XPS 15 9570 Owners Thread

Alright so I put the 1mm pads instead of the 1.5mm that I put there before..
Repasted again and checking tempature..
Idle is going from 46-60C ;/ Probably temp sensor isnt so accurate..
Anyways Intel XTU stress test again with -0.130 as before I repasted and this time indeed I can see more stable temp of around 78-82 which is great compared to what I had before when 3 cores got to max of 95 and temp was between 84-88..
I guess what made the issue was indeed the 1.5mm pad I put that made the heatsink not closing properly.
Thanks for everything will keep an eye on it,Hope my faults can help someone who might get to the same issue
Edo.

Edit-
Uhmm well stress went well and everything..But after connecting the charger to the PC thermal thorttling start to work when browsing the internet? HWINFO showing CPU went to max of around 93-98 on all cores! Any idea?

 

Do you use the Dell drivers for the GTX 1050 and Intel igpu or do you download directly from Intel and Nvidia?

 

Directly from,
I'm not pretty sure what happened when on idle temp went to that high when on stress it got to max of 82.. ;/ (Probably due to charging? But why?)
So far right now it's pretty much fine (On battery) Still trying to understand why that happened.

 

Well, the CPU is downthrottled on battery. You do stress tests on battery!?!? If paste is okay, a few minutes of Prime95 stress test (all threads) shouldn't get the CPU into the nineties, let alone web browsing.

 

After I did a clean install of Windows, the latest Nvidia driver refused to install. The only way to get it to work was via DDU in safe mode.

 

@edoroom I think that you (and many others) look to much at Max temps. The fan curves are, afaik, tuned to only increase gradually when needed, rather than the instant the CPU gets hot. This most likely to keep fan noise as low as possible. The problem is not that a CPU might reach those temps, it is if they stay there. Thermal paste is not a magic fix that changes the laws of physics

Sent from my SM-N960F using Tapatalk

 

Thats the first step you did wrong for a clean Windows 10 installation. One of the first important things to do, before doing a Windows update is, to deactivate driver updates through Windows update, for example via OOSU. Reboot Windows, now install Wifi/lan drivers. Windows 10 wont install OEM GPU drivers now anymore automatically through Windows update.

 

Actually, I had no problems at all install generic Intel GPU drivers after the Windows 10 installation a few days ago. Must admit I was a bit surprised myself not having to run DSU etc.

 

Uhmm well..After all I thought it's enough to be on battery and keep it on the best power settings..
Anyways with retest on stress test on power is connected I'm getting totally diferent result of around avg of 92C and max of 97,Where all cores got to thermal throttling..

Of course hehe,I never was thinking about it as a magic,Just wanted to get the best I can get out of that machine in terms of everything,So the Max Temp is just a good indicator when I compare it to other users from here if what I did was okay or not,But when I see on stress test with undervolt a temps of around 92C I guess I did something wrong? :/
Or maybe its normal and I'm just stupid to believe other owners ;D

 

Laptops typically perform best when they can access additional power from both the wall & battery simultaneously.

These are good points. But the fan curve impacts thermal testing across the board. On the other hand, thermal testing at idle is not viewed as particularly helpful (but if your laptop is frequently on idle and you are trying to reduce fan noise or battery consumption SOME of your testing should be at idle).

​

Not magic. But with the garbage thermal paste jobs coming on the XPS and many other laptops these days, repaste can boost performance significantly. And the 9570 is very happy to undervolt also so that can help boost performance more.

Albeit, the 9570s old cooling system is still under-engineered, but a repaste and undervolt help a lot.

 

@pressing probably mean the 9570

Thing is that if you read these forums, you can get the feeling that "if I just repaste, this will be the ultimate performane laptop". It won't, especially with Power Limit Throttling taken into account. But it will help in lowering temps, mostly under stress.

 

I'm pretty sure no one really except to get a ultimate performance laptop with just repasting.. ;/ It's more that from how it sounds and how everyone talk about it's improvments everyone want's to do it,Just in getting the 4-6 Lowered temps for hardware longetivity and of course reduce fan noise if possible..
For example the first thing I did when I got last week the 9570 was to come here and make sure I follow any tips possible to get the what you call "Ultimate Performance laptop" It's not that I expected it to do something that it cant do normally..Just to make sure it would be optimised the best it can,For example by making sure to buy the intel 9260 with it and replace it right on the first time it saved me probably wifi problems that some owners of the XPS complain about..Maybe it was just a waste of 30$ but from my point of view if there is already 121 pages here on that forum about the XPS so probably people with much more experience specially with the laptop can provide their suggestions to people for example like me who have no clue what that laptop can do ;D

 

Well, if I don't remember incorrectly, the impact of Power Limit Throttling was much less for previous version of the XPS15, so that the thermal throttling had much bigger impact. My point is mainly that there is always a risk that you can do something that goes REALLY WRONG if you try to repaste, especially if you are not used to these kind of things. Damage a heat pipe of something like that and no amount of thermal paste in the world can save you. Just so that average joe does not think it is an absolute must to repaste themselves and expect miracles just because us "power users" talk about it in the owners thread...

 

A lot of us got the original 9550 to work well with undervolting and repasting.

Only a handful of people were able to get the 9560 working relative well at extremes (see @iunlock extreme mod).

I don't recall anyone resolve throttling on the 9570 in more extreme usage. The old cooling system and vrm limitations are just not up to the task. But a simple undervolt and repaste gives you a significantly better performing laptop than what Dell delivers from the factory.

​

Repaste requires a bit of research and care. There certainly are risks but they can be reduced with a research and careful preparation. NBR has a massive amount of information and posts. Frankly, not many people have (admitted to) system damage based on 9550-9560-9570 posts here. I suppose the main risks are bending-breaking the very delicate heatpipe, overtightening heatsink screws (to strip or rip out interface "nuts"), shorting out the system, or inadvertently knocking off one of the tiny components/connectors on the board.

Repaste with liquid metal is rather risky and requires some special precautions and research. Also requires frequent checks to make sure not spilling onto other components.

Undervolt is easy and risk is low. ​

 

Well, the thing is that I have actually compared a repasted 9570 with one with standard paste (exact same specs of course), and while the difference in temps were there, the actual difference in performance were much lower. And this due to Power/current throttling. So while the repasted one was cooler, both throttled due to Power/Current rather than heat when running something like Prime95.

Undervolting however I only see benefits with, and the risk of actually damaging anything with it is slim, if they even exist.

Now, I don't claim this to be an absolute truth, but instead my observations.

 

Agreed. But most people are not running Prime95 or other brutal benchmarks all day.

So I think that the reductions in temps and power from undervolting & repasting can help a lot of people in day-to-day activities. And they can make the laptop quieter.

Still Dell needs to up their thermal game . . .

 

I really dont get all of this: oh I need to do a fresh install of windows.

The dell parttion doesnt come with many programs that are uselless and its a matter of just uninstalling. The dell partition also comes with the drivers and allows you to revert to factory conditions.

From what I see doing a fresh install removes all these benefits.

Sent from my SM-N950F using Tapatalk

 

What are those benefits that you loose? Maybe if you live in Internet free world I could understand it. That's not to say a reinstall is needed, that's more OCD. And yes, I do it myself always

Sent from my SM-N960F using Tapatalk

 

Dear All,

I am very new to this forum and I recently purchased my first XPS. I have the i5-8300h model with the FHD display and the 1050.

I am having a few issues and thinking of returning it but dont want to return it if I can get the issues solved.

1. When on battery , it seems like the 1050 is severely downclocked especially when running benchmarks. It doesnt seem like this is happening when the AC adapter is plugged in. I have set the slider to best performance as well as im using the ultra performance option in the dell power software but this doesnt help when on battery. I have GeForce experience installed and am thinking on uninstalling this and letting the dell updater keep the driver updated.
2. When on power and I run a CPU/GPU combined benchmark it seems like my CPU sometimes downclocks to 700Mhz for a few seconds then will bounce back up above 3Ghz. I think based on this forum it is VRM related but I am not sure as when I run a CPU benchmark without GPU, this does not happen and i get excellent results with the CPU consistently staying between 3.6 to 3.9Ghz.

Hope someone can help.

 

The fan logic is horribly designed on the XPS, and Dell has no wish to change them. Two most annoying problems:

- fans change RPM too fast and rapidly without need under some situations, for example, gaming. Instead of a steady silent RPM, they suddenly jump in one second from 50% to 100% then 3 seconds later back to 50% then jump again to 100% and so on, instead of a steady 66% spin. This is totally annoying for the ears but also mostly not good for the fans. Monitoring CPU and GPU temps at around 65°C there is totally no need for this behavior. This seems to be a bug or just bad programmed fan control.

- Bad bad bad bad low work fan curves. If you are fast browsing temps easily are balancing around 45-50°C. Then if you're too quick with opening new pages, the temps jump to 60-65°c for just 1 second, and the fans are kicking in, where there is no need for because temps are going back in an instant to around 45°C. Fans then keep on for 10-20 seconds for nothing and turning off again, until this repeats.

- happens a lot the fans just activate for 2-3 seconds and go off again with the above "low work" flow

- watching a video will mostly activate the fans sooner or later, especially on youtube, or with HW decoding. this is unacceptable. the temps are around 50°C and this is no reason to trigger the fans.

It is, like it is, and like I said. If you do a clean install you can install normal non OEM GPU drivers. If youre unlucky, the GPU driver trigger point via updates is triggered, this doesnt happen the first time you do Windows updates, it happens somewhen after first hours of use. If this happenes, you will have an issue installing non OEM GPU driver. No need for DSU nonsense just can remove driver via Windows device manager btw. If youre unlucky, Windows will overwrite newer GPU drivers with older OEM drivers again at some point, so you should totally still deactivate it with the tool or manally via registry...

 

1.Normal. Many youtube vids have shown that dell has chosen this route. This is normal.
2.also normal. When the egpu is used more heat is generated. It will compensate for it.

Sounds fine to me.

Sent from my SM-N950F using Tapatalk

 

Thanks for the reassurance, when plugged in, its a monster of a laptop and im actually glad i bought the i5 because it looks like its not throttling compared to others with the i7.

My only question would be that if the downclocked 1050 when on battery is normal, is there a way I can boost this back to its normal speed when I want to on battery ? Also, what did you think about my question regarding GeForce experience and drivers?

 

Has anyone ran LatencyMon on their 9570's to check for DPC latency issues?

 

1. Downclocking on battery is by design, for both longer battery life and lifespan.
The CPU is downclocked in the advanced power profile settings. It used to be possible to set the "on battery" values to the "plugged in" ones, not sure if still after the redesign that comes with Modern standby instead of S3 sleep?
The GPU is capped to 30FPS on battery in Geforce Experience, "Battery Boost".
But I wouldn't recommend you to change this, it might cause faster ageing of the battery, maybe swelling. Fast discharging at high temperature isn't healthy for it.

2. Get HWinfo64, open Sensors, maybe open graphs of CPU and GPU clocks and temperatures, also watch the throttling flags and power limits. So you can see what is going on exactly. Power limit throttling under load is kinda "normal" unfortunately with XPS. Try CPU and GPU undervolts, disabling CPU turbo when gaming, putting thermal pads on the VRM mosfets, maybe repasting - google up specific advice that has been posted many times.

 

I would like to bump my post as I really have no idea if it's a normal thing or not,
Always when I did the stress test it was on battery and the max temp was around 80-84C (After the repaste and everything..)
Now because I wasnt sure that my repaste was alright I did it again and this time I tested on charge..(I'm on Undervolt of -0.125mV) the temps getting from 92-98C (Like the picture I gave there ^)
When thermal throttling is getting mostly yes..When it was on battery stress test it was never like that,Just power limit thorttling kicked in every few seconds but temps were stable at 80-84..Now please someone just tell me it's a right behaivor and that's what I should have except it..I know in normal usage I wont come to things like stress test is doing but the stress test is a good indicator to know (As I can compare my results with other who posted their undervolt) If the repaste I did was fine or not..

 

It's hard to tell how your repaste is if you don't know what frequency your CPU was running at before. It's likely just fine. Also running on battery will limit speeds and lower temps, that's totally normal.

 

Yes, it is really bad. Not capable to do real-time audio stuff, without some workaround like process lasso. As typical ACPI.sys causes high latency spikes for me usually up to 2-3ms all the time. Another thing to add to the bad list and not optimized drivers. Turning off wifi doesnt help, so it's mostly energy saving. AC and high-performance setting doesnt help though.

You wont find anyone in this thread mostly, who will understand this issue btw.

https://www.reddit.com/r/Dell/comments/8rleyj/dell_xps_15_9570_i78750h_awful_dpc_latency_very/

The XPS 15 is mostly just a $2000 laptop you better not use and just sit on your desktop doing nothing. But it can do Word decently.

 

I have several posts on latency of the 9550 & 9560. They were terrible when originally released but eventually LatencyMon results got pretty good without requiring many tweaks. I don't know if it was updated drivers/BIOS/Windoze or some combination.

Can't predict LatencyMon results on a 9570 but I think the test should be done on idle with no programs running for at least 15 minutes.

Currently, I run huge virtual pianos with a RME audio interface on a 9550 (i5, 32GB RAM & Samsung SATA 850 SSD. I can run a 40GB+ piano at 88.2kHz and 96 samples with virtually no popping, perfectly playable at uber low latency. Even though this is a mid-tier laptop, I haven't seen anyone running Garritan CFX Full at lower latencies.

Currently, disabling c-states in BIOS improves LatencyMon results slightly. Also I removed all the nVidia drivers with DDU as that helped and I don't use the dGPU these days. I run SpeedShift (EPP=0) and Windows High Performance (Power Options).
​

I did not see any improvement with the 9560 for audio purposes.

Some of the guys at GearSlutz like the HP laptops for audio. HP is not the engineering powerhouse it once was, but they seem to have recovered enough from years of massive scandals to focus on producing good commodity laptops. Apple has some nifty new i7 laptops which are worth a look; MacBooks tend to play well with audio (one reason is good engineering but note OSX runs some hidden buffers so latency results are a bit overstated).

 

Glad that we have you in this thread then

I recall we had a DPC discussion that got quite heated in the 9560 thread some months ago. Can't recall that we were able to find a laptop that didn't have DPC problems or that we found a "magic fix" for the 9560 problems. I will soon have the ability to do some testing on a Thinkpad P52, will be intresting to see if a proper workstation can handle this better. Will report back when I know.

 

@pressing Interesting with your 9550. I guess that is with no WiFi as that is a known problem regarding DPC? Uninstalling nVidia drivers are interesting, I thought the were needed to get the Intel GPU to kick in or rather one would be stuck with dGPU instead.

Adding a TB16 dock to the mix I have found to introduce many DPC related problems...but also of course add a lot of practicality.

 

When the 9550 was new and had immature drivers/BIOS/os, I tried all kinds of latency tweaks and they had varying levels of success. Disabling networking & antivirus certainly helped before, but during the past year or so, I don't see a difference on LatencyMon or when I play.

Of course, if running critical low latency processes, like recording, then I would want to disable networking, antivirus, etc. just in case.

I saw those nVidia processes littering the LatencyMon pictures, particularly DPC routine execution time, so just uninstalled them with DDU. That improved the LatencyMon results and reduced audio dropouts at extreme performance.

Those TB15 & TB16 docks are garbage.

The only external device I run on the XPS is a RME BabyFace Pro USB3 interface.​

 

@pressing May I ask how come you landed on the XPS15 for such specific requirements and not something like a MacBook Pro or a Workstation laptop? Did you look into other alternatives?

 

Super unfortunate to hear about the DPC latency issues. I'd been looking to pick one up to replace my aging Razer Blade that I'd been using for sound engineering and production+live performance work.

Do we know if *every* 9570 has this issue, or if it's an issue in just some of them?

 

The 9550 & 9560s were very heavily discounted like 50% or more. And I assumed they would be "high-performance" based on prior XPS units & Alienware acquisition. I also like the screens, mouse, keyboard, form factor, (crippled) TB3 port, ability to add RAM/SSDs. Dell released the 9550 & 9560 to the public as beta models and required a ton of work to get thermals under control and running smoothly; some of the firmware and thermal engineering is still half-baked, unfortunately. I will say the 9550 and 9560 I used have good latency performance both on LatencyMon and using my huge virtual instruments (finally).

I have some older MacBook Pros but the previous generation was overpriced with disappointing expandability. I like the OSX audio integration and overall stability but there is a hidden Apple buffer that makes latency "appear" better than it really is.

Intel's marketing shows 8th gen CPUs provide a lot more audio processing power via faster speeds and more cores. But from a practical view, the higher speeds are just for a few seconds & thermals will significantly limit the number of tracks one can process. That plagues both the XPS & MBP. Still a decent bump up for many users from the virtually similar Skylake & Kaby Lake CPUs.

HP and Lenovo seem to have some fans in the audio production world also so those are worth looking at. When I was in the market, the XPS was much less expensive.

If Intel ever gets Thunderbolt integrated into the CPU (has been hinting at this for too long) and kills the sloppy TB controller board solutions, and if that performs closer to PCIe, I might consider upgrading the 9550.

All that said, a desktop with a quality motherboard, non-BGA K CPU, sufficient VRM, etc. will perform much better than any laptop can. An old PCIe audio interface will crush anything USB or Thunderbolt. This is probably what I will buy next.

Professional audio engineers & builders of audio PCs frequent the gearslutz forums so they will have more informed views on the newer generation of laptops regarding latency issues.

 

The 9570 has been around for a while so I would expect the BIOS/drivers would be somewhat sorted by now. I see a few people here with DPC latency issues however. I haven't seen many audio folks with the 9570 but check out the gearslutz forums.

In some countries you can return the XPS during the first say 30 days for a refund. That might be worth testing your work flow to see how it works for you.

 

...Yes?

https://www.dell.com/community/XPS/Win-10-Home-High-DPC-latency-XPS-15-9570-ACPI-sys/m-p/6159137

https://answers.microsoft.com/en-us...s/e7fa0ab7-bf94-4aff-a7b4-54570cc37f31?auth=1

https://www.reddit.com/r/Dell/comments/8rleyj/dell_xps_15_9570_i78750h_awful_dpc_latency_very/

https://www.dell.com/community/Lati...s-on-Latitude-7389-causing-audio/td-p/6074873

If you google about it, mostly all Dell laptops have bad latency, especially the 9560 too. My guess is, there is some bug/bad design with something related to battery/AC charging or EC poll of Dell energy saving drivers, because 90% of the latency comes through ACPI.sys. Of course is you have not swapped the Killer card, it will cause latency too, because of bad drivers.

It seems all Dell laptops of the past years have the same issue, so it is mostly a bug with their bios or EC, which is present in every laptop of them. And as usual, Dell doesnt care about problems like that.

I had tested it with Intel wifi card, clean Windows 10 install, ac power, no antivirus, no background programs, and ACPI.sys was always causing high latency spikes. Deactivating c states and powershift didnt help.

Dont buy a Dell laptop if you want to do realtime audio with it. Seen reports of user switched to Asus, Lenovo and HP and they had no problem at all and great latency.

If you have Nvidia Experience installed you have to deactivate its overlay, it will cause permanent CPU load and a 0.5W drain.

 

I hate it too. When coding on Ubuntu an AC fans go on for 2 sec and off for 1. Repeat this for whole day .... This is so annoying.
But when on battery, fans are usually quiet all the time.

On my other dell 7490 fans are working so much better.

 

There seems to be a shortage in parts. Many orders have been delayed.

My 2TB/i9 is going to take a long time.

Sent from my SM-N950F using Tapatalk

 

@maffle Curious, it's quite obvious that you dont like Dell or the XPS15 but what laptops and brands would you actually recommend instead? What/which laptops do you use yourself?

 

What's the chance to get Samsung 1TB PM981 SSD if ordered from Dell or Costco? Anyone has idea? I heard that only 512GB models have more chance to get samsung. 1TB models will get Toshiba trash most likely.

 

No one really knows sadly. Most people would never know the difference between a PM981 and the toshiba drive (most as in like 99% of people).

 

Right. It only shows difference in test scores. But it's still annoying.

 

Can’t say but I got the Toshiba 2TB and it benches very well, beating the Samsung in 3 out of 6 scores.

 

Which part is annoying? Not knowing which SSD you'll get?

 

And what is max teperature when you run ATTO Disk Benchmark 3 or 4 times in a row? Have you compared the 4K speed? From what I know the Toshiba NVMe drives Dell used in last years Alienwares boiled to death. Because they run all too hot.

 

Okay, I believe I have resolved my issues. I undervolted my CPU and managed to get to -0.140mV stable. I ran Cinebench R15 and managed to get a score of 877cb at a max CPU temp of 82C. I ran Realbench and got a combined score of just over 104,000 with a max temp of 91C.

I also resolved the issues of the GPU being underclocked when on battery. I didnt do anything in particular but I found that on battery you need to run something graphically intensive for about 5 minutes at the downclocked rate and then suddenly the fans start ramping up quite a bit and the boost clock goes to 1650Mhz and stays there with a temp of 72C.

The above was with the i5 8300H and the regular 1050. I am pretty impressed with these results.

 

As stupid as it may sound, I wouldn't recommend any laptop of 2018 or 2017. There is no good one. I was hoping for the XPS15 may be decent enough for my high standards, but it isn't, and I made the mistake to wait 30 days and can't get a money back anymore. I was long happy with an Asus Zenbook, and it is in most parts better than the XPS, better chassis, better keyboard, no coil whine, you can regulate the fans, good latency, has S3, no c state bug. The only reason why I did not buy a new Zenbook Pro was, because I read, that the cooling isnt actually that good, but I guess it is still better than the XPS anyway. Seems also, that MS literally killed laptops from 2018 on. Because if no laptop has working S3 anymore but just modern standby, it is not usable anymore as a laptop, at least under Windows.

You have no idea, how good it feels to be in that other 1%.

Switched from SKH 500GB to a Samsung 970 EVO 500GB, the differences are worlds.

WoW starts 2 times faster, loading times 3-4 times faster. This is mostly true for all games which use data blob files
Decompressing archive files 10-20x times faster (installing programs, addons ect, because all are compressed). SKH write speed was like 100KB/s where Samsung can handle it with 1-2MB/s
Opening large IDEs faster, working with large projects which has lots of folder and data structure way faster

 

Have you tried out the different workstation laptops like Thinkpad P52, HP ZBook G5 etc?

 

Thanks to all. The problem was solved.

The Delock adapter simply didn't work:
https://www.delock.de/produkte/S_63312/merkmale.html?setLanguage=en
I was surprised, because it worked with XPS 9350. Therefore I didn't expect malfunction with Precision 5530.

I got i-tec USB-C Display Port Adapter and it works perfectly.
https://i-tec.cz/en/produkt/c31dp-2/#pll_switcher

 

I'm glad you've noticed a difference. I've seen tons of game loading tests (even in WoW) and there is basically no difference between a SATA SSD and a NVMe. I do see the utility for working with large data files and video editing. For small files, optane is likely the answer due to its faster 4k random reads (which NVMe isn't much faster than SATA at). Sadly those drives are so expensive, and still not worth it if gaming is the priority.