[Alienware Refresh] Initial thoughts and impressions? Hit or a Miss?

You started the wrong way You purchased a BGA Ops

 

The 6820HK stock should do 3.2ghz across all 4 cores. And it's already hitting 92C with stock settings? Yikes. I'm hoping that dude's AW has a bad CPU paste job.. Coming out of the box that hot with a "mild" load is bad. 3dmark only stresses the CPU in the physics test and not a lot anywhere else. I'd hate to see what happens when he runs Prime95.

 

It's not that unusual to hit 90c under load without undervolting on Skylake, I think you have experienced that yourself.

Hopefully he will give it a try

 

I am running fire strike with a -100mv underclock and it's holding at 87c. Might need to paste it. Had it set to 3.5 on all cores. Hit 93 on physics test.

 

I experienced it on a 6700K stock (almost 90c) lol. But not on my 6700HQ (had 3 MSI's with the same CPU). None hit that high stock with small FPU Prime95.

 

And what would the Maximum processor temp been with pre installed dynamic overclock, advertised from Dell? I think they use 4.1 GHz. Of course on 1 core @quickie What was the stock load voltage?

Edit. 3DM Sky Diver and 3DM11 will load your processor a bit more.

 

this is with -125mv http://www.3dmark.com/3dm/15414062
scores are lower.... but the gpu was at 71 and cpu was at low 70s with a peak of 81.
Papusan how do you check the stock load voltage?

 

The 6820HK with AW sw might be pushing high enough to cause the high temperatures, it's worth taking the time to test things out 1 at a time.

Optimizing the thermals via a software solution is easier than pulling apart a laptop, one of the things I enjoy about software over hardware hacking.

It's best to take a methodical 1 step at a time method, characterizing each possibility for improvement of the thermals, and noting the effect.

Undervolt, maximum frequency - 4 core multiplier, cache multiplier, fan curves, and even application tuning.

Then if the result is enough - stopping the throttling while gaming or other use, you can save the mechanical solution - re-pasting - for later in the life of the *brand new laptop*.

There is still a chance for infant mortality for the hardware components, something might fail in the first days, weeks, months.

It would suck to rip everything apart on day one, only to find out something else needs fixing - RMA involved - only to have the RMA rejected due to the vendor finding out you pulled apart their carefully constructed laptop.

Sometimes that goes ok, after a nice conversation - promising you are an expert in such things and didn't actually muck things up yourself by doing the dis-assembly / re-assembly.

Sometimes it doesn't go so smoothly.

It's better to keep ones leverage in ones favor, rather than going out on a ledge and getting caught out without a net.

The first time I copied a file, then deleted a file, and restored it from backup, I realized it was far easier to do that than re-wire a fix by hand on the wire-wrap board

Software is always the first best solution, if you have a handle provided that lets you control things that way.

Otherwise you have to break out the sticks and stones, or wire-wrap gun, to make the fix

Introduction to Wire Wrap


Still in use today

Wire Wrap Tools - 2016
http://www.digikey.sg/product-search/en/tools/wire-wrap/1245295

 

The undervolt shouldn't affect the score, something else is a contributing factor.

Did you undervolt both the CPU cores and the cache? Or just the cores?

You can try making a change to the CPU cache add undervolt, or remove it, and rerun.

Make sure there isn't something else running, like a browser with active tabs, or a game control app - like Steam/GOG/etc.

You can use hwinfo64 to see the temps, voltages, and power draw live and enable logging for finer grain monitoring.

I usually quit/exit the XTU systray app when running hwinfo64, you only need 1 monitoring program accessing the probes at a time. Plus you can remove the CPU usage of the XTU systray app monitoring.

Nice first result, keep going - from others reports you might get up to -200mV undervolt on the Cores, somewhere along the way you will BSOD.

Make sure you aren't doing important work or doing disk copies when running on a new test setting - if you do BSOD, do a disk check on C: on next boot - reboot to do that check immediately.

 

We all know that enthusiast paste will improve temps, but now you are gaining a software advantage as well. Using both is better than using just one.

And, gaining data before re-pasting gets you a good base for comparison after re-pasting.

Try leaving the cores at stock, and only doing the undervolting first at CPU defaults. It's better to change 1 thing at a time.

Cores undervolt, then cache undervolt, then see how much you can increase the core multipliers at best stock CPU undervolt.

Then you can reduce the undervolt as you OC the CPU cores past stock.

Find the limits now, and again when you get around to re-pasting.

Are you using a laptop cooling pad? At least lifting up the rear an inch or so to improve air-flow?

 

@quickie hwinfo do that job as @hmscott said. Also put max performance and not balanced/power saving profile in Windows.

 

That's because you limited the cores to 35x, try going back to stock CPU settings except for core undervolt then cache undervolt.

Only change 1 thing at a time, and you can correlate effects more precisely - you are sure it's due to one specific change, and don't need to "guess" which of the 2 or more changes are responsible - you have a good chance of guessing wrong

 

You're on fire @hmscott. Great stuff. Keep at it.

Seeing all this temp stuff, OC'ing and UV'ing I'm tempted to see what the new AW can do with the new cooling system.

Hmmm...not bad of temp for out of the box. I'm up for the challenge. Just say go @hmscott. Lol....

A family member does need a laptop and was headed to get an MSI, but let me divert that course...(college student and games.)

::iunlock::

 

He should also try the pre installed Oc profiles out of the box. With and without undervoltage. And with more than the wimpy 3DM Firestrike. Processor benchmarks like Cinebench R11.5/R15 and Wprime 1024M stress tests. Those processor benchmarks is almost like real usage. And a kill-a-watt would be fine in test purpose to see if Batteri boost is used!!

 

This is tempting...repaste, repad, UV and OC... I'd run the heck out of it to see what it can do. Hmmm

::iunlock::

 

The stock on he program was 35x 34x 33x and 32x

 

Go man go!!

Save up some powder for the 1080 when it arrives in November

 

So 35x on all 4 cores should have improved the scores, I would do a compare myself, and will later, but I gotta head out now

You have attracted several really capable people to help you out, hopefully they can stick around and assist you now, and later too

Have fun

 

@quickie

That's a good idea @Papusan / @iunlock , if you have time maybe help him get control over how the AW software is tuned right now, and explore the Windows tuning options as well.

If the AW software profiles are work against your XTU settings it can be confusing, with one setting the CPU up initially and then the other coming in afterwards making changes on it's own.

Right now you want the power profile to have the CPU set to 100%/100% CPU - Windows High Performance setting - unknown what AW sw would call it...

See ya'll later

 

Try =150Mv on core and set all core x36

What about noise levels and gpu temps? Last gen AW had the same CPU temps but great sub 70C even on 980M, try to run Unigine Valley ExtremeHD bench. Does it have G-Sync?

Thanks)

 

Done. It'll be here in a few weeks. Going to bench the lights out of it, get some real world numbers and show what it can do. I'm pretty optimistic that with a repaste, repad and some tuning the cooling system will hold its own. After I'm done with this it'll be going to a family member who's looking for a gaming thin portable laptop anyway so it'll be within the household and I'll have full access to it anytime lol.

I already feel sorry for it before it even arrives, because you know how I drive lol...

Come Nov. I'll get the AW1080 as well as that's the one that we all have our eyes on to see how it stacks up to the many others that are currently available. I'm curious to see what kind of numbers it can produce with the GTX 1080.

 

I'd suggest him to go into BIOS, set all the multipliers to 40x across the board without changing anything else and boot it up. Then see how it handles it.

He needs to record the stock voltage at stock clocks first to see how much power it's eating up. It'll likely be a very thirsty chip as they all pretty much are across the board, which means that it should be enough voltage there to run the 6820HK at higher clocks no problem without messing with the voltage.

I would not recommend using the stock profiles that Dell has as presets. If it's anything like the R3, those settings are complete rubbish

If anyone needs help and you're willing to screen share, I'll help you do some quick tests to see if your unit can live up to the claims, because if it can't run at 4.0 at the least then I'd exchange it for a unit that can. If you get lucky there are chips that can handle 4.1GHz no problem (6820HK). This is totally up to the individual, but to be honest I'd be upset to have paid for something that claims something and for it to not work.

There are preliminary tests and things one should always do when they first get their computer. Seeing if the CPU can handle what it claims is just one out of the many.

....Now I really can't wait to get my hands on one.

 

Gelid Extreme and/or Grizzly Kyronaut.

Here is a thread that could help you with visuals and to get some ideas. Although it's mostly about liquid metal, the info found on there can be helpful.
[Liquid Metal Showdown] Thermal Grizzly Conductonaut vs Cool Laboratory Liquid Ultra / Pro

Here's one on CPU and Voltages:
[6820HK | 6700HQ] CPU Voltages at Idle and Max. (Stock + OC)

Here's one specifically on Under Volting:
[Undervolt Results] 6700HQ △ 6820HK (What's your MPG?)

This should help you out with answering a lot of questions....

 

Dell's Bloatware is similar like the **** one from Clevo. Not useful at all if you want tweaking the performance Entertainment or power saving. What a waste!! I don't use junkware. The only app that *can* be useful in the Bloatware package, is the one you can tuning the colors on keyboard. And I expect Dell also this time... Skipped to develop a app for tuning the fans in th ACC software. People have asked for implementing this in Dell's software. But no one from Alienware listening what the buyers want.

 

I'll do it when I get off work today. On benches the fans are spinning high, kinda sounds like a hairdryer in the next room, playing games it's not really loud. Gpu stays around 65-71 and CPU is the same on gaming.

 

http://imgur.com/u152Nb6
Here is temps and clocks -150mv and 3.6 across

 

43-44watts used on the package power and still hitting up to 90c.. :/ This is either a bad paste job or a underwhelming cooling setup. Take your pick. The voltage is at 1.225 which isn't extreme either.

Package power will directly correspond to cooling capacity / paste job.

Edit: It also seems like the BIOS is adding too much voltage under load.. 1.225v is pretty excessive for 3.6ghz. You should be able to push more on the undervolt.. Try to land around 1.15volt. Or keep pushing the undervolt offset until it crashes then back off by -20mv

And focus on stock clocks for now in my opinion. Just leave your multipliers alone and focus on max undervolt.. When you change multipliers the BIOS will automatically add more voltage to compensate.

 

Thanks) I strongly suggest to open up the laptop and repaste it. 79c and 90c between the coldest and hottest cores is too much.

Also, fire up XTU benchmark 2 times and record temp and score at the end. It stresses cpu quite a bit.

Sent from my SM-G930F using Tapatalk

 

Download Cinebench R11.5 and Wprime benchmark test and try out. Wprime with 1024M settings with 8 threads. Also download and use Throttlestop for tweaking your processor. And rather start to use Hwinfo64 for monitoring your hardware(put up Hwinfo like the image). And use maximum performance in Windows power plan - not balanced or any power saving plan/settings in ACC.

Spoiler: Hwinfo setup

 

i got it to -165 mv, -170 is instant crash.

 

after 2 tries with -165 mv
http://imgur.com/a/aGdnP

 

double post

 

So it's throttling. Recorded speed was 3.4. Try and test at -150 and -130 for the sake of experiment. But at at 3.6ghz with no throttling it should get you to 1200 mark. What ram are you using?

Sent from my SM-G930F using Tapatalk

 

If he use max performance in Win power plan, we should see a nice clean straight graph

 

http://www.3dmark.com/3dm/15420046

this is with -150mv, +200 gpu clock +400 mem, CPU temp max was 78 and GPU was 72
not bad, with a little paste I think the temps will be a lot better.

 

You dont happen to have a full list and with instructions on how to do them, a Youtube tutorial for example . I will receive my maxed configured AW17 in 2 weeks and like you said - i believe preliminary tests are a great idea to see that everything are in order.

 

I'll draft something up and make it available here soon. Sort of like a check list.

 

I agree. It's no surprise that Dell has the worlds worst paste jobs.

99.99% of the time having uneven core temps like that means that it has a terrible paste job. This is common even more so with it being just the bare die...ie...little room for error compared to the IHS on a desktop chip.

A Repaste is mandatory as stock paste deteriorates quickly and leaves little to be desired as time goes on... all paste will seem decent when fresh, but reality will hit shortly depending on how crappy the paste is and with Dell?...I'd rank them just above toothpaste.

@quickie, repaste with Gelid Extreme and/or Grizzly Kyronaut when able. I'll have a tear down of it once I get mine in. Hang in there if able...

 

Single channel memory. That would be a performance hit there, right?

 

@quickie , Could you please run a Fire Strike Ultra/Standard and post the power pull (Watts) with and without OC ? on the GPU (Use HWinfo Plus run these on the dGPU mode not the Optimus mode). I'd like to know what's the GPU TDP and the real world power usage. You can Install Rivatuner and Check the (RTSS) OSD tab in HWInfo settings for custom settings with games/gfx programs to fetch these values and monitor them realtime.

Also you can use MSI-AB for Undervolting the GPU. Since It's a shared HS undervolt the Skylake voltage, It's too much (Sad part is the HK and HQ use too much voltage compared to the 6700K that K series CPU clocks upto 4.6-4.7GHz at 1.4v while these need more than 1.3 at 4.2-4.3 or even 4 that's Intel BGA) and run Cinebench11.5, WPrime 1024 to test stability. Can also run the FireStrike in Loops for the stability. And a little cut off with the 1070 stock voltage might also let you drop temps further, With Maxwell It was not possible to Uv the GPU voltage this is truly remarkable for us to control the GPU voltage with Stock vBIOS give it a try, If it works with Dell vBIOS then you are lucky.

Dell/MSI/<ANY OEM> stock pastes in this world are the WORST you can even imagine. Period. I *STRONGLY* recommend a re-paste. Else you will experience too much dust if fans are running in Auto, HighTemps and degraded UX, It's pretty simple once you get a hold of it. X or 2 lines on the GPU and one single line on the CPU works & get Gelid GC Extreme using it since 2-3 years and works like a charm, More Info here

You can control the fan speed from HWInfo on Alienwares try to max the fans out when Benching and gaming stuff. Also make a note of CFM rating on the Fans while you re-paste and Please kindly let the people know what are values, Finally you can start AIDA64 when running these tests and check that Voltage power graph, If it says ACline only then there's no stupid Hybrid Battery boost gimpstuff.

Thanks and have fun with your machine. Best of Luck.

 

Mmm I have the impression that the 6820HK heats more than the 6700HQ @ stock (even with the factory added frequency boost), do Intel apply more voltage on the 6820HK ?

I think the GPU temps are awesome for Pascal, was used to hitting 89°C in 3DMark and Unigine valley very rapidly with my old (now returned) GL502VS.

Regarding CPU temps they are a bit higher than what I was expecting at stock, was expecting 80-84ish°C.

Anyone knows why my software undervolt with XTU are not always persistent with my dell XPS ? I apply a -75mV but sometimes I boot up the computer, XTUservice process is running in the background as usual, but when I open the xtu app my profile is @-0mv stock configuration. It's not frequent, but I don't know why it happens, any idea ?

 

Thanks, that would be great.

 

The 6820HK does run rich (overly-volted) and has a lot of room to tone it down via Under Volting. However, depending on what lottery ticket you are holding, the 6820HK just may need all that power, but most often times than not you can always get a decent amount of under volt to tame the CPU temps.

So far the GPU temps out of the box seem pretty decent. With a repaste it could hold the crown for having the lowest temp GPU with the new cooling system. Basing off of the first chinese reviews that rolled out, there is definitely room for a lot of improvement from taking what is decent to perhaps having the best cooled GPU in gaming laptops. After all, the new cooling system may fair out to be worthy and I'm feeling optimistic about it...

CPU temps are easy to tame...work on your under volting and plan a repaste here soon if you so desire. Highly recommended.

XTU and TS are like sworn enemies lol. They don't seem to get along all the time. However, the main issue that I know of is that with TS, if you make changes and save it, it'll create a .ini file in the TS folder; which will load every time you launch TS and can cause some bucking of heads with any other tuning software. Therefore, I'd highly recommend that you delete the .ini file if you don't want it interfering with XTU or any other program. (You'll have to go to view -> Show hidden files to see the .ini file.)

Tip: Use XTU to set your multipliers and nothing else. Then use TS to adjust your voltage.

Hope that helps.

 

yes that helps, good thing i took a 6700HQ, It will heat less imo. I don't have TS installed, but might think about installing it, since I'm just planning on UV -100mV on the A15. I want a simple software UV that doesn't require me to set it on each windows boot . XTU works great, but sometimes I discover that my UV hasn't been activated for days --'...

 

Can you do gaming benchmarks? I believe that's important. Games that needed to be benched are:

a) Witcher 3
b) Arkham Knight
c) Crysis 3

 

Why install XTU at all? Is there a reason Throttlestop can't be used for adjustment of multi on newer AW machines? This is like butter on top of bacon

 

Because TS isn't perfect nor is XTU. I've never felt comfortable due to funny business with OCin'g via TS. It's great for adjusting voltage, but as much as I don't care for XTU as well, it seems more confident with adjusting multipliers via XTU...after all it is made by Intel and I think the Skylake is made by Intel? Or is it AMD?

 

TS works perfectly for the vast majority. And pretty easy to use Although XTU is made by Intel... Does not mean that this soft bloatware is better. *Maybe* a bit more fancy designs on the graphs. But in return.., the XTU bloat use more processor cycles. Not desirable solution for overclockers or people who want all power from their hardware.
Is it AMD?

 

I didn't say that it doesn't work. Personal preference.

Yea it's bloat lol..

I think Skylake is AMD FX100000.


::iunlock::

 

You can adjust the voltage curve (on my MSI with 1070) but all it does it force the card to run at lower clock speeds when you start undervolting.. The card just autocorrects itself to the undervolt per clock ratio or something.. Seems fruitless for the most part to try and save on temps.. Unless i'm doing it wrong.. But if I lower each tick by -50mv the card just never tries to turbo any higher than base clock.