XPS 15 9570 Owners Thread

Maffle you seem like a man possesed to get to the bottom of this. Hope something good comes out of it eventually

Sent from my SM-N950F using Tapatalk

 

I know, it was only a polite way of making him realise that he was mistaken.

 

I think I may have found a worthy contender.... been holding on to my £'s since I sent the 9570 back and using my old XPS17 as I couldn't find anything I liked.

Alienware Area 51M - 17 inch.

https://www.dell.com/en-uk/shop/gam...1m/spd/alienware-17-area51m-laptop/n00aw51m02

Now this looks like it can cool itself, look at those exhaust vents!

 

Hey guys,
after padding all VRMs and not letting the GPU use more than .93 Volt at 1.800 MHz as well as turning off turbo boost for gaming, I am pretty happy with the laptop, even as a gaming machine. The 1050 ti can run Battlefield 5 and modern games in FHD at low/medium settings. But I have two questions about further improving the performance:

Would uninstalling DPTF (and keeping it uninstalled through the Windows pro Group Policy editor) get rid of the 74 degree thermal limit of the GPU?

Do you think there would be any danger, as long as I make sure that the GPU does not go above 78 degrees?

And would setting the internal GPU rather than the the DGPU in the Nvidia device manager get rid of the annoying fan spinning that came with BIOS 1.7.? I would then only set the 1050 ti to be used for specific applications, but change the global setting to the internal GPU?

Thanks for your help! And Jesus, what is wrong with Dell?

 

@ Mulgul I wanted to test that too (though I had no reason so far because I dont have a game where it reaches over 66°C), havent done it so far. I have upgraded to Pro though for $4 and disabled once and for all the nasty auto driver updates with gpedit.msc, too bad the other workarounds dont seem to work anymore properly. Not sure how safe it is though because of bad modern standby, might be risky if your laptop fries to death because of modern standby going nuts and then no thermal throttling kicking in in your bags or next to your bed.

My 1050ti works stable at 1709 and 0.875V, havent tested 1800 actually so far. I would like the lower voltage more than a tiny bit of more performance. Dont overclock the VRAM btw, it gives literally zero advantage because of the capped 128bit memory interface.

I actually am testing a workaround for the nasty fans with 1.7.0 right now, by simply deactivating the Nvidia card in device manager until I use it, then deactivate it again. It works so far and I havent encounterd issues. I read before that people were talking about problems caused by driver toggling on/off for a GPU under Windows, but so far it works, and totally gets rid of the fan issues.

You can download the MS tool devcon ( https://docs.microsoft.com/en-us/windows-hardware/drivers/devtest/devcon ) and then create a script toggle like this:

devcon enable "PCI\VEN_10DE&DEV_1C8C&SUBSYS_087C1028&REV_A1"
devcon disable "PCI\VEN_10DE&DEV_1C8C&SUBSYS_087C1028&REV_A1"

That is my 1050ti device ID, and then just create a shortcut with admin rights. Havent tested yet though if this causes a modern standby drain, which wouldnt surprise me... (Update:... ok it does (-: https://i.imgur.com/NB63Ado.png how could it be different, you deactivate a device, and it kills modern standby and fries your device. I love it. and that kills this workaround too. I just love it. I love it. I really do.)

I actually found out, that there is some nasty bug in Windows 10 with Optimus / MS Hybrid. Remember when I said, one dGPU poll was when you clicked right click on an EXE file in Explorer, causing the fans to spin? Well, I find out, that it is actually not caused by the Nvidia context menu dll hook which I disabled. There seems to be some stupid bug in Windows, where some context menus triggers a short 100ms dGPU poll the moment they are loaded into RAM, then when theyre there, it doesnt happen anymore, until next load. I tried around right clicking everywhere randomly in Windows 10 and most of the times I could trigger the dGPU for randomly context menues, even in Chrome, on the start menu, Explorer, and others. Also just scrolling the scroll bar in ThrottleStop caused this sometimes... (mostly some HW accelerated GUI elements like scroll bars menus ect all over Windows) And if you have Nvidia Experience installed that also randomly causes a short dGPU poll... so it is just not accetable, Dell tries to go through the wall like this with this nonsense lazy programming.

Deactivating the Nvidia card in device manager then totally got rid of the fan spins. If you reboot the laptop with a deactivated or not installed dGPU the fans btw never go off (-:

It is just RIDICULOUS how STUPID and IGNORANT and INCOMPENT Dell is, by binding the fans to the dGPU like this, without probing the temperatures and totally ignoring them. And they mostly will then just say, well, too bad, not our problem, MS has to fix the random dGPU polls in Windows... which I agree to a certain point, but Dell also has to change, that the fans just turn on bound to the dGPU temperature.

 

The windows dGPU polling issue has been happening for years now. I don't think MS will ever fix it...

 

Yepp. I know. It got a bit better with the last major updates, but there are countless little places still here and there, where a short dGPU poll happens, like these context menu polls for example, or whatever, mostly a LOT. The problem now is... this was irrelevant before Dell changed it with 1.7.0 to the AMAZING EPIC IDEA, to spin on the fans the nanosecond the dGPU has activity (and dont put this in the changelogs btw). They cant let this be in this state. I have actually HwInfo open and everytime randomly the fans spin on, there was a short dGPU poll, just while typing this, there was a poll happening in the background by what I do know... It happens a lot durent a first Windows boot and then settles down a bit, but then it happens still everywhere around all over Windows, mostly with context menus, or start menu ect.

 

Yeah what a dumb move for Dell to spin the fans up whenever the dGPU is activated. I was going to pick up another XPS but decided not to until they fix the GPU throttling and fans...

 

Thank you for the awesome reply @maffle!

I am just not upgrading to 1.7 then, since the only real bugfix (GPU not clocking up when cold) is not that important to me. By the way, I also use .893 Volt at 1683 Mhz for lower voltage and 1797 Mhz at .931 voltage if I have the temperature headroom (not having it was also why I asked about raising the 74 degree temperature limit.

I would not be worried about hybrid sleep turning the XPS into a fiery hell (another genius Dell / Microsoft idea), since I just use hibernation now (thanks Dell). Still: no one knows whether deactivating DPTF would get rid of the 74 degree GPU limit?

 

I don't think that getting rid of the DPFT would help with the 74C GPU limit. It's in the vbios (attached to the bios).

 

am I the only one here whose dGPU doesn't trip constantly (on 1.7)?
And before the pitchforks come out, I have standby issues, crackling issues (fixed with modded drivers), replaced wifi with intel and undervolted, I had the 48 degree dGPU throttle in 1.6, so yeah I'm in the same boat LOL.

 

Nope. Mine doesn't either however I have it virtually "unplugged" using linux, and fans never turn on. What I do experience is sound distortion as if the speakers were loose vibrating against the metal chase, which makes a horrible noise even at medium volume.

 

Isn't the real problem here people thinking the XPS15 to be a gaming laptop? Or rather Dell for making people believe that by pretending it to be in marketing etc.

Sent from my SM-N960F using Tapatalk

 

@improwise Well, if there is a 1050 ti in there, you want to be able to use it, right? Otherwise no need to put it in there. No one expects the performance of e.g. A GTX 1060, insane overclock capabilities etc, but you should be able to use the hardware you are buying.

 

Yes, you """are""".

...

Nope. I knew before what the 1050ti can do and what not, otherwise I had not bought this laptop. I use it for mild games, where the dGPU never goes over 66°C, encoding and CUDA libs for deep learning programs.

 

So I just got an email back from the Dell "engineer":
"We have received an update from product group team. There are no plans on releasing a BIOS with the GPU throttling at 78C again."

I told him that was not an acceptable answer so we will see. I'm glad I sold my XPS now...

 

Oh wow! That's REALLY upsetting if that is 100% true.

 

Yeah I was fairly shocked about it. I will say my response was probably less than nice. It really is not acceptable though. And if it's really the case then I will make sure that everyone knows about it.

 

Exactly!

 

Can't a manufacturer get somehow sued for this? They're technically decreasing your machines performance without previous update. They should get somehow in trouble for this.

 

@custom90gt I told you so all along (-: It was obvious they reduced it secretly and never intended to change it back. They will mostly also not revert the dGPU poll fan "problem", they implemented it by their will and wont change it back. Mostly all to reduce the return numbers they get by fried laptops, mostly a lot by modern standby, and just people who dont know how to use a Windows laptop properly. Problem is with all these people who game on the XPS with no FPS lock at 60, no undervolt, and then complain it melting. It would be hilarious, if they throttle the CPU too now with a future bios and wont revert/allow a revert. I am, unfortunatly, back to 1.6.0 now, the fan problems were just not accetable of 1.7.0. The dGPU polls happen every few moments still, but it wont trigger the fans. With that being said, everyone should have a 100% allowed money back actually, regardless if you bought it like half a year ago. Dell shall rot on tons of returned 9570 they cant resell again. #DellCares Kappa

 

Could you let us know the email of the engineering? I wouldn't mind to send him an email as well.

 

What do we know about other similar laptops running Windows 10 - do they have the same problems? Like running Windows 10 on other similar laptops or Linux on this one, is it equally bad?

I don't doubt that many of the problems are Dells fault, but if Windows 10 suck in general, its a bit hard to blame Dell for that and it should then be obvious on X1E etc.

 

Sadly I do not have a direct email address, it's just a generic [email protected]. I'm really disappointed in Dell. The XPS is a great laptop with a few issues that I think should be fixed. I am trying to figure out the best sources to put Dell's response out there. I'm hoping some negative publicity may change their mind.

 

I found by random luck another awesome "bug", Dell would call it a feature (not sure if this just happens on 1.7.0, I found it on 1.7.0): If you have selected "silent" in the Dell Power Tool, and then if the dGPU rises over +90% load, it throttles to 800-999MHz REGARDLESS of the temperature (could be even 40°C). I found this behavior by luck, because I wanted to try out HDMI output on a 1080p monitor for a game.

I had some FPS issues, but then noticed the dGPU clocked to 800-999MHz for no reason it seemed. I found out, this mostly was related to buggy 4k to 1080 window scaling by Windows. so if you scaled a windowed game to windowed fullscreen on the second monitor, after you moved it over, it was internally still drawing with a 4k resolution.

This caused the dGPU use a 99% usage, and that caused automatically a 800-999MHz downlock regardless the temperature was fine (60°C), minimizing the window back, dGPU clocked up again. I can easily reproduce this, by starting a game on the 4k monitor with HDMI connected and extended desktop. Then move over the game to the 1080 monitor, then toggle window min/max. The moment the window goes max, the dGPU throttles to 800-999MHz because of usage goes over 90%.

You can see the behavior here: https://www.dropbox.com/s/v5tzvf1vdl8mm1m/dgputhrottle.mp4?dl=0

Second temperature value of the first line is dGPU (is fine, 60°C).
Last line first/second values ared GPU clock and usage. (automatically throttles to 9xxMHz after +90% usage, regardless of temperature).

I can reproduce this on the 4k panel alone too, by opening a game in 1080p windowed mode, normal clocks, usage around 60%. Then max it to 4k, dGPU rises to +90% usage and it automatically down clocks to 8xxMHz, even dGPU temperatur is 58°C. Or just like here, slowly rising the dGPU load by zooming out step by step and then back in:

https://www.dropbox.com/s/pl2hbk2ithjfm6r/dgputhrottle2.mp4?dl=0

There is totally no need for this, becaue fans and temps are all low and fine. I just love this laptop (Dell). The bios/EC is a pure joke.

 

A small start How Dell cripple performance explained by....

 

Apple did that with my wonderful iPhone 4S. That is presumably to sell more new units via accelerated obsolescence(TM).

I just won't buy any new products from Apple or Dell.

 

Yes, this should completely fall under this category the same as Apple did with the iPhones.

 

And that's why I blame Dell, I would have preferred an Intel gpu only option as the cooling has enough problems handling just the CPU.

Sent from my SM-N960F using Tapatalk

 

This isn't about what the 1050 can do, it is what it can do in an XPS 15. Don't think I have to tell you about the shortcomings of the later

Sent from my SM-N960F using Tapatalk

 

You'll soon get a third one

Sent from my SM-N960F using Tapatalk

 

Hah third? You mean 4th or 5th. I don't recall, lol. I have 3 sitting in my office that I either need to sell or return to Dell (silly customers changing their minds).

 

Are you talking about your customers or Dells?

Sent from my SM-N960F using Tapatalk

 

@custom90gt anyone know an easy/reliable way to push the 1050ti of the 9570 to 75°C? Ive tried several ways so far and cant push it over 73/74°C, it literally stays at 74°C with proper clocks, regardless what I try (except for that one Farcry5 example you gave, I dont own that game), and therefore wont throttle. I tried even opening two times MSI Kombustor fr 30 minutes (it is not throttling by driver to 800MHz like Furmark) + Prime + caffee and the only thing I would achieve is throttling because of the CPU power share, but not because of reaching 75°C of the dGPU.

Honestly, no. Ive run benchmarks and they are on par what other 1050ti can achieve in other laptops.

 

An option for some people who want to bypass some dgpu issues is to get an egpu if you want to game. I use an Aorus Gaming Box 1080 which I have had since my 9550 and now 9570. I play all games in 4k (connected to 4k TV) but not everything maxed out (depends on game) and it works well (plug and play). I have the dgpu disabled due to many of the issues people speak about here. I am still on BIOS 1.5. I also have my 9570 CPU undervolted -125mv on core + cache. Haven't repasted yet but have some AC5 ready if I get round to it. I know this isn't probably an option for most of you but it may help someone.

 

Xps15 is not a gaming laptop.
What programs do you use to overheat dgpu over 75c?
I played gta v and latest nfs and i have 60-65c in 1080p ultra.
I rendered file with Resolve and also have 60C.

 

Looking at my benchmark thread I see that realbench typically had my GPU running at 75C or greater, maybe try that? I know Far Cry had no issues getting it nice and toasty, I could watch the FPS decline. My guess is any new AAA title would do the same (Battlefield V and the like). If you really wanted to try, you could increase the TDP limits of your cpu with throttlestop.

@Cooler-master, GTA V is more of a CPU test than a GPU test.

 

That wont work like I said, there is also a mechanism, that if the CPU is going near its TDP limit, the dGPU will throttle down, regardless of its temperature, maybe this happens with Farcry too? Because if I do a Prime95, the dGPU will throttle to 800MHz even under 75°C.

 

I did not experience that issue. All of my GPU-Z logs show that it was throttling due to temperature and not power limit on the 1050ti.

 

@maffle you can try Heaven, If I let it run long enough it will hit 74c. Also Luxmark (the OpenCL/GPU part of Realbench will hit the GPU pretty hard).

 

My XPS have PM981 512gb. It is too small for me. I want to upgrade it to 1TB TLC nvme SSD.

What is the better: Samsung 970EVO or WB Black 2018 3D nand(not 2017 with 2D nand).
At first place is energy management and power consumption.

 

@Cooler-master The new WD Black is mostly very close to the Samsung now but I would still take the Samsung. The XPS doesnt support PCI Express Active-State Power Management (ASPM) anyway.

 

@ abujafar Ive run Heaven on 1080p@Extreme now for 30 minutes but the dGPU wont go over 70°C, having 99% load. I even reset undervolt so it would run at 1V, which wont help anything. Maybe it is too cold right now in my room -.- I wanted to test if removing the Intel DPTF would have any effect on the 75°C throttling but I just cant get the dGPU to 75°C.

Edit1: After 45 minutes it is now at 73-74°C but it doesnt want to jump over to 75°C ...

Edit2: After 1h 30mins there is still no serious throttling, the thermal throttle flag is triggered, yes, but the dGPU stays around 1500-1600MHz all time, no extreme "shutoff" throttling to 800-900MHz. I thought this was the case we were speaking about? I even added one prime95 thread to heat up the CPU, because these benchmarks have so little CPU usage:


This shutoff throttle to 800-900MHz actually, like I said before, I just saw when the CPU power usage was going over >=40Watt, then the dGPU was throttleing to 800MHz, even temp was below 75°C. I noticed this actually if I do a prime95 with all threads + benchmark. This could maybe happen in games which cause a lot of CPU usage.

 

Here @maffle
I went through some of the data I have and here are my results from 16 minutes of Far Cry 5 gameplay. The system was undervolted by 130mv on the CPU/Cache and GPU was set to run at .9v. The table is the average over that time-span and the chart shows what the GPU clock rates were doing. Looking at GPU-Z logs from similar game-play, the throttling reason is temperature.

 

So if turning off / uninstalling Intel DPTF does not get rid of the GPU 74 degrees power limit and if it lets the CPU get as hot as 97 degrees (mine does) then what is the use of uninstalling Intel DPTF at all? Removing Intel DPTF cannot get rid of power limit throttling because that is essentially a hardware issue, right?

 

Uninstalling DPTF would likely exacerbate the issue because the CPU would heat up the GPU and cause it to throttle even more.

 

There's a thread on Dell's forums, where someone disabled DPTF and said it disables the 75C (or any other temperature) limit. That's basically the only workaround.
But yeah, as said earlier, don't do things like putting your laptop to sleep in your backpack with DPTF disabled and be more conscious of pushing the temperature limits...
The nVidia GPU actually has a "maximum" temp limit of 94C, as reported by GPU-Z. It's the DPTF that enforces the "current" limit, which is 75C.

Well, IF one is ready to put up with inconveniences, one could disable the CPU turbo, so that CPU doesn't heat up when DPTF is disabled. It IS an inconvenience to keep enabling and disabling the turbo, of course... Less so if you do it with ThrottleStop profiles though.

 

I would love to see some proof that it comes from DPTF. I would find it odd that the same DPTF installer sets one bios to 75C and the other to 78C. Also on a clean install of windows without connecting to the internet or installing drivers GPU-z shows the 75C limit. I think someone is pulling your leg.

 

Here, a post in your own Dell forum thread about the GPU 78C set point:
https://www.dell.com/community/XPS/...U-set-point/m-p/6201913/highlight/true#M19258

I haven't tried it, so I don't know. But the guy in that post says he disabled the DPTF (although Windows enables it on its own every now and then), and the GPU throttles around 90 degrees.
If the guy is lying then he's lying. But if he's telling the truth then disabling DPTF works to bypass the GPU 75C limit.

I suppose it makes sense - the 75C limit is set in BIOS and can't be changed. But the DPTF is the software that does the limiting based on the number set in BIOS. If someone can hack the DPTF code and make it to limit the GPU at some fixed temperature instead of the temperature from the BIOS then problem solved!

 

Hah that is literally my Dell thread. I don't think it works though. I welcome you to prove me wrong. If you can maybe I'll end up with another XPS, lol...