Something is Wrong with my XPS 15 When I try to undervolt

I am sorry that I am posting a lot of threads, but I am really concerned about my new baby here. Ok, I am really skeptical about my undervolting method. I recently got my Dell XPS 15 9570, and was shocked by the power limited throttling, that when I play games, the CPU goes down to 0.8 GHz. So I thought I gotta do some undervolting by xtu and throttlestop. Now here comes my skepticism.

I was able to make Core and cache voltage offset to -200mv and Intel graphics card voltage offset to -150mv, which sounds crazy for me unless I am really lucky. I ran prime 95, and it does not show any blue screen, and the temperature is around 75 ~ 80c, so it looks good so far. However, whenever I enter XTU, for some reason, it is recommending me to reset my voltage to normal which is a bit odd in my opinion.

Also, when I watch a youtube video, sometimes the sound crashes and I am not so sure this is due to undervolting or not. What do you guys think? Is getting -200mv reasonable or just something wrong with my unit?

 

ok finally my -200v showed some blue screen, now I might be able to find my sweet spot.

 

You can not undervolt to -200mv and expect no crashes.

Review the posts here and you will see some people claiming -150mv. That varies by system. You should start with a less aggressive undervolt and stress test for a few hours. If that is good then maybe another -10mv.

I would hope your CPU can be stable in the -120 to -140 range but not much more under stress.

 

That's what I was thinking as well. Since I got my first bluescreen on -200mv, now I am slowly dropping my voltage to find some sweet spot.

 

Yeah I don't think I've seen anyone run -200mv stable. Start at -150mv and go from there. I'm able to run mine at -190mv stable, but I consider myself one of the lucky ones

 

Don't use XTU, it's broken. Use throttlestop.

 

Fully agree. Throttlestop is a great little utility and does a good job as long and you test your settings 1. actually are working, and 2. leave the system stable/reliable.

 

Did you repaste that system? Throttling to 800mhz is extremely severe.

 

I was able to have stable -190mv but after I restarted my laptop, for some reason, throttlestop changed my setting to default. So it seems like 190mv is no good for me

 

Not yet. I will get my thermal pad today!

 

Good decision.

 

It seems like my unit doesn't have thermal throttling, but hard power limit throttling because even when I ran prime95, there was no sign of thermal throttling according to XTU.

 

Set it to "save settings now" in FIVR

 

I'm currently testing the i9 8950hk in my Xps and have found that it does not like -150mv core and will cause my system to hang.

Currently trying -145mv core amd -142mv cache with 0 Speed Shift EPP

 

I think you need to match undervolt of CPU core & CPU cache.

-150mv is a pretty aggressive undervolt. Also, each chip is different so you can search for undervolt ranges that worked for others. But you can't trust that the higher numbers were tested in more extreme conditions so may not be very stable.

Normally you will see people test with very conservative undervolts, say -100mv or -110mv and run some tough benchmarks for a few hours; maybe some games. If no issues maybe bump up in -10mv increments and full testing again. Then once issues arise drop back -10mv or -20mv.

 

Yes I have a feeling my values of -145 and -142 is too much. When running Realbench stress test my Dimm temp goes through the roof 90c and starts underclocking my cpu significantly. System then blue screens.

 

This guide from @Eason is very helpful with undervolts

https://www.notebookcheck.net/How-t...ife-The-ThrottleStop-Guide-2017.213140.0.html

 

I am still testing my undervolt, and it's currently -160mv. What's funny is when I was using -190mv, it passed the prime95 test, but when I played outlast 2, it immediately showed a blue screen.

 

You need to do a variety of stress tests and your own usage to see how it holds up. Your stress tests should be for several hours also.

Your gaming pushes the power module hard due to GPU use. Also has different CPU demands. Bottom line is that you should not expect to run undervolt near -190mv if you are looking for reliability.

 

Typically a CPU will work fine at a decent voltage when at it's full turbo boost due to the way the voltage tables are setup. -160mv isn't much when you're at a much higher voltage. The problem is when the system lowers the clockspeed the -160mv isn't stable because it's running at a much lower voltage due to the voltage table.

An easy way to test this is to set your undervolt, start prime95 for a couple of minutes, and then exit prime95. Do that a couple of times and you'll probably BSOD (for sure at -200mv). That's why I recommend AIDA64 because it varies the CPU speed (and the voltages that it runs at).

Hopefully that makes sense.

Here is a quick example with just random numbers that I made up.
Say at max turbo the cpu voltage table says use 1v so 1-.016=0.84v
But at min clock speed the voltage table says use .4v so 0.4-0.16=0.24v

 

Good explanation! Here is my scenario.
i9 32gb ram with stock voltage.
Aida64 Stress Test.
CPU 80c ish average upper 70s
Dimm upper 60s low 70s
CPU average clock 2.9ghz


I then run Unigene Heaven by itself and use Aida64 to monitor and my Dimm temp levels out at 93c and GTX 1050ti at 76c.

I cant run Realbench stress as it causesy DIMM temp to shutdown the system i think 100c?

I did a repaste and cpu and gpu temps look fine but the DIMM see to just scream north with a heavy gpu/cpu load.


Sent from my iPhone using Tapatalk

 

DIMM temp in AIDA64 is mislabeled. The DIMM temp is actually the first ambient temperature near the majority of the VRM.

You could try the Iunlock mod that I did and it will help drop that temp, it is a lot more involved though.

 

Is it labeled correctly in HWinfo64? or what is it labeled as in HWinfo64?

 

I think it is labeled as Ambient 1, but don't recall offhand. I thought I had a picture somewhere of it.

 

Hmm.. Not sure. As the "Dell EC ell XPS 9570" sensor area is grayed out.

 

Not sure why it would be greyed out, but I know for a fact that AIDA64 is mislabeled.

 

I was able to add that section to be monitored. Here is where I sit now.

 

Ok, I Did repaste the thermal compound, and it does not help to change the extreme throttling, and I assume this is due to power throttling. Maybe it might change it after I change apply thermal pad.

 

Ok, I did everything to make my XPS 15 stable - Clean windows 10 install, update, replace thermal compound, thermal pad, and undervolting -, and in conclusion, XPS 15 9570 CAN NOT have stable 142fps in overwatch in lowest setting at FHD. Still, the laptop is very very stable, and maybe if I get an external gpu docking station, it might be better due to no stress to the gpu.

 

At least it is stable. What sort of FPS are you getting out of it?

 

For first 5 min, I get stable 150 fps, but after that, it goes down to 80 ~ 100 fps, then it goes up to 144fps. Temperature is somewhat nice and cool, and I assume it is something to do with power limit throttling.

 

You can do logging with hwinfo but I bet you're right. With an egpu setup you would help take some load off the system but the CPU will still be limited to 42W. Thankfully the GPU wouldn't be limited though.

 

It's really unfortunate that despite the brand name -XPS- the performance is limited due to the design of the laptop.

 

Between the DIMM throttling, low power limits, non-existent VRM cooling, and terrible heatsink design I agree with you. I was still pretty happy with the machine after modding it though. Dell isn't going to change anything though, they've shown that time after time.

 

I love this laptop, probably one of the best laptop that I've used in my life, but at the same time, there is a lot of room to improve: Better cooling, More IO, more m.2 slots, and such. However, I also don't think Dell's gonna fix it considering they technically downgraded XPS 13 2018.