Dell XPS 15 9550 Power Limit Throttling

Squiggly,

Sorry to hear about the throttling problem you are having. A few quick comments, which you have probably already addressed:

1. The 9550 thermal paste on the CPU and GPU is terrible; people are getting significantly lower temps with a repaste

2. One or two of the VRAM chips is not properly connected to the heatsink pipe; replacing these two thermal pads helps

3. I saw you posted on the ThrottleStop section; make sure you delete XTU as it conflicts with ThrottleStop.

4. I did some novice learning about the 9550 VRM in another thread...you can see that and the response of a laptop engineer below.

5. A few guys got a big cooling benefit from holding the laptop off a desk or using a cooling fan, so that might help you a bit...

Keep us posted with your solutions...

 

Thanks for the suggestions, pressing. I've undervolted my CPU (-160mv core/cache on the 6700HQ), which helps somewhat. Unfortunately the dGPU voltage appears to be BIOS-locked, but I did underclock it by about 150mhz. It's a real shame I can't undervolt the dGPU or underclock it further, since the GPU draws probably 65W+ while gaming and therefore is the main strain on the VRM, as opposed to the CPU which draws maybe 25W while gaming. And again, when throttling occurs due to VRM heat, it is only the CPU and its relatively insignificant VRM strain contribution that is throttled.

I tried a cooling pad, which reduced CPU/GPU temps by a few degrees, but I'm not sure if it really did anything for the VRM.

I'd rather not repaste/order and install thermal pads but perhaps I will have to.

In response to Gonzo, the VRM reaches 100 C when I game, even after a hefty CPU underclock. So this isn't just a stress test problem. This appears to be the case for others, since there have been plenty complaints of throttling during gaming.

---------------

Also, I'm considering gunning for a motherboard replacement. Lots of people are having VRM-induced throttling, but perhaps my VRM gets unusually hot. Or maybe Dell has improved upon the issue a bit in newer revision of the motherboard (I know that there exist newer revisions).

Could fellow XPS 9550 owners please do the following test to help me (and others) judge whether they have unusually bad VRM heat problems that might be solved with a motherboard replacement:

Run the Furmark burn-in test with the following settings:

- stock CPU/GPU voltages/clocks
- 1280x720
- AA off
- laptop on flat desk, display open to 90 degrees

Report the ambient temperature, which CPU you have (6700hq vs 6300), and how long it takes for the VRM to reach 100 C (you can use HWInfo or Aida64 to check). If the VRM never reaches 100 C, report the equilibrium temperature of the VRM.

----

Also, has anyone had success undervolting the dGPU? How?

 

How can i verify the VRM temps? Im having A LOT of throttling while gaming, with undervolt and removing "turbo" from the CPU.
The CPU temps are not even that high, but throttlestop and Intel Xtreme Tunning (in different tests) showed TDP Throttle and Power Throttle respectively.

As many others, i understand a bit of throttling, but going to 800Mhz is just way to much, at that point is better to throttle the GPU tbh.

Oh and any guides/link on the VRAM repad? I find a lot of posts of people talking about it, but no instructions

 

http://forum.notebookreview.com/thr...rature-observations-undervolt-repaste.785963/

Aida64 shows VRM temps after I managed to find the sensor and confirm it (it showed as DIMM temp before)

 

Wee,

Quote #1 below - You can see my positive results VRAM repadding and repasting; undervolting made a bigger difference but at the extremes, fixing the defective paste & pads is important in my view.

Quotes #2 & #3 below - I did some novice VRM "research" and GoNz0 was kind enough to provide some guidance; he is a laptop engineer so his post provides a lot of color...

 

Hi all,

I was having the same issues with my XPS 15 9550 (4K i7) throttling down to 800Mhz/8x Multiplier under more or less any load. I tried undervolting and various driver fixes but nothing seemed to alleviate the problem. So, today Dell sent round an engineer who replaced the Motherboard and CPU assembly and... no throttling now! I didn't get a chance to see if the newer motherboard was a later version than the one that came out, but I'm fairly confident it's being caused by a bad batch of motherboards or CPUs so if you've got the same issues I'd suggest badgering Dell to get the part replaced.

Before, running Prime95 was a sure-fire way of causing the CPU frequency to tumble - I'd be lucky if it got to 20 seconds before doing it and the CPU package was reporting 100 degrees C at that point (my VRMs were only 65-75 incidentally) but now I've been able to run Prime95's stress test for 15mins and it's steady at about 80 degrees and, most importantly, no tumbling CPU frequency!

 

This is very interesting, and gives me hope. Was the 100 C on the old motherboard with or without undervolting? What about the 80 C on the new motherboard? Which of the prime95 stress tests were you using? 8 threads?

Could you run a benchmark that stresses the CPU and GPU and report CPU/VRM temps?

Thanks.

 

Hi,
Sorry for the slow reply - the 100C on the old motherboard happened at both stock voltages and undervolted (but just took a few seconds longer to appear undervolted).
I've been using the default Blend Prime95 torture test to stress the CPU and have had a chance to do some more tests to see what temperatures I'm getting: After after 10 minutes running the tests I'm at:
CPU Package: 85C
VRM: 77C
But most importantly my CPU frequency stays on a constant 31X for all 4 cores throughout with no signs of throttling.
The other thing I've really noticed is that the machine is a lot quieter than before - I think it was probably kicking the fan in more or less all the time at 4000rpm (and up to 5000rpm under any load) whereas now it's mostly whisper quiet unless I'm stressing it hard.
Feel free to let me know if there's a particular benchmark you're interested in but I'd say if you think you've got the same issues I had before this motherboard was swapped definitely get onto Dell and ask them to get it replaced!

 

I'm probably a bit late, but I can confirm the strange throttling issue during gaming. I can't explicitely state that it is due to the VRM temperature, but all the other sensor readings (hwinfo64) seem fine. Worst of all, I can't seem to find a pattern as sometimes i can do an extensive gaming session (2 hours in a hot space) without issues, and sometimes will just start the notebook up and get bad performance 5 minutes in.

TurboBoost disabled, GPU temperature in normal range + stock clock, CPU temperature in normal range BUT when performance drops i notice that the CPU clock goes down to ~800MHz. Interestingly, hwinfo does not detect any thermal throttling as it was the case while I had turbo boost enabled in bios.
For the moment, switching the power scheme to maximum performance seems to get rid of the CPU throttle, and my temperatures are sort of alright.
Normally I use the Dell custom scheme, and funnily enough, I have not seen any specific difference in config for AC mode.

Is it safe to say VRM is the culprit here?


EDIT: The power scheme change did NOT solve the issue in the end, don't waste your time.
I have just experienced a cpu downclock to 800MHz after only 5 minutes of play, even though I've managed to play for hours without issues earlier today.
I'll try to get some VRM readings from hwinfo (or aida64) and get back here with more numbers.

EDIT2: After 2 hours of gaming today, have not managed to reproduce the issue while Aida64 reported a maximum temperature of 94C for VRM
Is this actually working as intended? Is 100C for VRM something we really want to avoid, since some people are stating that the fallback is only there to keep the laptop more pleasant to hold.
(They uninstalled the Intel Dynamic Platform and Thermal Framework)

 

Hi everybody, i'm a newbie proud owner of the i7 9550. When trying to optimize battery with undervolting and low cstates, running hardware monitor make me found something strange: my c state core package goes down to C8 state... Did i miss something or this is a real good news? I'm waiting for the new version of throttlestop for confirmation... Feel free to confirm nor deny!

 

Thanks for the reply. Is the 85C/77C when undervolted? By how much?

Could you run a benchmark that stresses the GPU, or both the CPU and GPU, and report temps? Could you run Furmark? And then run Furmark and Prime95 (blend, 2 threads) simultaneously? (I think that Furmark + Prime95 8 threads is unrealistic and unfair.)

I get CPU 69C/VRM 78C when running prime95 blend 8 threads (with the CPU undervolted by 150mv), but if I run Furmark, or a game that stresses both the CPU and GPU (like Overwatch), I get way higher temperatues and throttling once the VRM hits 100 C. What's your package power when running prime95 8 threads blend?

 

Several posters here have warned against using FurMark. Extreme stress combined with the weak cooling system of a laptop, particularly one so thin as the 9550, risks permanent damage.

 

Yep furmark has to be the worst thing you can do and expecting people to reduce the life of the GPU doesn't go down well to help other members

 

Can you verify what BIOS version you are running?

 

I can confirm that in my case, the CPU underclock does NOT have anything to do with the VRM temperature, CPU or GPU temperatures.
I've just had a day playing under some very comfortable cool temperatures, and still dipped into the 800MHz range while playing overwatch.
Maximum performance power scheme was set but did not help at all.
It seems completely random and I cant work out the cause.. I will try completely disabling SpeedStep in BIOS and see what happens,

 

Does throttlestop + high performance mode allow you to take control of the clocks?

 

Throttlestop didn't really help in my case.
However following those steps helped a lot :

Source : https://www.reddit.com/r/Dell/comments/4tus0u/another_cpu_throttling_post_xps_9550/

I also downgraded bios to 01.01.19
and uninstalled Intel Dynamic Platform & Thermal Framework Driver.
Thanks to this it is not getting as hot as it used to be and I get better FPS.
Please give feedback if it works for you as well.

 

I did not try throttlestop, and I will avoid using it if i can.
It is simply against my principles to pay this amount of money for a notebook, only to end up relying on 3rd party apps combined with obscure temperature optimizations for a case where downclock happens programmatically without any visible pattern at all.

EDIT: Disabling SpeedStep did not help. Will check C states next.

If this is a tdp induced throttle, i suppose it is there because of some system temperature, but temperature of what?
Who do we even contact in regards to this, Dell, Intel?

Interestingly, unplugging the battery charger an interesting thing happens.
My game fps gets capped at 30 (probably some app setting), but the cpu clock goes up to about 1200-1300. It is still clunky but not so unevenly unplayable as in default mode (which shows over 50fps ingame but is actually unplayable with crackling sound)

Can this have something to do with the battery or charger itself?

 

Some interesting things. Look at the PL1 Power go really low and at the same time the CPU throttles really heavily.
Another interesting thing i could almost certainly make the CPU triggle the throttle by just scrolling the HWiNFO window up and down...
(check the last spike down)
The rest is charts while running Prime95+Heaven



PS: VRM is what i believe is the VRM, i have 4 temps that show as Ambient.

Cheers

 

VRM is correct in that as I tested it with a cotton bud soaked in isopropanol when chasing it down in Aida64 (Aida64 said the VRM was the RAM until the dev fixed it)

 

Ok, i just got a new Motherboard replacement, really odd it contains BIOS 1.2.12 (not released yet?). Still does the 800Mhz throttling, although i have to perform some more extensive testing.
I do have lower VRM temps, and if you check my old temps, i dont know if the throttling is due to the VRM. I believe they have been tweaking something on the BIOS that causes the CPU to throttle really hard.
Does anyone have a BIOS < 1.2.0 so i can test? and anyone knows how to backup my current BIOS?

thanks

 

Maybe you can find your current and historic drivers here:

http://downloads.dell.com/published/pages/xps-15-9550-laptop.html

I have seen instructions for backing up current Dell BIOS so they are floating around somewhere

Could you tell me if adjusting screen brightness by holding down f11/f12 while audio is playing causes audio crackling noise?

 

New BIOS eh, at least ones in the pipeline if they are on the replacement boards.

 

It is still there if i do while playing music.

 

Thanks WeeDv2 - sorry to hear that

 

Just posted my analysis of this issue over on another thread:
http://forum.notebookreview.com/thr...ndervolt-repaste.785963/page-47#post-10333286

Worth a read for anyone experiencing issues with throttling.

 

Just tested with BIOS 1.1.19 and still has really heavy throttling. It lasted a bit longer on, i think this could be easily fixed by just throttling a bit to reduce heat/power draw, without going all the way to 800mhz, but go get DELL to do that.
I have a Dell L3 Engineer (allegedly) say the device is working according to design (this is, backlight bleeding, bad wireless card, the 800mhz throttling, and the screen key up key down noise) which is ridiculous, so of course im disputing it, this is because they cant fix the problem (well of course, if the only thing they do is replace with refurb parts) and just want to get off the hook.
Its a shame there is not other device with such a nice screen, otherwise ill buy something else.

 

Get a refund and buy again?

Sent from my SM-G920F

 

Another note about 1.1.19, i have higher screen brightness, and i believe better battery life.

@GoNz0 i would love to, but i cant since i purchased on a MS Store in the US, and i live in Germany. I purchased an additional NBD Premium support warranty here, since i was scammed at the MSStore into believing that the MS warranty would cover me over here (hopefully i will be in the us in 4 month and ill complain at the store as there is not other way of communicating with the store and all phone/chat/mail services are for Online Purchases only, ridiculous)
The bad thing is, Dell says "well you purchased at a 3rd party so we dont replace the device", another scam, as this was not said to me when i purchased the NBD warranty.

 

Some information update incase it could be useful.

I had recently requested a change to the palm rest piece (the soft carbon material) from Dell due to small bubbles increasingly forming on the left side of where my palm rests while typing. I forgot to tell Dell about the heat issues I've been having since I have remedied it myself just with TS. Interestingly, Dell themselves asked if I had any heat problems and if I wanted to replace any of the materials that help reduce heat, so they also replaced my heat sink and I think the fan on the left side in the process (replacing the carbon fiber material means taking the whole laptop apart). I also had the bottom aluminum plate replaced because after I replaced the carbon fiber material, the old bottom aluminum plate did not align with each other very well. After the replacement, I tried testing out my thermals without UV with TS and I was down by 3-5C from normal when playing demanding games. With UV on TS, I've decreased by ~10C compared to before the replacement parts. My idle temperature was more or less the same, but it feels much more stable than before.

An interesting side note is that the replacements for the carbon fiber material and bottom aluminum plate feels different. The carbon fiber before the replacement was a lot softer to the touch, but also more prone to fingerprints and oils. The replacement felt less soft and as if it was coated thinner or something. It still feels nice, just not as cushion-like when I rest my hands on it. The bottom aluminum plate also felt more plastic and slightly lighter. During the replacement I tried tapping on the old and new plates, and they sounded different, not much, but enough to be noticed. It also has a slightly smoother texture to it than before.

 

Today, I did some stress testing on my 9550 (6300HQ) for fun and it shocked me as the laptop did not spin into PL throttling with ROG RealBench StressTest. That always killed my laptop even with max thermals below 75*C. I am trying to figure out what changed.

1. Best guess - Last month I was messing around with the newest version of Intel XTU software which lets you bump up:

- turbo boost power max
- turbo boost short power max
- turbo boost power time window

==> Bumped "turbo boost power max" to 65 watts (from 55 watts). That 65 watts shows up in HWiNFO64 also.
==> Also increased the "turbo boost power time window" to 96 seconds.
==> These settings seem persistent as they stick in registers even after several weeks of shutdowns.

Notes

- several versions of Intel XTU have CAUSED power limit throttling so after setting above factors, I don't have that buggy program running on my computer. Set it and never open it again
- XTU also will conflict with ThrottleStop so those two programs can not be run concurrently
- I tried adjusting those items in ThrottleStop (turbo power limit window) but don't think the settings did anything on my system

** DON'T EVEN CONSIDER INCREASING TURBO SETTINGS IF YOU HAVE NOT UPGRADED CPU & GPU THERMAL PASTE AND CHECKED FOR CORRECT CONTACT OF VRAM THERMAL PADS

** TRY AT YOUR OWN RISK. SEE DISCLAIMERS IN XTU.


2. Other changes in laptop that may have improved performance

- enabled "Ultra Performance" (fans-processor boosts) via Dell Command-Power Manager
- enabled SpeedShift EPP=0 (via ThrottleStop)
- disabled c1e states (via ThrottleStop)
- updated BIOS, drivers, Windows 10

fyi I always run ThrottleStop with undervolt to CPU and GPU