XPS 15 9570 Owners Thread

Has any other notebook manufacturer put forth a design like this? How different is this from the standard copper pipes? It looks legit because of the 8 point mounts, and the fact it clearly covers VRMs as well. I know the vaper chamber tech in the XBox One X is legit in producing a very quiet machine. Although isn't this vapor chamber tech also said to be used in the 9575 as well - albeit it looks to be a combo with the standard heat pipes?

 

It's not too bad until we have some options from dbrand)

 

It's hard to tell if this is a good idea or not. We don't know yet if this is just a piece of copper or a big vapor chamber.
Regardless, it's not clear to me if this will benefit the VRMs or hurt them. At the end of day, you might end up "transferring" the heat produced by the CPU and GPU to the VRMs with this design.

In general, this one piece cover everything will tend to equalize the temp among all components...
I am curious to see how it performs!

 

I suppose we will see how good that heatsink is as people test it. It looks nice. If it fits well, the radiators are big enough, and the fans are powerful enough, I think it could be quite effective. There is a lot of heat to dissappate. And some Razer laptops have a history of running hot.

On another note, some good laptops have been using vapour heat pipes for quite a while now. A tell-tale would be a hollow pipe with an obvious sealed nipple. Here is a quick primer on the technology:

http://odm.coolermaster.com/manufacture.php?page_id=8#!prettyPhoto

 

Agreed. Merging CPU, GPU, and VRM to same heatsink is a problem of engineering in a small space on a small budget.

I suppose this can work on a case-by-case basis with a bit of trial-and-error and a bit of luck. Not sure our laptop engineers bother with either. So that leaves the real beta testing for us : )

 

I mean the very definition of a heat pipe is a vapor chamber to transfer heat. It's pretty much a universal technology and I would guess 99% of laptops have it. The idea of a large vapor chamber like Razer is doing is interesting, let's hope they figure it out. Aside from the 17" razer pro, I can't think of one that didn't overheat like mad.

 

In comparison with Razer Blade 14"

 

Well the mating surface is more serious & the radiators are much larger.

Not sure the larger heat chamber would perform better than traditional heat pipes in small laptops.

 

I like that Razer is offering a full redesign of the aging Razer Blade 14 chassis design, and I like that they have retained a centered keyboard with speakers on the deck. It does further reveal some of the shortcuts and oddities of the XPS 15 9570, including the wireless card, which is the same old Killer 1535. Even Dell's G5 and G7 have the Intel 9260 Bluetooth 5.0 card as standard, while the XPS 15 2 in 1 is even worse as it has the Killer 1435 integrated. Pricing is the the major strength of Dell, with more discounts than what I have seen from Razer, Gigabyte, and MSI.

 

Here are my stock benchmarks for the 9570 i9, 32gb of ram, with the 1050ti. The temps can get high, but the laptop has been rock solid stable...no crashes no bluescreen.

 

Now if only they could make their batteries stop exploding...

 

Thanks for the post @cpaek72 . I see a 98C max on RealBench so perhaps you are hitting some throttling; with a simple undervolt perhaps you can get well over 101,562.

One guy showed his i7 factory paste was a disaster so you may have additional upside there with a good repaste using an extremely small amount of thermal paste.

 

Yeah, undervolting will be my next test. I'll post new benchmarks after I do that. Any suggestions which settings I should use for Throttlestop for the i9?

 

Start with -135mv on both core/cache. Other users have reported success at this level.

 

Ok, undervolted -135mv on both core and cache. Here are the results:

 

That's pretty good! You max temp dropped (no more hitting hard 100c wall). Your cpu-bound parts of the test improved by almost 60% (heavy multitasking).
Probably you were throttling during this part of the test in your previous run!

Try if you can lower the core/cache more if you have time.

 

That is a good increase in score and good decrease in CPU max temps.

You should look for better CPU temps as 91C is still hot; I would aim sub 80C but I don't know if that will be possible on 6 cores at high multipliers with those tiny radiators.​

You can run a temp monitor at the same time (like HWiNFO64) to see what the problem areas are.

- Not sure if you have more room undervoltingbut try -5mv increments and run some tests for an extended period of time (say Prime95 for a few hours and maybe some gaming). At this level of undervolt, even another -10mv can provide a significant thermal benefit. If system is stable you can keep trying then back off say 10mv when things get unstable.

- Repaste with good quality paste should give you more headroom after that. I would use absolute min thermal paste possible, after repaste take everything apart to at least once check that you get full coverage and redo, make sure cores are within say 2C or 3C of each other (else you may have thermal paste distribution issue or bent heatsink).

What are your SpeedShift EPP settings?

 

In general yes, however, I think in real bench the max temp will always be around 90C because the cooling system kicks in only 5 seconds after a load is applied. (rationale: keep the computer silent if there is just a short burst)

HWInfo is better because it will shed light on how the temp/clock evolve over time.

I am skeptical that repasting with normal paste (not liquid metal) will bring tangible benefits. Curious to see what people are able to achieve!

 

Why not a binary search approach to find the best threshold?

 

Look at the recent photo of the terrible Dell factory paste job from the Reddit 9570 thread.

That is consistent with the factory paste jobs on the 9550 & 9560. Very thick gobs of cheap thermal paste are good insulators. A high quality paste with uber-thin application shows 5C+ improvements on these laptops.

As the case, heatsink and CPU are virually the same in the 9570, I also would expect a material improvement in 9570 thermals with a good repaste. Time will tell as actual results roll in

 

That looks legit - I may have to consider the 1060 model when it launches in a few weeks, if this cooling system facilitates quiet fan acoustics (truly quiet <43 dB). Although with such a beefy system, the Max-Q chips seem kinda lame.

 

It sure looks like something new for a laptop and promising.

Question is if all the components are happy with the resulting heatsink temperature. I mean, in the 9560 the CPU and the GPU seem to be quite coupled thermally with the heatsink, even though the throttling threshold of the GPU is much lower than that of the CPU. This doesn't appear to be causing issues*, presumably because the heatpipe temperature is still sufficiently below the GPU's threshold, so that enough heat gets transferred from the GPU (but may not always be the case). It shouldn't be a problem if you are sometimes actually heating up a component which happens to be inactive, as long as you are below its operating threshold.

An alternative workaround for the 9560 using another thin heatpipe(s) from the VRMs towards the fins or the main heatpipe, without adding VRM "fin surface" and redirecting some air towards it, wasn't tested AFAIK. I mean, it might work, or it might cause the heatsink temperature to rise too much (which would choke the GPU), or it might not cool the VRMs enough.

*The most difficult issues we've seen are PL throttling while the CPU and GPU temperatures are okay, due to "ambient" VRM overheating. But in many cases there is also direct CPU or/and GPU throttling, which mostly go away with proper repasting.

 

@_sem_ totally agree with what you wrote. We just don't know yet if it works better or not.
I sure hope it does. If it works, other brands can pick up and everybody will benefit.

 

I helped another user on reddit test his machine. He has an i9 and is getting similar results as @cpaek72.

https://www.reddit.com/r/Dell/comments/8ldqtv/recieved_my_xps_with_i9_huge_performance_upgrade/

 

Does any one have benchmarks for comparison of i7 with i9?

I read on reddit that the i5 gets nearly the same scores as the i7 as the i7 throttles so much?

 

This is from an undervolted i7 8750h

 

Humm must be throttling pretty good during the heavy multitasking since my i5-8305g scores 72,583 on it. Wonder if it's thermal or power throttling.

 

@trebuchet

your results look subpar, especially in the heavy multitasking section. How much did you manage to undervolt the core/cache?

 

Not mines I´m still waiting for mine

I took from reddit:
https://www.reddit.com/r/Dell/comments/8kh009/xps_9570_repaste_photos_and_stress_test_benchmark/

What are your other scores? image editing, encoding...?

 

The other areas are not near as stressful. My cpu power limit throttles to around 2ghz when doing the heavy multitasking section.

 

Just for fun, the desktop at 4.4ghz with the wife streaming plex (transcoding) in the background:

 

Unfair
How about that 9560 of yours?

 

It's certainly unfair but it gives a frame of reference what a 12C/24T desktop cpu not thermally constrained can do.

I feel that Encoding and Heavy Multitasking are the parts where improving thermals (reducing downclocking) will benefit the most.

 

Sadly the 9560 is gone and I don't have any realbench benchmarks from it, just tons of logs running realbench stress test. I'm sure there are some results on this forum with the 7700hq though.

I also re-ran realbench on my laptop and OpenCL causes power limit throttling on mine. Image editing only uses about 8% of my CPU in the desktop. I really think iunlocks mod on the 9570 is the way to go in terms of helping to limit throttling.

 

What for? Undervolitng or anything else? I´ve heard people is using throttlestop to undervolt xps

 

iunlock's mod for the 9550/9560 helps to cool the VRM and repaste the CPU/GPU. That coupled with a stable undervolt (I use throttlestop) will help the i7/i9 in the very thermally limited XPS design. If you're curious you should read the 9550/9560 threads on throttling like:

http://forum.notebookreview.com/thr...ures-benchmarks-xps-15-9560-kaby-lake.802345/
http://forum.notebookreview.com/thr...rature-observations-undervolt-repaste.785963/

Doug Black also has a good post here:
https://www.ultrabookreview.com/14875-fix-throttling-xps-15/
and here:
https://www.ultrabookreview.com/10167-laptop-undervolting-overcloking/

 

Great thanks for that buddy

 

From what I have gathered so for by looking at early users reports, the 9570 doesn't seem to have problems with VRM throttling or collapses below the base speed.
Under cpu-bound stress tests, stock machines settle at around 3Ghz/80c, while undervolted CPUs settle at 3.4Ghz/80c.

I think further mods (repasting, etc) can further raise this bar.

Still early, I hope new reports come soon!

 

hehe thanks

 

I just dit the Realbench software for my 9560. -130mV undervolt on CPU (7700HQ). -100mV undervolt on GPU, +100MHz core, + 220MHz memory.

The encoding tasks went really bad for some reason, though there were no issues during the image editing / OpenCL part). It wasn't thermal throttling, but the it wouldn't hold the clock speeds at 3.4GHz (it was constantly jumping between many different clock speeds). Even though if I encode a video via Adobe Mediaencoder myself, it will hold the clock speeds (and it also kept a steady 3.4GHz for the image editing / OpenCL part). I mean, I expected a big jump in performance for the 6-core CPU's. But not over a doubling of performance.




Edit:: Just did the Handbrake encoding again. Max temp was 65 degrees, 100% sure no thermal throttling. Score was 47,000. The FPS of the encoding was constantly jumping between 18FPS and 31 FPS. Seems the Handbrake encoding is just botched for me, Adobe MediaEncoder is much more consistent and faster.

 

While not a fan of his previous "reviews", at least there is something while waiting for PRO reviews:

 

@MrBuzzkill Your scores do seem low, but It doesn't look you are thermal throttling in your 9560. Windows update in the background?

 

Yeah it seems odd that every score is lower than what I would expect. Almost like power saving or too high of a speed shift number is at play there.

 

That score sux.

Something is wrong. Repost to the 9560 thread along with your ThrottleStop, HWiNFO, etc. snapshots.

 

Is it just in North Europe that it is impossible to buy the FHD, the biggest battery and the i9? I don't understand this nonsense from Dell, especially considering that every laptop is built to order!

 

No, it's the same way in the States too. Dell is the new Apple don't you remember? Kind of like how I couldn't get 16gb of ram with an i5 in the 9575. They make you purchase all the stuff you don't need to get the stuff you want.

 

I just found a very interesting video on YT. It's the XTU stress test on a 9570 with i9 CPU. No mods all stock.
In short, it thermal throttles up to 3.5Ghz. There are flags of Power Limit throttling. NO VRM throttling.

 

The Intel XTU stress test is rather lightweight. It provides excellent graphs and easy to follow data so is worth running as a first test. Then use those results to tweak and move on to more aggressive troubleshooting based on your usage.

Intel XTU provides good access to some tweaks, including undervolting and power limits. It was buggy with the 9550 and 9560, however.

 

Indeed, those flags were "VRM throttling" in the 9550 and 9560; the name was misleading but the typical cause was that VRM ambient sensor getting hot (well hot per Dell which was over 77C).

Not yet sure what is the cause in the 9570.

 

Is it possible to swap out the battery yourself, ie upgrading the 56W for 97W? Don't mind opening up the machine but if it involves soldering etc it might be a bit to much work...