*OFFICIAL* Alienware 17 R5 Owner's Lounge

@Papusan after a bit tweaking my results are better now and nearly on the same level as your 8700K Scores (within the 0.5-1% range because of your faster rams)

Ambient temp. 24.5 degree

4.3ghz stock



4.5ghz



4.7ghz



so i think we can say: mobile coffee lake is not slower per design... its just a tweaking thing

 

Or maybe you won the BGA SL!

There aren't many results like yours in the BGA category, but it won't be hard to find similar scores in the LGA category.

I guess we'll find out only when there will be enough BGA laptops that will have similar thermals after a LM repaste + other mods and tweaks

 

After numerous reboots, BIOS to 1.1.6 and 1.2.1, it seems I have a power throttle which keeps my CPU just below 93C, not matter what I do. This power throttle kicks off after 2-3 seconds of a thermal throttle, it is like taking over. This is a variable power throttle, sometimes at 60W, 65W and sometimes at 70W depending on room temps, and if the GPU is also generating heat, down to 50W. It seems the intention is to keep the CPU below 93C. But instead of thermal throttling it is a power throttle.

This is a new installation of Windows 10 Pro, and have no idea what it is that is doing this. On the one hand I know who wants to have their CPU higher than 93C, but on the other hand I would like to understand what it is that is doing this, because in the past the CPU would happily reach 100C and would happily throttle thermally. All Alienware 17 laptops here will also very happily reach 100C and have only the thermal throttles to tame them. My thermal throttles have disappeared and I would like to know why. If anyone has any ideas.

 

Uninstall XTU with profiles, go to Bios, load defaults, Boot and restart twice, reset Bios again. Reinstall xtu.

I had a similar Problem after to many bsod because of unstable undervolt on my old r5. Dunno if you need to do everything twice but this helped me...

 

I still have the white pads. I kept them just in case. According to my eyes, they are 1mm in thickness and not 1.5mm which I used afterwards according to iunlock's guide. This may keep the heatsink away from the CPU hence my issues. In fact, I do not see any 1.5mm pad but 1mm maximum. The 0.1mm white pads (stated as 0.1mm on iunlock's guide) are in fact 1mm on R5. I used a common ruler to measure thickness. I also compared my 1mm gelid pads with the white pads and are the same thickness.The thickness stated on iunlock's guide are from R4. This may be the R4 pads but not the R5 pads? Could be because one of the heatsink's screw is placed differently compared to R4 too. In addition, when I opened my machine, the white pads were on the motherboard and not on the heatsink...

I will reduce 1.5mm pads to 1mm and see what happens.

 

I have loaded BIOS 1.2.1, 1.1.6, back to 1.2.1, dozens of "reset to defaults", restarts and complete shutdown and restarts. I have disabled XTU's service and it has not run in ages, I use Hwinfo to monitor. I have also disabled some other Intel's Power services, just in case.

What I seem to have is a system which keeps the CPU at under 93C come what may. No matter overclock, or underclock, frequency, GPU also running: under any conditions, as soon as the CPU hits one thermal throttle then this system takes over, imposes an appropriate power ceiling and limit, and stops the CPU from exceeding 93C with no more thermal throttles for hours.

Actually this is not a bad system, but I wonder what it is. I will call Dell.

 

uninstalled XTU? completly with all profiles and settings?

if yes then sry ^^ was just an idea and it helped me

 

No, just disabled its service and made sure it is not running. I will now uninstall it. Hang on.

 

Dont forget to reset Bios after uninstall its worth trying

 

Nothing has helped. Running more tests it seems the cut-off temperature is 94C. As soon as the CPU hits 94C the power limit decreases power to keep the CPU below 94C.

Boot machine, run hwinfo, run P95, power limits active, PL1, PL2, PL3 and PL4 (or a combination of). It appears to be at 67W, CPU temp 92C.
Now start WoW in the background, GPU now goes to 150W-180W, additional heat in the system. Power limit now magically drops to 45.9W, CPU temp 90C and rising ever so slowly. Power limit stuck at 45.9W. After many minutes CPU reaches 94C. Power limit magically drops to 42.9W and CPU to 90C!

I am now categoric that there exists an algorithmic power throttle that keeps the CPU below 94C by adjusting the power limits on the fly. It uses PL1, PL2, PL3 and PL4 according to HWinfo.

Not sure if it is the BIOS, I have tried versions 1.1.6 and 1.2.1 . Intel XTU completely gone. Not undervolted anymore. BIOS set to its defaults 100 times.

 

strange... but it was worth a try. I dont know if this is normal behavior... never reached this temps (and never want to) but i remember that you had hwinfo Screens with 99 degree, so something has changed

 

Now looking at "Intel(R) Dynamic Platform and Thermal Framework service" which I have stopped but God knows what it may have done.

 

It might be worth to try an EC reset or the full reset package EC/CMOS!

 

How do I do that? Without unplugging the CMOS battery? Wouldn't the BIOS flash not do this too? Secret BIOS setting maybe?

 

I managed to sort out my instant high temps after re-pasting. Replaced 1.5mm white pads (which are in fact 1.0mm on R5) with 1.0mm gelid thermal pads. As I guessed this was the issue. Even replaced the ones stated as 1.5mm with 1.0mm. The ones stated as 1.5mm on iunlocks guide (which is for R4) are at the sides of the heatsink seem to be 1.0mm on R5. This means that placing thicker thermal pads (we are talking about 0.5mm difference) on the sides causes CPU/Heatsink not having contact at all. This machine is painful but worth trying.


Unfortunately, I will have to re-open to use LM again (removed it because I did not want to play around with it while checking the thermal pads and heatsink-cpu contact. I used phobya nanogrease extreme which has good temps generally (lower than 50C during idle) but I am getting some rare spikes (90C once without gaming) knowing what is going to happen while gaming (farcry 5).

Damn! this is by far the most awful to dissasembly machine I ever had.

 

I have started Windows 10 in Command Prompt mode, run hwinfo and Prime95 and the laptop refuses to go over 93C.

I have re-installed Windows 10 from USB stick. I have not installed XTU or any other OC software. I am still having same effect: an invisible, variable power throttle that keeps the CPU up to 93C and lowers the power if it ever touches 94C. Currently gaming CPU is being power throttled to 38W. No thermal throttles at all. It then slowly increases power to 40W, temp hits 94C and power drops again...

Someone must know what kind of throttle this is ?

 

Try just the EC reset (for CMOS, I don't know where's the jumper for this AW model or if it has one), this would be the general procedure:

-turn off,

-unplug all peripherals,

-unplug the ac adapter,

-unplug the main battery (it could work even if you skip this step),

-hold the power button for 1 min,

-plug in the ac adapter and turn on the laptop.

 

Captn.ko what undevolt level do you use for CPU now? Do you use throttlestop or intel XTU? Whats your maximum temps?

And what undervolt for GPU? maximum temps?

 

I think I will do the repasting too, do you have a photo or anything showing what pads you have used where?

 

Just a short test . Tests ran until temps stayed stable and start hovering around this value. So i think temps may rise over time just a bit. Can make a longer test later. Dont have much time now. No graphs. Believe me or not. As every CPU has a different VID you cant use my offset. Maybe you could run a better, maybe you crash instand. You see my voltage in the hwinfo window (current)

CPU



GPU

 

captn.ko thanks for the screenshots but can you post a screenshot from FIVR inside throttlestop? This is only to have an indication where to start from. I am miles away form reaching your temps with phobya paste. I just want to see where to start from. I will have to re-open to re-use LM (had to remove LM because I was sorting the bad CPU/heatsink issue first). But while using LM I saw temps of 36 for the GPU in idle! CPU was cooking of course because of bad contact with heatsink.

I am now using undervolt because temps can go over 90 without it using phobya paste (I wont bother trying any game or stress test before repasting with LM because temps are giving spikes even in idle, even though maximum temps are in the 80s).

 

Just use 1.0mm pads for the ones stated as 1.5mm on iunlocks guide for GTX1080. I recommend not to touch the white pads as I did but if you do use 1.0mm pads for all of them. I worked by comparing the thickness of the original pads with my 1.0mm gelid pads. There was none at 1.5mm. You can double check. Compare the thickness of your pads with the original pads.

By the way, prepare to spend much time. This machine is a nightmare to dissasembly and re-assembly

 

start with -100mw

 

I have it at -125mv which is the maximum. seems stable. does this mean it has a good VID?

 

Done that held power button down for 2 mins and 15 secs - no idea if it cleared anything. Symptoms persist however.

 

I do not have the guide at hand. Do I take it then all pads are 1.0mm everywhere? And if you need something thicker you can double them up?

 

You guys need to check each pad. This will take hours.

 

I now remember that I had noticed something similar on both Alienware 17 laptops I have had recently.

Test: boot laptop and on BIOS choose "standard" (or whatever the heading is for "not OC"). In Windows then observe PL1=45W and PL2=90W - hwinfo will show you this. Now set it to OC in the BIOS and then PL1=PL2=110W. So far this is fine. However. Go back to the BIOS and set it back to "standard", ie PL1=45W and PL2=90W. Go to XTU / TS and try to set PL1/PL2 to anything higher, eg try to set PL1 to 60W or to 90W. It does not work! XTU cannot override BIOS settings. Even if XTU displays for example "80W" you will see you are still stuck at 45W. I do not know how PL1/PL2 are set - what is the interface - who controls it, but it seems to me that the BIOS PL1/PL2 cannot be overridden *higher* by Windows applications, but it can be overridden lower, so if you boot at 45W/90W you can se it to 20W/30W if you wish and that "catches". But if you try to set it to "60W/95W" it does not even though it displays as if it does.

Based on that observation, I think there is more at play that meets the eye and I need to dig deeper to understand how this works.

 

Oh yes I have seen the video on youtube it looks unreal that you must remove the motherboard!

 

You have to check each pad thickness but generally used 0.5mm for the pads stated as 0.5mm. 1.0mm for all the pads stated as 1.5mm and 1.0mm for the white pads stated as 0.1mm (this must be mistyped anyways). Again the thickness I used was by comparing with thickness of original dell pads.

 

Can someone who has the latest bios for the R5 with i9 run a stress test on the cpu, what speed does it hold?

I feel like Dell may have crippled some speed to keep the cpu cooler...

 

This is what I am recording just now.

BIOS 1.2.1

With Prime95 small FFTs (maximum heat stress). No GPU.

With 30C in the room (yes 30C in London), GPU off, no undervolt, 2.9GHz, ~60W, ~93C (I cannot get it to run hotter as you may have read).

With -140mV undervolt: about 3.33GHz, 60W, again at ~93C.

Around 1200cb at 3.78-3.83 GHz 59W-64W and again 92C ( -140mV)
(I have got up to 1290cb at various runs with same settings, but I have also got less)

However "speed" is a bad way to describe the heatsink/chassis performance because at the same frequency you could be performing integer arithmetic which spends less energy than floating point which spends the most. So it depends on the code you are running at the frequency stated. Intel define their "TDP" as " a blend of representative instructions at the stated frequency" (2.9GHz for the i9)

 

Guess those countless hours of dealing with support and incompetent techs wasn't for nothing…

 

Ohhhh you're more screwed now than before buddy... I did this same swap and the new one is absolute trash compared to the r4.. more heat and more hopeless to mod and resolve.

 

LOL I'm aware. Crossing my fingers and if not a nice sale considering I payed 1.5 for my original machine.

 

What limit reasons does throttlestop show?

Run the test again with the limit reasons window opened and clear the previously registered limits with those 3 buttons (core, ring, gpu).

 

Here is my new findings for whoever cares

We are at idle and nothing loaded except P95 and one of Hwifo, XTU or TS.

We start P95 Small TTFs with >= 6 threads. Thermal throttle, fans start, a few seconds, power throttle on, thermal throttle off. Power throttle stays for as long as we load the CPU. Power throttle is PL1, PL2, PL3 and PL4 and any combination. Power throttle is varying from say 40W to 70W and the ONLY constant is the CPU temp <= 93C.

Where have my thermal throttles gone ? It is obvious the system is keeping temps just below 93C by applying power throttles. I did not know that.

Today another discovery. We can "trick" the system to screw up. By varying the power this system affects all cores. What if only 1 or 2 cores are used? How can this system now throttle just the right cores just when they approach 94C ? The answer is it cannot !!!

Proof: we can start P95 with only 3 threads. See now how we get thermal throttles all over the place and the system above cannot cope.

The questions are:
1) what system is this that implements a variable power throttle to keep the CPU <= 93 C ? Where did the 93C come from? Is this in the BIOS to pacify customers? Is it in Windows?
2) why is this system in existence rather than a variable temperature throttle. Why do we have a system that tries to tame CPU temps by varying the power rather than directly by an adjustable temp. throttle? It does not look like this is an Intel system design, more like a Dell/BIOS hack, if I were a betting man.

Two screenshots; the first is P95 12 threads







And this one is P95 only 3 threads





As you can see when all the cores are busy the "varying power throttle" works and there are no thermal events. As soon as we "burn" just a 2-3 cores, the "varying power throttle" cannot cope and we now at last get thermals.

 

Forgot to add to the above, PL1 = PL2 = 110W and we are nowhere near those values. When it says "PL1" and "PL2" we are at 40W, 50W or thereabouts.

 

Be carful with bios updates. And god damn if you are forced to have to update firmware due very needed security fixes...
Some insights into How Dell cripple performance
Be forced to use old unsecure firmware due what’s can come ain’t nice!!

 

There is a possibility that there is a throttling mechanism that is triggered at that temperature, but it's more likely that you have some sort of a bug (a primary indication of that would be the fact that you reach the PL1 and PL2 limits, although you say that they are set at 110w).

Your system is behaving like it's stock (without raised power and current limits, hence you have the PL1 and PL2 throttling and edp other limits, all at a much lower package power than you say your limits are set to, don't know at what value your current limit is set).

I presume that the thermal throttle limit for the aw17r5 is around 100C (maybe 97C), so when you say that you triggered the thermal throttle limit, do you reach that temp (about 97C)? Or you somehow trigger the thermal limit at 93-94C?

Another odd thing is that you reach thermal throttling with 3 cores/ 6 threads.
It might be possible since those cores should be running at a higher frequency at the same wattage, but I can't know what happened without having the rest of the relevant data.

How does this theory sound: you started the test, you reached the thermal limit because your power limitation is at a higher level while being in the time window for the PL2 limit, after that your PL1 limit is triggered (that limits you at a lower value, usually its 45w, but yeah it should be 110w if you set it to that value, but in this theory, we're presuming that the there is a bug and you limits are interpreted as defaults) and while you are being PL1 limited, it is possible that the wattage at that limit coincides with 94C.

I can only speculate until I have more data to work with. It would be great if we could read the values (VID, clocks, temps, power draw, the power limit values and when they are triggerd) in a graph so we can know what happened during every moment of the test.

 

Need to fishing everywhere. Just a start... From my older posts... This options if not Dell has changed how their new OC App works for the new Coffee BGA models.

 

Start fresh. Shock the CPU with P95 it hits thermal limits immediately at 98C-103C. Fans are still off. It takes a few seconds for fans to catch up, thermal limits still on. After a total of about 7-8 seconds from initial shock, the power throttles take over and replace the thermal throttles and the temps are held at 93C or just below, eg 88C-93C. We then modify parameters, eg we start the GPU so it also generates additional heat or start a large fan in the room at the laptop, and the power throttle changes between 40W, 50W, 60W, 70W - any of these values depending, so that the CPU stays just below 94C. How do I know the power level of the power throttle? Because it is a straight line. If the temps go down to about 88C the power increases a notch and if we hit 94C, as soon as, the power drops a notch. Clearly a dynamic power throttle, as explained by papuscan above. Called Dell Dynamic Power something or other.

Why? Each core has its own temperature sensor and you can hit cores individually. You can easily do it from Task Manager even or programmatically, it is 1/2 API calls SetThreadAffinity or some such

Of course, Package Power is the total including all the cores, GPU, and whatever else - and you cannot tell which core uses what power unless you make an educated guess from the multiplier? Because if you could read individual power consumption per core then it would be displayed in Hwinfo. I am guessing here.

My power limits were not 45W/90W because I booted in performance mode and it sets them to 110W/110W. It seems that the BIOS has the final word on what the limits and the algorithms are and neither XTU or TS can do anything about it. Explanation in

 

doesnt sound wrong to me... why should it work different? I Would expect my cpu to do exactly this what you describe

 

You missed my point here ^^^ since I stated:

There are some things that don't properly fit.

Firstly, you said that this is a new behavior l and before that you didn't have this kind of limitations. What could have changed to experience this new behavior (or you had it all along from the start and you only discovered it recently?)?

TS can modify the power limits in the MSR, XTU has access to both MSR and MMIO, but if this throttling behavior is controlled by the EC/PECI, then you should've had this behavior from the start, since you can't do anything to overrule the EC settings.

The fact that the PL1 and PL2 limits are triggered, might indicate that the usual power limits are triggered. If it would be another throttling mechanism that is separately controlled, then it might not show in TS reasons (this doesn't exclude the possibility that you could have trigger the usual power limits and also be separately controlled by another throttling mechanism).

There is still the strange fact that you say that the PL limits are triggered at 50w and a few watts above, but your PL1 and PL2 limits are set at 110w! TS indicated that you triggered those PL limits that you mentioned that are set at 110w.

The dynamic throttling shouldn't be detectable with TS.

Regarding the 3 core usage, I know that you could set the affinity, but that wasn't the point, since the only logical explanation for having higher temps with 6 threads vs 12 threads, would be that with 12 threads the power limitation kicked in fast enough, before the temperatures had enough time to rise to the throttling limit.

When stress testing you need to force the fans to max speed from the start (don't use auto fans).

I also don't know how much does the alienware command center control, because it is possible that it overrides the power settings and could also be responsible for triggering some throttling mechanisms, similar to how CCC (clevo control center throttles the current BGA models).

 

I have seen articles saying the R5's speakers are improved but no mention of how exactly anywhere. Anyone have more info on how the speakers have changed,if at all?

 

Speakers haven’t improved over the r4
They are the same spec. 1 sub and 2 small front speakers in the same location as the r4.

However the software is limited now as you can’t set a custom preset and adjust the bass. This means you get less depth and bass on the r5 compared to the r4.

 

Really? That's horrible. I live by my custom preset in Sound Center, that's actually genuinely disappointing. I was referring to this anyway: "The design now includes a couple internal smart amplifiers that monitor audio waveforms (a graph that displays amplitude or level changes over time), and the system uses that to better regulate the speaker thermals for better sound."

https://www.laptopmag.com/reviews/laptops/alienware-17-r5
(Speaker thermals? Wtf?)

 

They have come back with bad English and bad technical explanation alluding to and implying and I am inferring that the change happened on the last two BIOSes, ie 1.1.6 and 1.2.1. They suggested that I go back to an even older version but that I should not because the new versions are great.

Has anyone got the < 1.1.6 BIOS so I can try?

I have installed two new BIOS, 1.1.6 and 1.2.1 and cannot remember what was there before. I have re-installed Windows twice. So a lot has changed, yes. I used to remember that the CPU would very

happily hit the thermal ceiling (100C ++) on the smallest provocation and would continue doing so until I reduced the clocks and/or the power - which is incidentally what this new behaviour does for me.
​

OK if you could explain this to me I would be very grateful. I do not even know what the EC is.

Throttlestop and XTU and hwinfo mention PL1, PL2, PL3, and PL4. All being breached together and separately. Looking at the actual TDP at any given time it is like say 40W or 55W or 65W, and then looking at PL1 it is say, 80W, ie nowhere near. Where did this "40W" or "55W" or "65W" come from? It is obvious it is dynamically being set to XYZ watts to keep the CPU from exceeding 93C. I love it, it is a great system, it fails of course when not all cores are being used at once (read below). If we only knew how to program it, I might want to set mine to 85C instead of 93C. I think 93C is too much.

No, I did not say that. I said that if you only use 2-3 threads then the unfortunate cores to be handling those few threads will jump to the 100C, whereas other cores are at much lower temps. If you use 8, 10, or 12 threads, now you are employing all the cores and the temperatures rise evenly. Why does this matter? Because this new "let's keep the CPU below 94C" system, modifies the clocks based on TDP which is core agnostic (it seems), so it fails, especially when threads jump cores. Then the cores cannot be contained and the temps escape the 94C barrier and hit 100C and then the thermal throttles do come in. So if you use only a few threads you will continuous thermal and power throttles, but if you use all the cores then you see only power throttles.

Not installed in the last two installations of Windows 10 Pro.

 

There is no sound centre anymore, the sound presets are within the new Alienware command software. But like I say there’s no custom reset, just aprox 5 presets. In an earlier post on this thread, I took an image of what it’s now like.

 

The essential thing regarding what I mentioned previously about the PL, is that they are stored in different places. TS can modify them in one place (and they should be cleared after a restart), XTU has access to some of the limits in BIOS, so they stick after restarts.

Those that are controlled by the EC (embedded controller) can overrule all the previous limits that were set (so it doesn't matter what limits you set in other places if the EC has the final word).

The limits that should be reported are: PL1, PL2 and the EDP other (they shouldn't have been triggered at lower power draw values, unless they were ignored and controlled elsewhere).

That is also what I'm trying to figure out and now it appears that there are even more variables: you mentioned that you set the both the PL1 and PL2 to 110w and now I see this:

So you set the limits to 110w, but the monitoring software is reporting the limit at 80w and you say that it is triggered at values as low as 40w.

The thing is that, this kind of dynamic throttling shouldn't be reported (different ways of control in different places, but it could be possible to have an algorithm that uses a different triggering behavior for the same limits) by the monitoring software, so this is why I mentioned the possibility that there is a bug or something that doesn't add up.

Can you post some pics (using this: https://postimages.org or the app https://postimages.org/app , with imgur they are not easily readable) that will show all the relevant info (clocks, core usage, core temps, cpu power draw and the temperature and power limits when they are being triggered) during the stress test?