurs around 78C.

To get your XPS 15 to run at its maximum potential, we will need to solve both these problems. First, let’s take care of the CPU and GPU heat directly. For this, you’ll need your Grizzly Kryonaut, 99% rubbing alcohol, and those microfiber cloths.

  1. Use a T5 screwdriver to remove the 10 Torx screws around the edges of the bottom panel.
  2. Open the service hatch and use a Phillips screwdriver to remove the two Phillips-head screws.
  3. If this is your first time opening your XPS 15, it will take a bit of effort to take the bottom plate off, as there are many small clips that keep it attached. Get your fingernails or a plastic card under the sides of the base near the hinge where the casing is most robust, and apply even, steady pressure to pull the bottom off.
  4. Remove the battery connector by pulling gently.
  5. Use your Phillips screwdriver to remove the five screws of the heat sink assembly.
  6. DO NOT PULL ON THE PIPES TO REMOVE IT. The heat sink assembly bends very easily. Carefully remove the heat sink assembly by getting your fingers under the upper edges of the left square (this is the GPU’s heat spreader) and gentle pulling upward on the top edge. The assembly should come up fairly easily once you overcome the glue-like suction of Dell’s abominable paste-job.
  7. Use the microfiber with the rubbing alcohol to remove the existing thermal paste. You’ll want to clean both the CPU and GPU dies as well as the heat spreader. Keep cleaning with the alcohol until you don’t see any grey marks on the cloth after rubbing.
  8. Put a little (half the size of a grain of rice) dollop of the thermal paste onto each die. Don’t worry, it will be pressed flat by the pressure of the heat spreader.
  9. Carefully replace the heat sink assembly flat onto the dies. Keep applying firm downward pressure over the heat spreader areas with one hand while you reinsert the screws with the other.
  10. Tighten each screw only a little at a time, taking turns tightening the screw diagonal to it so the pressure on the heatsink is uniform.
  11. Reconnect the battery.
  12. Place the bottom case cover back on the unit. No need to screw it back on yet, though.
  13. Turn on your laptop and run some tests! If you’ve done it right, you’ll see a pretty good reduction in CPU and GPU operating temperatures.

The next step is using software called ThrottleStop to reduce the power consumption of your CPU. This can reduce the load wattage needed for the CPU by 10W easily.

Generally, all the i7-7700HQs on XPS 15 9560’s I’ve seen can undervolt to -125mv for core and cache. I run -125mv on core/care and -75mv on the iGPU. i7-8750H CPUs on the XPS 15 9570 do well at around -120 mV, but I’ve seen some that were onyl stable at around -11- mV to -110 mV. The link above explains how to use ThrottleStop and how to test for stability.

The final tweak to take care of your CPU and GPU temperatures is to go to the Nvidia Control Panel and change the “Maximum pre-rendered frames” value to “2”. This will prevent the CPU from needlessly being overtaxed to send data to the GPU.

The next step is dealing with the temperatures of the VRM head-on. Take another look at this picture:

Our goal is to cool those MOSFETs with the highest temperature (above the heat spreader) so they don’t hit their throttling temperature. How? We will be using stacked thermal pads to connect the MOSFET to the aluminum bottom cover. Why? When the right type of pads are used, the bottom cover will wick heat away from the MOSFETs.

There is trick to this, and it explains why we use lower performing 6W/mK thermal pads instead of some very high quality ones: if too much heat is transferred to the case it will actually end up heating up the VRM instead of cooling it. This is what will happen if you use an extremely conductive (16W/mK) thermal pad, and it’s why we want the 6W/mK pads for this purpose instead.

The mod is fairly straightforward: cut out small vertical strips from the thermal pad and place them on top of the MOSFETs above the heat spreaders. If you are using 1.5mm thick thermal pads, you will need to stack 3 of them on top of each other in order to reach the case. You want to leave as much open space around the pads so what little airflow there is doesn’t get impeded.

That’s it! Once you have stacked your little padded pillars over the MOSFETs and put the case bottom back on, you should be able to run any game or benchmark without VRM-induced throttling.

After following this guide, your Dell XPS 15 9560/9750 should now be a finely tuned beast of a machine, able to leap mountains in a single bound. If this guide helped you, please let me know in the comments and share this guide.

Based in Hong Kong, Douglas Black is a veteran editor of Notebookcheck, university lecturer, researcher, and writer.

201 Comments

  1. Bill

    June 7, 2018 at 10:11 pm

    I've been reading about the thermal throttling issue the XPS 15 has, and many people say it's caused by a lack of airflow in the region of the VRMs. Would it be possible to drill a small hole (1/8" to start) in the side of the fan cases to get some air moving over the VRMs?

    • Cristian

      July 23, 2018 at 10:50 am

      Only one way to find out!
      Let us know if it works, if it doesnt then you can cover that hole and go for the pads.

  2. Daniel

    June 17, 2018 at 5:08 pm

    Hey I noticed that on the cpu you put a very small dollop but on the gpu you put a much longer one. Can you confirm that applying the coolant as shown in the photo is the best way? Your instructions are a little at odds with what I see. Thanks!

    • Douglas Black

      June 17, 2018 at 6:40 pm

      That's just how I do it. :) If it's square I do a dot. If it's rectangular I do a longer line

    • k morgan

      July 8, 2018 at 10:52 pm

      You got the ICs switched. The CPU is the larger, rectangular die with a long strip of thermal paste applied. The GPU is the smaller square die with visible components on the substrate with a small dot of thermal compound.

  3. Spencer Clayton

    June 21, 2018 at 6:19 am

    WARNING!!! READ THIS!!! You actually mismarked which screws to release when taking off the heatpipes and spreader.
    You marked the screws holding the fan assemblies instead of the remaining two holding in the heatpipes.
    I sorta blindly followed the guide and nearly bent my heatpipes in half.

    • Alex R

      June 26, 2018 at 12:25 am

      Yikes! Thank you for the tip! I was thinking about doing this for my 9570, so thank you for letting us know…

  4. Ahsan

    June 24, 2018 at 7:17 am

    Hi, for the second type of throttling (VRM) how did you record those temps for each mosfet? I'm using HWinfo and none of the ambient sensors exceed even 60C when my XPS 15 9560 throttles the CPU to 0.8GHz under load (adobe lightroom export). HWinfo claims power limit throttling and the PL1 power limit is reduced from 45W to 7W right before throttling begins.

  5. Anthony

    June 30, 2018 at 8:13 am

    With the pad placement it looks like my 9570 has the VRMs in slightly different locations. Followed everything else pretty closely and temps are looking awesome. 37-40 with no fan action detected at all. Then again I don’t game anymore and totally did this to prolong battery.

    • Douglas Black

      June 30, 2018 at 8:30 am

      It seems the 9570 doesn't need as much help to prevent massive throttling like in the 9560. I just put a little padding over the top cluster and never ran into issues after that.

  6. George

    July 5, 2018 at 7:16 pm

    Hi Douglas,
    I've followed you guide very precisely, and to my surprise I think overall it has made my 9560's endurance worse. My ambient/neutral temp was about 37 and it's more or less the same since the modifications. However,when under strain the throttling now seems to kick in quicker than it did before. It's a nightmare when playing games that i used to be able to play for a few hours whithout many issues, but now experience PL1, PL2, and EDP and BD PROCHOT much quicker. I tried repasting 5 times, as initally I thought it could have been that, but that doesn't seem to be the case. Also I experimented with taking in and out the thermal pads, and adding extra ones too, but this doesn't seem to make a difference.
    Any idea what could have happened?? It's driving me crazy!
    (Also don't know if it's worth mentioning but the original paste job was very excessive – paste all over the overflow margin. Starting to think that was necessary now, and i should recreate a similar pastesplosion.)

    • Douglas Black

      July 5, 2018 at 7:44 pm

      That is quite strange. You'll need to use hwinfo to graph your ambient sensor temps under load and see what is causing the throttling. Take all the padding off and put the pads on one by one until you see what is causing your ambient (or dimm2? I can't rememeber what it's labeled in the 9560 now) to hit the throttling point

      • George

        July 6, 2018 at 2:04 am

        Thanks for the speedy reply man! Great that you're still actively monitoring this page months later.
        Okay, i'll have a go at that. Cheers for the advice. Oh one thing i wasn't clear on from the article was if the fix should in theory prevent throttling happening pretty much altogether..? How much is likely to happen even if it's a successful operation?

      • George

        July 6, 2018 at 3:34 am

        So i've done some further monitoring and the point at which the throttling kicks in was when the CPU and GPU was at 74 (should they even be getting that high after the repasting?), and there was one of the ambient sensors reading off a whopping 91. Is it that '91' chip which is causing it to kick in, or just the cpu and gpu do you think?

        • Douglas Black

          July 6, 2018 at 6:40 am

          It will still happen depending on your ambient room temp and your undervolt.

          The cpu and gpu temps aren't what causes the pl throttling, it's the dimm2 sensor (but ambient at 91 is waaaay too hot!). Are you using Dell power manager to set the fan profile? Make sure it's ultra performance.

          • George

            July 6, 2018 at 5:52 pm

            Last one mate, sorry for bombarding you.
            Sorry, should have been more clear. That 91 reading was under full load. I did some experimenting and found that chip and applied a thermal pad to it, which brought it down to about 75. (it's the one in the picture in the first post on this forum, if you're interested: http://forum.notebookreview.com/threads/xps-15-9550-temperature-observations-undervolt-repaste.785963/page-27). The throttling still seems to kick in when the CPU and GPU are at about 74 each though… I got Dell Power Manager, which actually increased the fan speed and is helping but sadly it's not quite enough to keep the temps down.
            It is very hot here at the moment, about 28 degrees in the room, perhaps it's just that? Also, the more I undervolt the lower the chip temp should be right? So I should undervolt as much as possible, while remaining stable?

          • Douglas Black

            July 6, 2018 at 6:03 pm

            The high ambient temps have a major effect. I don't think the cpu and gpu temps will cause the throttling. The fact that they both happen to be at the same temp suggests that the heatsink and case are being saturated by heat.

            And yes, undervolt as much as you can without bsod :)

  7. Harry

    July 18, 2018 at 5:19 pm

    I’m really not sure if vrm cooling is entirely nessescary. Yes they get hot but their throttle temp is up around 125C. Having the bottom open when you measured your temps certainly helped and I’d expect them to be a bit hotter. I’m more concerned with the vrm’s heating up the bottom plate and reducing the cooling for the ssd.

  8. Trevor

    July 19, 2018 at 12:23 am

    I raised my XPS 9560 up slightly from the rear to allow more air to be drawn in and it worked, the ambient temperature dropped and games were playable (but you don't want it too high its worse then).

    Will follow your tips for the CPU paste that one always works in my experience.

    • Chris

      July 22, 2018 at 2:36 am

      What do you mean by too high. I have the entire laptop raised a little over an inch off my desk, idk if that is too high.

      • Chris

        July 22, 2018 at 2:38 am

        Also i cant find the pads that specify 6W/mK on amazon so should i just get what ever they have?

  9. A. Nony Mous

    July 20, 2018 at 8:00 am

    Can you please write a similar guide how to fix the thermal throttling in the new MacBook Pros with i9 chips?

    • Douglas Black

      July 20, 2018 at 8:32 am

      Physically impossible. :)

      Mbp has a single heat pipe, almost no space for cooling, and undersized fans. No 6-core chip is ever going to keep turbo in that design.

    • Staur

      July 25, 2018 at 2:42 am

      Software fix already released by Apple.

  10. Max

    July 23, 2018 at 1:21 am

    Hey guys, can someone explain to me what Power Limit Throttling is, or even better how to solve it. My XPS is fine. But lately after BIOs update, I quite often see that Power Limit Throttling occurs in XTU. I've undervolted, though I have not repasted it yet. Thinking on doing that, though not sure it will help.

    Also, does it worth it disabling max turbo boost, and also reducing normal boost? Should it help with Power Limit Throttling.

    It is so annoying. Temperatures are fine, those never reach above 80. Makes me mad, as it seems that new BIOs update broke my laptop.

    Thank you for help in advance.

  11. Tony Bellomo

    July 23, 2018 at 7:49 pm

    I just did everything you said, and my improvement has been nothing short of dramatic! Thank you so much for publishing this. My 9560 is running much cooler and not throttling!

  12. Erica

    July 25, 2018 at 1:21 am

    I have a 9570 with the i9 upgrade if you would like me to test anything on mine and report back. We run repair shops for a living, so not afraid of opening it up. Let me know—I’ll leave my contact info here for you.

    • Douglas Black

      July 25, 2018 at 4:18 am

      Sure! Would be great if you could run cinebench r15 multi 64 10 times in a row and mark the scores. Helpful to keep graphing CPU frequency and temp as well

      • SeanSkiVT

        August 9, 2018 at 12:59 am

        Doug,

        Thanks for posting all of this. I have the i9 9570 as well, just came in yesterday. I haven't modified it at all – no repaste, no undervolt. I ran Cinebench r15 multi 64 10x in a row per your request. The temp was definitely hovering around 97c during the testing period and the fans were running up and down each time, with an indicated speed in task manager when hot of ~3.32 GHz .

        Here are the results:
        1054 cb
        999 cb
        1065 cb
        1050 cb
        1123 cb
        1070 cb
        1023 cb
        1022 cb
        1009 cb
        1033 cb

        And FWIW, the OpenGL score was 123.71FPS

        I'm not planning to undervolt, but would be interested in a repaste. I've only done it once, and it was on an old gaming PC, not a new $3k laptop :yikes: so I'm a little nervous.

        • Douglas Black

          August 10, 2018 at 1:25 pm

          You'd get better results from an undervolt, imo. I think you'd get 1150-1200 on repeated runs

          • SeanSkiVT

            August 10, 2018 at 11:09 pm

            Thanks, that's good feedback. I'm downloading throttlestop now to play with it. I suppose I'm unrealistically afraid of frying something only because I haven't done it before.

            Side note, I turned off turbo boost in the BIOS and I'm still pulling ~ 970 cb in Cinebench and the temp peaks at 74C. Not a bad drop in performance considering it caps at ~2.9 GHz now rather than 4 GHz.

          • SeanSkiVT

            August 14, 2018 at 5:49 am

            As it turns out I never got a chance to install or try Throttlestop. Dell SupportAssist threw an error that I had a memory chip failure. The laptop was only a couple of days old, so I initiated an RMA exchange. I'm nervous now, and have a bit of buyer's remorse, wondering if I should have just returned it for my money back. We'll see how the replacement is I guess.

          • Douglas Black

            August 15, 2018 at 7:11 am

            That's unfortunate. Best of luck with the HW swap!

  13. Cristian

    July 26, 2018 at 2:23 am

    Great article Douglas!
    Its just what i needed to know right after i bought the xps without knowing about these issues! Thanks a lot!
    Already got the pads and the kryonaut paste (i was using artic silver, didnt know there was something better), now im just waiting for that puppy to come home for surgery.

    • Cristian

      August 11, 2018 at 9:19 pm

      Ok, it works, The 9570 FHD has a different VRM and choke distribution but you get the idea.

      Out of the box it will start to thermal and power throttle about 15 seconds running the stress test (XTU).
      The paste helps a bit (if you do it incorrectly you will see the temps worsened, keep that in mind), also I applied 6Kw thermal pads not only to the VRMs but also to the square chokes (one layer 1.5m pad)
      The processor is a 8750H so i went for 0.150V (i heard people even go as far as 0.200V) and saw a great improvement there (I didn't touch the gpu voltage)

      Now idle is always at 38°C to 41°C, normal use won't go higher than 59°C (maybe 70°C if you start to multitask stuff).

      The stress test starts at 60°C and keeps going up slowly from there usually staying between 80° to 90° during the last 3 minutes of the test. I been using it for a day and its completely stable.

      • G

        August 15, 2018 at 12:43 am

        Likewise. Did the same as you with my 9570. Stopped throttling while I was playing games. OOB it was dropping my FPS from 60-40 but after repasting & putting the pads on the chokes and MOSFETs, was able to maintain 60 without throttling.

  14. Nathaniel Stacey

    August 10, 2018 at 12:10 pm

    hey so I followed these instructions and after booting up I was met with the 2 orange 1 white light indicator! fantastic! I always wanted CPU failure…

    • Douglas Black

      August 10, 2018 at 1:24 pm

      Did you unplug the battery while working? I can't imagine what could have happened to cause that unless the battery was left plugged and something shorted

      • nathaniel stacey

        August 15, 2018 at 6:25 am

        Omg you know what, i didnt unplug the battery. But when i turned the computer on at first it worked well for like an hour! its once i opened pubg and stressed it that it went black. after i got the signal i open it back up to look maybe i pasted it poorly but it was a great looking paste! so im pretty confused.

        • Douglas Black

          August 15, 2018 at 7:15 am

          What paste did you use? Try unplugging the battery and holding the power button down for 30 seconds, then plug the battery back in and see if it starts. Try re-seating the RAM as well. If that doesn't work then you'll need to get service for a new mobo, unfortunately

          • nathaniel stacey

            August 15, 2018 at 8:33 am

            Ive done the no battery 30 seconds reset a couple of times, I even re-pasted. What exactly do I do to "re-seat the RAM"?
            So, I bought this laptop off a guy on Kijiji like 9 months ago and when I contacted Dell for parts/repair they said that this laptop was suppose to be sent back as an exchange by the first owner. So basically they sent him a new XPS with the trust he would send this one back but he instead sold it and kept the new one. Dell is telling me that they will not do any service work or sell parts to me because this service tag is black listed. Super fun times lol.

          • Douglas Black

            August 15, 2018 at 9:42 am

            That sucks. Re-seating the RAM means take the RAM out and put it back in.

            Your last bet is to buy a 9550 mobo off ebay and upgrade. (Might even be able to do 9560, but you'd need the 97 whr battery to match)

          • Nathaniel Stacey

            August 15, 2018 at 9:25 am

            i used the Noctua NT-H1 termal paste for the jpb

          • Nathaniel Stacey

            August 16, 2018 at 9:27 am

            its a 2017 xps 15 9560, you think i could get all those parts? btw, I appreciate your help on this, thanks.

  15. Serg

    August 12, 2018 at 2:43 am

    Hi, I have the i7 9570. I noticed that the MOSFETs look like they're in a different position from where the 9560's are. Can you upload a picture of where to place the pads on the 9570? Also, for the 9570 would it still be 3 pads stacked on each other?

    • Douglas Black

      August 15, 2018 at 7:08 am

      I actually found that mine doesn't throttle at all with just a little padding in one area around a cluster of 4 MOSFETs. I will try to take a picture in a day or two.

  16. inference

    August 12, 2018 at 10:08 pm

    In which order should the heatsinks screws be tightened? The are labeled 1,2,3,4,5, so should I first tighten 1, then 2, an so on? What is this numbering for?

    You say you tightened iterativley and diagonally.. how much do you tighten in each step?

    The reason for this is that I have re-pasted but the temps did not change, or even wen slightly higher on idle… I did not tightened in order, so maybe the paste was not uniformly squeezed. Is this possible? Also, the heatsink is not bent, I checked

    • Douglas Black

      August 15, 2018 at 7:10 am

      It didn't actually matter which one you do first, just do opposites. I tighten them approximately halfway each time

  17. Matthew

    August 17, 2018 at 4:12 am

    Hey! Could I ask the average temperature difference your laptop decreased by after your cooling solution? And how would you compare it to a solution like the OPOLAR laptop fan?

    Thanks in advance!

    • Douglas Black

      August 17, 2018 at 4:43 am

      I've never found those fans useful, so I can't say.

      Undervolting seems to reduce temps at max load by 7-10C, while repasting seems to reduce temps by between 4-10C depending on your original paste job and paste used.

Leave a Reply

Your email address will not be published. Required fields are marked *