Dell XPS 15 9570 benchmarks + temps

I'm thinking about the same truck!!! Congrats man - and good review that you actually really liked the 9570. I'm hoping I will too.

 

Yeah the laptop is great, but in the big scheme of things I really don't need it when I have the desktop and another laptop I can use. I do feel bad returning it to dell, but I suppose that's why they have the no hassle return policy.

The truck is awesome, still not mine yet (and I'll likely get one without the dusk package), but it rides so nice. I'll be sad to see my Tacoma go, but I need a more family friendly option. I've only been eyeing the Colorado since GM first said they were going to release it, and even more when they started putting the baby duramax in them.

 

I had a Tacoma as well, 2012 Model. The new Tacomas are actually too small for me as I'm 6-5 and they offer no height adjustment on the seat. The Colorado and Canyon were more spacious inside.

 

Cool, Been going through these same trucks for my family. We do a lot of backcountry exploring here. I had a Jeep Wrangler unlimited and had a sponsorship deal with teraflex for our travel blog. We had pretty well the entire teraflex catalog on it minus bumpers. We need something to replace our jeep, So we were looking at compact pickups. The Toyota, while awesome, it mentally expensive here. I cannot order a base model with short box and 4 doors. So the Colorado is the winner, W/T trim level, 4 door short box. Seeing as the entire suspension system, tires and rims etc will be tossed anyways, we are not needing the z71 package just to take it all off. We use aftermarket for all of that. The desiel, while awesome, is not worth the extra coin IMO. You don't get Massive milage gains from it, and I am not towing heavy things, so the v6 will be the better option for most. How about that dell xps!

 

Indeed, how about that XPS. LOL

 

Although you could say that some folks treat their laptops like Off-road trucks - you buy it and strip it all down, replace the bad parts (thermal paste), fix the mediocre parts (VRM cooling), and put on the aesthetic touchs (dbrand skin?).

 

It's awesome. Like buying an AEV built JK wrangler. Already built to tackle anything.

 

Hah half the fun of owning laptops and cars is modifying them in my opinion. Shipping the 9570 back today and I'm actually quite bummed about it. Oh well, once I recoup some of the costs from buying the ZR2, I'll probably pick one up again. Perhaps I'll wait till they show up in the outlet.

 

@custom90gt, @pressing I am working on mine now. I also seem to able to hit -145mv on the cache. I can set any value for the Vcore and it does seem to still "work". I wonder if it really works or if it's actually set to Vcache

 

After a series of experiments and observing the voltages, I am inclined to agree with @pressing, If Vcore is set below Vcache, Vcore=Vcache is forced.

 

I've been doing the same with mine today and I agree with you and mines on -145mv cache also. Any more an I get TS bench errors.

And I also found if you use Dell Power Management app and select Thermal Management - Cool, it totally disables Turbo and the processor uses about 15 - 19v instead of 45-50v.

 

Had a bit more of a mess about with Throttlestop and added a -50mv to the Intel GPU and System Agent and its running nicely now.

Userbenchmark scores

http://www.userbenchmark.com/UserRun/9432945

UserBenchmarks: Game 45%, Desk 113%, Work 80%
CPU: Intel Core i7-8750H - 100.2%
GPU: Nvidia GTX 1050-Ti (Mobile Max-Q) - 37.9%
SSD: Samsung 960 Pro NVMe PCIe M.2 512GB - 221.2%
SSD: Samsung 860 Evo 500GB - 109.4%
RAM: Unknown 4ATF51264HZ-2G6E1 2x4GB - 67.6%
MBD: Dell XPS 15 9570

Not sure why the Samsung 960 Pro is saying below average on Userbenchmark, the Magician Software is reporting about 3K read and 2k write and I've tried it with both the MS NVMe and Samsung drivers, (The Samsung drivers were slightly better but not by much).

 

So you have actually bought one now?

Sent from my SM-N950F using Tapatalk

 

What exactly does the iGPU undervolt achieve? And what does the system agent actually do, is it worth undervolting?

 

Undervolting the iGPU does exactly what it sounds like. The intel internal gpu needs power and can easily be undervolted. Even in my carbon x1 the igpu undervolts by 100mv. I recommend AIDA64 to test the internal GPU if you choose to undervolt it. Undervolting the system agent really doesn't do much so I never messed with it.

 

I agree, and found undervolting the GPU and SA as well as the CPU made for even lower temps and I left it running benchmarks all day with no problems at all so left it like that. I think a chap on Reddit said that if you do just the GPU without doing the same to the SA it doesn't stick in Throttlestop (similar to the Core/Cache issue where I believe TS uses the Cache value over the Core value).

Pleased with its performance and temperatures now so its a keeper! Now if only i can find some actually compatible Ram.....

 

"Uncore" is a term used by Intel to describe the functions of a microprocessor that are not in the core, but which must be closely connected to the core to achieve high performance. It has been called " system agent" since the release of the Sandy Bridge Intel microarchitecture.

I'm guessing other 'stuff' that needs power and makes more heat. Plus my wife likes lower electricity bills!

 

I've never touched the SA and the igpu undervolt holds. I've verified this with GPU-z and hours of igpu undervolt testing.

 

Ok, good to know and I'd happy take your knowledge as fact over the other chap, as you clearly test & document your changes properly and know what you are doing. My logic for this was that the less amount of power I can give the processor as a whole (without affecting performance too much or becoming unstable) the cooler it should run. I was seeing the power useage in TS go well above the 45w TDP so did as much as I could to get this down to the 45w to allow the cooling to work at its design limits.

While I'm on here I've just found an interesting quirk with the RAM, if I put a stick from the 9550 in slot a on the 9570 and origional RAM in slot b, the machine boots and sets the clock speed to that of the lower stick.I think this machine uses slot b to set the RAM clock speed and when replacing 2 slots at once it has to be CL 18 or 19 to boot.

 

There isn't a set slot to set the SPD on ram. The system will look at both sticks and run the profile with the slower/looser timings.

 

But it wont work the other way around? With exactly the same Ram but 2 sets of sticks work this way. I tried this after seeing it on a DELL forum where the chap mentioned it may work.

https://www.dell.com/community/XPS/XPS-15-9570-3rd-party-32GB-RAM-issue-booting/td-p/6088199/page/2

 

Maybe I didn't read it correctly. You're saying if you put a stick of ram from the 9550 into slot A it works but if you put it in B it doesn't?

 

Yep. I can put other RAM in slot A but as long as I have the OEM stick in slot B it works perfectly but wont boot at all the other way around (but lowers the clock speed to that of the lower RAM). Weird huh?

 

That is weird, and if you put just the 9550 ram into A or B does it boot?

 

Nope, not at all. Won't even let me to go into BIOS either.

 

Weird, I know the 9570 has been picky with ram, but that is just stupid. Way to mess that up dell.

 

I know right?! I don't feel like Dell will address this compatibility issue with future BIOS updates judging by their replies to postings on the Dell forum. Their RAM is very expensive for what it is, but the 3rd party sellers should actually physically try the RAM in one of these models before they claim is compatible.

I also believe the only Ram that will work together in both slots has to be CL18 or 19.

 

Thanks for the info guys. I guess my real question is: even when using the dGPU during gaming, does undervolting actually do anything because the iGPU wouldn't technically be engaged?

 

Yes in theory as although you aren't using the iGPU mainly, everything else on the main processor is undervolted and producing less heat so helping the system as a whole to stay cool and stop the throttling. That's my understanding of it all but I'm just a noobie to all this.

 

For what it is worth, the iGPU is actually never disabled. The dGPU doesn't output any display, but only does calculations. Once the calculations are done, they are sent over to the iGPU that does the actual displaying. Usage of the iGPU during gaming is fairly low though. I believe about 5% for gaming at 1080p, about 20% for 4K gaming.

 

I've bought the XPS 15 9570 with the following configuration a few weeks ago,
and got it last thursday, replaced the SSD and the thermal paste...
• i7 8750H
• 1050Ti MaxQ
• 16GB DDR4
• 512GB 970 Pro
• FHD Screen
• 97WHr Battery
→ Repast with Thermal Grizzly Kyronaut
→ Undervolt Core -175mV | Cache -150mV | iGPU -50mV | SystemAgent -50mV

So, because it is very similar to @custom90gt I decided to run the same
benchmarks to compare his results with mine. Here are the results...

After mods → @custom90gt
Repast + Undervolt → Me ( PCMark 8 isn't right, I've used other settings, sry)

The most results are totally okay for me, but all the results who go up to
10% and more are strange in eyes. For example look at RealBench, his score
was over 110.000 without doing anything and over 120.000 with the mods,
but mine is only around 90.000.
Of course I also know that every CPU/GPU is different, but are the differences be so big?

For the benchmarking I set up Windows 10 totally freh/new after I installed
the 970 Pro SSD. Than I installed 'ThrottleStop' and was doing the undervolting.
And befor I used the benchmarks I closed all the programs in the backround.

Can someone tell me what i've done wrong, or if this is a normal difference?

 

Can you post your settings in TS? You might want to zero offset on iGPU and undervolt 1050Ti MQ using MSI AB to get better scores. Nv GFE services and OS optimisation is needed to get a boost.
Did you install Samsung NVMe driver if you're on AHCI mode.

 

@custom90gt did more mods than just a repaste and undervolt. Read the first page carefully.

Also @custom90gt probably has repasted dozens of these XPS laptops over the past few years so his repaste job maybe better than yours is. One way to check your paste job quaility is to run a stress test and see if all the cores run at about thesame temp (say within 2-3C).

And maybe he won the lottery and just has better performing CPU and GPU.

Also that RealBench test is brutal on the XPS as it really pushes the power and cooling solutions. Not really applicable for day to day but good for testing and tweaking...

 

I just saw that the PCMark8 was with other settings, need to give them a new try.

@Vasudev
Here are the screens
https://www2.pic-upload.de/img/35563439/1.png
https://www2.pic-upload.de/img/35563442/2.png
https://www2.pic-upload.de/img/35563440/3.png
https://www2.pic-upload.de/img/35563443/4.png
https://www2.pic-upload.de/img/35563444/5.png

I've installed the newest FW/BIOS for the whole system, also the samsung FW for the SSD
and set AHCI mode on in the BIOS.

@pressing
I know that he did more mods, but as I wrote for e.g. RealBench has a big difference also
compared to his resulsts without any mods/ at stock.

That with the repasting is a point, but i've repasted alot of PCs so I don't think i made big
mistakes repasting the XPS. I will check the temps while I run a benchmark .

The point with the lottery could be the reason :/

I also know that RealBench is not a all day use case.
But as you wrote it is good for testing and tweaking


Don't get me wrong, I don't care about the benchbark results, but if I have such a different
to a simliar system I think there is something wrong.

 

Did you use TS bench to test and set your undervolting to your unique processor, just asking out of curiosity as mine cant run with your settings where you've left Turbo Ratio Settings on 41 for all 6 cores? Plus, what setting are you on using Dell Power Manager, Cool - Otpimum etc..

And you were on mains power, right?

Oh, and change your TS settings (bottom right) to OK - Save Voltages Immediately.

 

@Woodking
I've tested it witch RealBench, Core does max. -175mV (all above crashing into a bluescreen).
I us for testing the Benchmark and Stress Test of RealBench, but need to test it with Prime95 for a longer time.
While testing I plug in the power and set the settings to 'best performance' ("beste Leistung" in German).

Why change the settings? I've done it like in the linked post of @custom90gt said it ... https://www.ultrabookreview.com/10167-laptop-undervolting-overcloking/

 

You MIGHT boost performance with these tweaks:

1. Dell Command I Power Manager --- Thermal Management --- Ultra Performance

2. In ThrottleStop--- main screen ---check SST --- type 0 in the EPP area ---type enter ---click save

Some people need to disable turbo or mess around with SST due to the 9570s weak cooling or to get around the throttling schemes...

 

"While testing I plug in the power and set the settings to 'best performance' ("beste Leistung" in German)". Just checking - that is in Dell Power Manager - Thermal Management - it should show Optimised, Cool, Quiet, Ultra Performance?

Maybe its thermal throttling.

 

@pressing
Nr.1 is done, it and change to "Ultra" for testing
Nr.2 is also done, typed in 0 saved and restart
After that I run the RealBench benchmark again...

Still no good results.. looks like the i've got bad numbers in the lottery

@Woodking
I changed it to "Ultra" ... in the first case I do not use the program from dell, just the Windows-Energy options

 

Like @Woodking said you need to adjust active core multipliers and Save voltage immediately option in FIVR.
In Options, you need to uncheck Ac on battery off. Set customized profiles for AC and battery.
In TS main window, uncheck speedstep,c1e and bdprochot. Enable EPP(SST) in main and TPL windows.
Make sure you set cpu and cache to same value to apply undervolting successfully.
@pressing can give his insights on XPS since I don't own neither an XPS nor 8th gen BGA chip.
In MSI AB you can undervolt it using ctrl+F and it will need better tweaking to adjust between max performance at optimal MHz to get consistent GPU perf.

 

Hahaha - that is no improvement. My XPS with 6th gen i5 has slightly worse results.

Are all core temps within say 2C or 3C during Prime 95 or similar benchmark?

Check out comments by @Vasudev

Without further mods, you might not be able to do better . . .are there other posts here with RealBench score with just undervolt & repaste so you can see reasonable score...

 

@evomod,

From my experience, @custom90gt 's stock realbench is anomalously good.

Your results are poor. Eeither:
1) Your undervolt setting are not really applied correctly
2) You repasted poorly
3) You ambient temp is significantly higher than @customgt (which was 21c?!?)

 

@Vasudev

- What do I need to adjust by the active cores?
- Save voltage immediately is set now

- Done, too

- Okay, but first I need to read how to work with Afterburner, never used it

@pressing

In my case it is about 4°C

@abujafar

1) You can see them at page 24. They are around the ones from @custom90gt
2) Don't think so, but can do it again since @pressing think the same way
3) Mine is 25°C now, was about 20°C while I run the benchmark this morning


In my option, the temperatur is not so important as long as it is kept within limits.
Like in my case, the ambient temperatur is mostly the same compared to @custom90gt .
So it look like there are only a two ways that can explain the differences...
1. There are big differences between the CPUs/ GPUs
2. @custom90gt made other major changes with his XPS (like Afterburner, OC or other thinks)

I will try to do the pasting again and do it better this time to get a slightly better temperatur.
If this don't show any better results, I just let it like it is, because it doesn't look like there is
anything I can do to improve them.

Btw. is there someone who can share his changes with his XPS (settings with different programs)?

 

Hit Reset Turbo button to revert to original turbo clocks. 8750H isn't unlocked cpu.

 

@custom90gt, are you sure that the igpu undervolt works? No matter what negative offset I use, the igpu voltage never goes below 1.095v.

 

That is not so bad. If you run a few days it may improve. You have experience repasting so that should mean your paste was pretty good.

I would not repaste as this does not seem to be a fundamental issue but difficult to be sure without lots of data. . .

You need to look elsewhere for improvements.

 

Try it combined with undervolting system agent, thats similar to what I found.

Its almost as if undervolting the CPU core & cache in TS have to be the same amount to 'stick', along with iGPU and SA.

 

@Woodking, you are correct! it seems that iGPU and Sysagent are connected like Vcache/Vcore!

 

@Woodking how much were you able to push this undervolt and how did you test it?

I am at -150mv on Unengine Heaven and still going!

 

Mine seems

I was stable at -150.4mv on CPU core & cache. On System agent and iGPU -50.8mv. I used TS bench just to check before I did any other benchmarks - if there were no errors there it seemed to be ok in all others (12 threads on 1024m with no errors - result around 100 -102). Pushing either one of the other values gave me errors in TS bench so I had to drop back.

What seemed to make mine a lot better, stable and cooler was lowering the Turbo ratio to the following (In TS).

1 core 41
2 cores 41
3 cores 40
4 cores 40
5 cores 40
6 cores 37

When benching it uses all 6 cores and just sits at 3700 ish solid, stable and able to cool itself. Any higher turbo ratio and the temps creep into the 90c's and it starts to throttle. Its a good trade off between performance and cooling for me.

My logic behind this was to get the total power as close to the 45w (by watching PKG Power value when running TS bench) as I could, as thats what the TDP of this processor is meant to be.