How to explain Thermal Throttling to senior representatives of Dell(India)?

First up, I will give the configuration of my Alienware 17 R3:
i7 6700HQ
GTX 970M
16GB RAM
256GB SSD

The whole issue started around January when core temps were going over 90C and I went to Service center for repasting(back then, I had no idea aboubt good quality paste, way of fixing heatsink, core temperature difference: Core 1 was hitting 90C while core 3 and 4 were barely touch 75C, even when running Prime95)

When I got my laptop back after repasting, to my horror, core 1 and 2 were spiking to 100C almos INSTANTLY when I started Prime95. Let's forget Prime95, it was the same with SupportAssist Stress test too. Even Dell's low stress ePSA reported 97C, but saying all tests passed!
At this stage, maximum core speed were around 2.6GHz, while previously I was getting nearly 3GHz even when running Prime95. I felt I dug my own grave

Then, I imported thermal grizzly kryonaut and asked the service center guy(not the same as above, a different one) to paste using this.
Initially, max temps using Prime95 was 85. I was overjoyed. Things are back to normal again.

BUT!! It was short lived. Very soon, the entire laptop started heating like an oven. Mind you, it was still february and pretty cool weather in India. Unberably hot to touch on the surface. Checked my CPU temps and to my horror, All cores were again going over 85C with Core 1 and 2 spiking to 97C.
I called up the Serivce center guy and asked him how he repasted it and if he had cleaned it properly before applying the new paste to which he said "No sir. I applied it over whatever was there. The paste is original from dell. It has to be there, I've applied Thermal Grizzly Kryonaut over the original paste". At this stage, my faith in Dell took a nosedive, like how can anyone be so foolish to do such a thing

This was followed by a series of replacements:

1) Heatsink + Fan replacement - Not fixed. Same heating, same thermal throttling

2) Second time heatsink and fan replacement - Not fixed.

3) Motherboard replacement - This time, I had no choice but to use Arctic MX-4, after clean wiping whatever Junk Dell had provided me with. Temps show 76C after this? WOW! Finally, it's fixed!

BUT, again, like my fate suggests, it was short lived. The problems returned back within 2 days. Massive Surface heating. Massive CPU heating to 95C with thermal throttling.
I called up customer care and asked them to elevate the issue to L3 level. The L3 guy, don't want to give his name here, said " It is working according to design specification" and closed the issue entirely. I am pretty sure this is a very famous term in Dell. Any problem that consumer reports is always within their " design specification". Horrifying, this was absolutely horrifying

Then, I was approached by Advanced Resolution Expert who said he wants to pick up the laptop and give it to R&D team for further testing. Well, I did as necessary. Later I realised that the R&D team consisted of the same L3 guy who closed my issue in the first place! What horror. My laptop was returned back to me after another repasting, again with exact same words " It is working according to design specification. Temperatures can go upto 100C. It is OKAY". Still, with temperature spiking to 99C and thermal throttling to 2.2GHz on Prime95. Which rubbish junk has thermal throttling as their "design", rather a "design flaw": Throttling to as low as 2.2GHz whence I should have got 3.1GHz on 6700HQ which is still "working fine" for them!

This wasn't the end of horror. Not just throttling to less than 2.7GHz all the time, there were occasional drop to 798MHz, while gaming, stress testing, ffmpeg, Visual Studio Compilation. You name it. Anything that is CPU intensive, there are ocassional clock speed taking a nosedive to 798MHz. Here are few Videos demonstrating it:







All Videos have timelapse in the description showing the instance where FPS drops very low due to Clock speed dropping to 798MHz. Insane! It happens at crucial times in CS:GO

Yesterday, still fighting after 3 months, I got a call to bring over my laptop and "Supervisor" of Advanced Resolution Team wanted to speak with me. Well, I went to Dell Internation Services at Domlur, bangalore. The conversation with him was one of the most non sensical discussio I've ever come across in my lifetime. Here are a few arguments and counter arguments:

1)
My argument - I am running Aida64 FPU test. Clock speed is 2.4GHz. I expect, at max load, the frequency to be 3.1GHz. This is case of thermal throttling.
His argument - OH! IT IS WORKING FINE! Your background processes is consuming 0.7GHz. So, 2.4GHz + 0.7GHz is 3.1GHz, isn't it?
***ARE YOU KIDDING ME??? How insane is anyone to "add" clock speed. Nevertheless say that some " clockspeed is consumed" by background processes!

2)
My argument - I run ffmpeg. CPU usage is about 90%. I am getting 2.5GHz. Don't tell me again that there are background processes "silently eating up clock speed"!
His argument - Oh! Again, 90% of 3.1 GHz is 2.79 GHz. You are getting 2.5GHz. You are good
***WHAT IN THE UNIVERSE ARE YOU TALKING ABOUT??? I show him Prime95 consuming 100% CPU and clock speed at 2.1GHz. Sir, 100% of 3.1GHz is 3.1GHz. Then why is, your beautiful Aliencrap, giving me 2.1GHz, oh good sire?. He cannot answer my question

3)
My argument - 100C is too hot. I believe 85C should be the absolute maximum a CPU should reach before the PC shuts down and this, should then be considered as a thermal issue
His argument - * Starts talking about 1000CC Ferrari car running at high temperature without problems *
My argument - You are comparing Combustion engine with Integrated circuit. I don't see any fuel burning inside my laptop and powering it to produce that much heat. It is due to such high temperatures the clock speed is throttling down to below 2.5GHz
His argument - Isn't that a good thing? Your CPU is being prevented from burning out
My argument - But sir, where is the 3.1GHz as advertized? Isn't that what I am supposed to be getting
His argument - You will have to overclock to get that speed. You are running stock speed. 3.1GHz is only the max speed. You will not always get 3.1GHz.
***At this point, I loose my cool at him and stop arguing anymore and nod my head at absolutely horrifying stupidity

4)
My argument - While playing CS:GO, there is massive FPS drop. * shows video where CPU clockspeed takes a nosedive to 798MHz. But, that happens to be during the time I am almost standing idle*
His argument - You are idle. Nothing is changing on the screen. You CPU went into idle state! That is why clock speeds came down!
My argumet - What? No. Even if I am perfectly idle, the frames are rendered irrespectively. There is a problem with the CPU
He - NO! IT IS WORKING ACCORDING TO DESIGN
My argument - *shows a part of video where I am trying to shoot someone but aggravatingly missing due to FPS drop to 40FPS, clock speed is 799MHz* What do you say for this sir?
His argument - Fans would have stopped working. When you suddenly start aiming, CPU usage spikes up and clock speed goes down to protect burn out!!! And your clock speed is back to normal after few seconds(Still throttled down at 2.8GHz, Meh, who cares anymore)
*** I am stunned at his dumbness.

Is this the kind of Engineers Dell is housing?? I am stunned at how dumb the above arguments sound! "Adding clock speed", "Consuming clock speed", "Percentage CPU usage * 3.1GHz"!!!! Dell, what is wrong with you? Why are you hiring such uneducated idiots? How in the 21st century of rapid technological growth is such a person at such a top position?

Well, atleast, he finally agreed to take in the system and 'have it tested at chip level'. Sounds Bogus to me. Still doesn't stop saying "Your system is absolutely fine. You cannot buy a 1000CC vehicle and always expect it to perform at 1000CC ability". Cannot believe he is comparing an Internal Combustion Engine to a microprocessor! he still holds stron to his argument that "it is working according to design". He doesn't take into consideration that thermal throttling is a flaw!!

It is impossible for me to explain that there is some insane issue with their laptop and I am not getting 3.1GHz, as advertised, under 100% load and below 45W TDP. He is either trying to fool me or doesn't know squat about Laptop design

I have Warranty till February 2018
Ambient temps are well below 35C(that is their specification in the user manual) at the time of running all these tests.

How can I make Dell understand that, at 100% load, below 45W TDP, I must get 3.1GHz?
How can I make them understand that Games render frames continuously, even if I am completely idle?
How can I make them understand their Design is extremely flawed and it needs to be addressed?

I just want a working laptop which gives me 3.1GHz under 100% load(Lets forget Prim95, it is too much for any laptop. Let me get that using ffmpeg). Dell just doesn't seem to understand such a simple statement and brings internal combustion engine into comparison with a microprocessor.

If any Dell representative could go through this post and address the concern here, I will be happy!

 

I believe you're out of luck. Dell redefined overheating to 3 cores hitting 110C or higher on the cpu. @D2 Ultima can chip in, though I think that's right (Do you have the source on that, I read it once in a ss but can't find it again). If not, 100% they say 100C is fine which is ridiculous. I've seen and you can find many cases of Dell techs saying that it's all good to have cpu temps like that.

This was because their Skylake generation alienwares had really poor heatsink contact so a ton of them were getting rma'd. So instead of fixing the issue they redefined what overheating means.

 

Yeah you're out of luck, though not holding base clocks is cause for warranty repair... But you're basically out of luck

Sent from my OnePlus 1 using a coconut

 

Not just overheating. Above 90C, it is thermal throttling, way down to 2GHz when pressed with Prime95 for more than a minute, and occasional drop to 798MHz. Is this also in their " design specification"?

Is my statement "I should get 3.1GHz at 100% load, else there is a fault in the product", valid cause for RMA?

 

I mean that should be cause for a warranty repair because it's not performing correctly.

 

Dell has an 800MHz throttle drop bug that's been plaguing their skylake models. Since Dell and Alienware rebrand Compal, makes sense that your AW has that bug.

Thermal throttling at 90c shouldn't happen, though.

2GHz shouldn't happen either, ever.

But no, your statement "you should get 3.1GHz at 100% load" is not valid. You need to throttle under base clocks (which you do), because that's why mobile CPUs have such high turbo multipliers and such low base clock multipliers, so manufacturers can make crap cooling systems and say "well boost isn't guaranteed", despite the fact that your CPU will throw throttle flags which can be read by software if it's EVER under max turbo while under sufficient load.

But you're dealing with Dell, and clearly some idiots. I didn't even need to read your whole post... if they won't replace it, you're simply screwed.

 

It doesn't throttle at 90C. But it easily hits 99C, almost instantly, when I start ffmpeg and throttle flags are set in HwINFO and clock speeds are 2.6GHz to begin with and gradually drop to 2.5-2.4GHz.

2GHz happens when running Prime95 for over a minute. I've never run it more than 1min fearing an explosion of the laptop. Wouldn't be surprised if it goes below 1.5GHz due to extreme heat

Oh, then let me get the question clarified:
Speaking by the terms of warranty, if the CPU usage is 100%, what clock speed should I see for my i7 6700HQ(base clock speed of 2.6GHz and 4 core turbo boost of 3.1GHz), below which the product is considered defective under warranty and should be RMA'd?

If they won't replace it, under warranty, when it is thermal throttling and having a clock speed drop bug, they are welcoming a law suit. Simple

 

Under base clocks, so 2.4GHz or less.

Sent from my OnePlus 1 using a coconut

 

So AW doesn't even guarantee Turbo boost? This is antagonizing

So if my clocks are above 2.6GHz though thermal thorttling is indicated by the processor, it's all well and good?! I'm baffled!

Yet p95 goes to 2GHz. That at least I believe are under warranty terms to be fixed.

 

no company (except probably HID and Obsidian) guarantees turbo boost. It's a farce, that turbo isn't guaranteed performance. Why do you think nVidia made turbo boost so bloody high on their newest gen of hot cards too? You know how much my card boosts? The base clock is 1557MHz, it boosts all the way up till 1898MHz at stock. That's a solid 350MHz, or nearly 25% of the performance of the card.

Yes, if your clocks are AT OR above 2.6GHz and thermal throttling without 3 or more cores being 100c+, then you're "within acceptable regulations set by intel" and you're fine.

Welcome to laptops, where the consumer gets shafted so hard it's a wonder we can walk comfortably, but everybody and their mother who runs a website or talks about it is furiously trying to prove that it's fine. You know, just like windows 10 being "fine".

 

Oh, well, Aida64 shows 2.5Ghz, Prime95 2.1-2.0GHz(Which are still fine for the tech) plus CPU clock drop bug when doing any CPU intensive workload. Temps are like CPU 0 and 1 are at 99C, CPU 2 at 95C and CPU at 3 at 92-93C in HwINFO.

Will testing by these software provide enough proof of defect ? Combined both throttle below 2.6GHz and the bug, is the laptop RMA eligible?

Out of ideas with this malfunction junk

 

@sagarbhathwar: Like I said earlier, open it yourself and get that high uneven temp fixed or else look for another laptop that suits your needs.

 

you could fix that easy ... changing the thermal paste for somthing like IC diamond 7 or thermal grizzly kryonaut. also you can change the fan profile using HWinfo64 and you will have room for overclocking

 

@judal57: Only you convince @sagarbhathwar to attempt a repaste. Because you helped in fixing my heat issue too.

 

The problem is, what I also said in the other topic. If yo are operating in high temperature places with higher than normal humidity you are operating out of spec.

You can only optimize it by using the best paste job possible.

 

OP, here's a little suggestion, since you're out of luck anyway.

Why don't you try to force the undervolting of the CPU and GPU by using a lower wattage PSU? Say, if you're running a 180 or 200+W PSU, move down to around 130 or 150. I can almost guarantee you the CPU will run at the same clocks, only with a lower max TDP, hopefully resulting in much lower temps and power consumption. Users with the AW13 did the same trick (90W or 130W OEM PSU vs 180W), and their temps MASSIVELY went down from mid 80's to mid 60's. GPU scored 10% less at most benchmarks (GTX 1060) and the CPU reported almost exactly the same scores with much lower temps.

Sent from my SM-N900 using Tapatalk

 

bad idea !!!! since Dell use hybrid BIOS. that will discharge the battery at a high rate and can damage the cells
and why to force ??? he can easy undervolt the cpu using intel xtu or throllstop :S a kid can do that, is just soooo easy

 

Tell those "kids" to undervolt the GPU. Good luck!

 

why the hek do you want to underpower a gpu ? that is totally insane ..... in fact you can undervolt a gpu by using nvidia inspector

 

You can't, Voltage offset is totally locked in the bios.

If using a less powerful PSU means my temperatures will be way lower (better durability, who cares about the battery cells if you own an AW? ), I'll use it no matter what the consequences are. You just can't refuse running the CPU at mostly the same performance with unreal temps, and again, knowing an undervolt made by "kids" won't even get you close to mid 60's.

Sent from my SM-N900 using Tapatalk

 

@SkidrowSKT i am using a custom VBIOS so i CAN!!!! if you dont have the skills of a kid to undervolt is not my problem. you should be able

 

Who the HELL even talks about CPU undervolting here? You brought it up, the suggestion I've made has NOTHING to do with an undervolting technique - which I entirely know, so please quit throwing groundless assumptions - using a lower wattage PSU proves to be extremely temp-efficient, even CPU and custom vBios undervolting won't even get close to.

Sent from my SM-N900 using Tapatalk

 

It works because it is throttling your hardware, and it might wear out your battery sooner than expected. You cant expect the same performance without tweaking the usage with a lower PSU. That would go against all laws of physics .

 

But here is the thing: I'm giving an example of the AW13, which mainly uses the 180W PSU. Multiple users switched to the 130W PSU from last year (Which was ideal for a dual core i7 + GTX 965M seen in the R2). The GTX 1060 dropped by 10% in performance, while retaining sublime temps of under 70 degrees, even under extreme load. CPU didn't throttle AT ALL, showing identical benchmarks results (130, 150 and 180W PSU were all put to test by a fellow owner, and the CPU benchmarks were within 1-2%, which is quite normal if you run multiple benchmarks), what complately changed, though, are the temps. I'm not sure how they explained it, but the temps went down from mid 70s to mid 50s, under load with the same setup!! (XTU Undervolt and repaste).
The hardware only starts to throttle when using the much weaker 90W PSU from Inspiron machines.

Sent from my SM-N900 using Tapatalk

 

Just now remotely tested a friend's Alienware, exact same model.
Ran Prime95

Without undervolting:
Core 0 and core 1 - Max 95C
Core 2 and 3 - Max 82C
Throttling to 2.9GHz and stays there. Constantly.

Undervolting by 140mV:
Core 0 79C, Core 1: 85C. Core 2 and 3: 72C!! With ambient temperature of about 30C!
No throttling whatsoever. Constant at 3090MHz, with p95. That is as good as a laptop gets!
One difference though - Bios 1.2.3, I run 1.3.10

Ran both for 2 minutes. Didn't want to push much further

I dug my own grave when I gave my laptop for repasting! I should have simply undervolted! Max temp ever recorded on my machine before dell opened was 91C on core 0 with p95 maybe my machine was better than this? One can never tell after it has been handled by monkeys

I feel like a fool now

Before anyone suggests: I've tried undervolting my machine with no success. Throttling to 2.8GHz with Aida64 and 2.6GHz with p95! If it's still not fixed, I'll repaste it myself and live with it

 

When opening the laptop, mind the following things,

-The bottom door screws only need to be loosened but not fully unscrewed, if you do, you ruin the threads. They have a locking ring underneath.

-When you remove the screen lid from the laptop base, first screw out the back screens next to the graphic amp port, hold the screen lid while unscrewing the 4 screws on the inside, after you removed these 4 screws, wiggle a little bit while pulling the screenlid up and it will come of easily without the risk of breaking plastic clips.

-Make sure the SD dummy card is not inserted while pulling out the motherboard,

 

This is why Dell says the stock paste shouldn't be changed. Although it is cr*p compared to enthusiast pastes, It is supposed to favor durability and last at least a year before needing a repaste. Those A/M pastes can show very good results, but you will be stuck with having to do a repaste every month/3 months/6 months depending on the paste used...
And like I said in my previous comment, I encourage you to just try a lower wattage PSU, and attempt the same benchmarks. If you have a way to get ahold of one of course!

 

In my case dell stock paste was good for 4 months until it began to overheat.

 

Yep, it's very normal! It also depends on the amount of stress tests done on the CPU and GPU. Have you done so many benchmarks upon receiving the laptop?

 

Yes I participated in few rookie OC competitions.

 

Aha! I totally forgot you can OC the CPU and GPU. Don't you think 4 months is very generous for all those stress tests you've done? Running stock numbers will definitely give more breath to the paste.

 

Mine CPU is 6700HQ completely locked, i can OC gpu thanks to @Prema Bios and @judal57 for setting optimal OC values.

 

What? I'm shocked

 

I knew you would say that

Check the AW13 thread for more info and actual benchmarks, I was shocked as well. It really works!

 

Oh'well. Maybe the next will be using the psu only for charging the battery. Thereafter unplug the psu and start use the machine on battery Back again And so go the days.

 

The human testicles can sustain a TJmax of 45c, but do yo really want it that hot?

 

Hahahaha!
Honestly though, knowing how horrible that could be, I never knew using a weaker PSU could be so handy, especially when it comes to "wanting a quieter and cooler machine". The lads reporting these info said the maximum fan noise value under load was 42-43dB instead of the 50'ish! (which is great, considering the GTX 1060 retains around 90% of its power)

Sent from my SM-N900 using Tapatalk

 

gtx 1060 mobile is 1/2 tdp of the 1060 desktop

1060N = 80-85w
1060 = 120~140 FE, AIB is higher

only MSi and clevo MXM 1060 have higher tdp

 

From the first released AW Echo models in 2015... 180w psu was the shipped psu. Maybe Dell should have continue with it? And rather put in more recourses for making a better working Battery boost?

 

I'm well aware of that, I wasn't referring to the full desktop variant, but rather, the massive PSU limitation which led the GTX 1060N to retain 90% of the 1060N's power when driven by the 180W PSU.

Sent from my SM-N900 using Tapatalk

 

I got back the Laptop from Dell after two weeks! And looks like the temps and clock speed are holding up pretty fine! (Though not sure for how long because last time temps broke loose within a week the previous two times)

To be sure everything is working fine:
1) What test should I run?
2) How long?
3) HWiNFO or Intel XTU for monitoring?

Intel XTU stress test holds below 80C for 10 minutes with undervolting of 130mV. No throttling. Is that a good enough test?

I want to be absolutely sure the CPU is fine before closing the case with Dell

Thanks

 

AIDA64 system stability test (Only stress FPU and GPU checked. Uncheck all other boxes). You can also run same Aida64 with stress FPU checked + Heaven simultaneous. CINEBENCH_R11.5 - CINEBENCH_R15 - wPRIME v1.5 (Run as adm. Select advanced settings and 8 threads. Fire up Wprime 1024m stress test. Hwinfo is good enough. Remember use max performance in Windows power plan.

 

Shall I run with undervolting or with stock settings?
How long to run before concluding everything is fine?

 

Test with Stock. This is what you got from DELL!! Run as long as needed... Eespecially AIDA64 system stability test (Only stress FPU and GPU checked. Uncheck all other boxes). And Aida64 with stress FPU checked + Heaven simultaneous.

 

Here are test results:

Ambient temperature: 25C

Prime95 FPU test + Unigine Heaven with undervolting to -130mV:
http://i.imgur.com/vDI1PyI.png

Prime95 FPU test + Unigine Heaven without undervolting! : Thermal Throttling starts!

After 4 min:
http://i.imgur.com/Xs8BfZT.png

After 5 min: Down to 2.6GHz. Can't hold max turbo boost. High temp even at low frequency
http://i.imgur.com/7nLw4uz.png

Full 10 min: BUG detected! 798MHz throttling. Experts at Dell said system is working fine
http://i.imgur.com/gPtN0cx.jpg


wPrime score:
http://i.imgur.com/1e8PGfM.png

Cinebench R15:
http://i.imgur.com/9ewoW23.png

Cinebench 11.5:
http://i.imgur.com/8qWi0XU.png

Your tests reveal everything!

I think my option is to demand a motherboard replacement. A strong feeling that there is CPU problem

 

You should post your pict with Cpu Package Power. Need those numbers. For same test below. Do same test, but also run test with FPU and GPU' checked only.

Spoiler: How you should put up Hwinfo

Remember also set Average monitoring as before

Edit. For comparison of scores. See HWBOT.ORG

 

Here you go sir. Power values too : http://i.imgur.com/E0iQmuq.png

 

You should have better thermal paste(better paste will lower the temp. Aka processor will hold higher clock speed). Your processor thermal throttle as now. Do same test but with undervolt. Test both scenario. Test 1 = Only FPU. Test 2 = FPU + GPU
Edit. Fire up Hwinfo after you start up the test. Or reset min/max.

 

Undervolted by -130mV

With only FPU:
http://i.imgur.com/65RuGeW.png

With FPU + GPU:
http://i.imgur.com/yulvUo0.png

 

Can you see if you can increase PL 1 and 2 in TS or XTU and run once again? As well Undervolt a bit higher if your processor can take it. Test only with FPU and GPU. This test is your problem (with load on cpu/gpu).

Edit. Post pict of XTU settings.