P150SM Temps

No!- any and all screws should be tightened adequately! I normally use 99.9% Isopropanol for cleaning

The poor contact is because the heat sink is not the correct shape exactly, if your good enough, BEND it to the right shape, but go easy - it requires very fine adjustment, you will have to repeat the paste job AT least 2 or 3 times using a tiny tiny tiny thin screen of paste - just to check for contact NOT enough to run the machine on!(ideally using " engineers blue" - (do not run the machine with "blue" it is purely for measuring flatness)

You can use the washer method but its not solving the problem, and its not a great idea IMO although it will work no doubt - it's very bad for the life of the solder that bonds the chip to the board to apply more than the " absolute minimum required" pressure on a BGA chip - especially a hot one, and the 780m are certainly capable of throwing heat out!


So, to summarise your options are both a bit poor, bend the GPU cooler - risking damaging it, or use the washer method - and fail to fix the root of the cause, and just employ brute excessive force to get contact...!

 

With respect to temperature issues when the notebook is under heavy gaming load, would anyone recommend the p157sm over the p150sm? A few have mentioned in this forum that the p157sm has an extra vent where the GPU is...this theoretically should reduce the temperature?

 

I could be preaching to the choir here given the number of computer enthusiasts here but I'll explain anyway.

You could actually have a defective heat sink. Not in the sense of inadequate contact with the chip but in the sense that the heat sink is actually non-functional. The copper tubes (heat pipes) that run between the chip and the radiator are not solid copper. They are hollow and filled with a small amount coolant and sometimes a wicking material that coats the inside of the tube leaving a hollow core (see image). They actually work similarly to an air conditioner without a compressor. At the hot end a fluid is evaporated, taking heat energy along with it. The vapor works its way over to the radiator end where it cools and condenses back into a liquid releasing the heat into the radiator to be dissipated by a fan. The wick then pulls the liquid back to the chip where the cycle repeats. If by dumb luck you got a cooler that has had the coolant escape, or was simply not put in at the factory, it could very well cause your overheating problem.

Spoiler

 

I have wondered this exact same thing myself, the case and exhaust vent is different, however in my opinion the exhaust port on the P150 is less restrictive than the P157 - however temperatures might not be a good indicator, temp&fan speed along with ambient temps (which can all be monitored and logged with HWinfo) would be a much better test and easy for anyone to achieve. throw in noise levels and we'd have a really great comparison as typically a more restrictive case will make more noise forcing air about - but this isn't a factor that could be monitored well enough, it would have to be 2 machines, each tested in the same room at the same time, with the same dB monitor

 

and these are ethanol based working fluids with a sintered copper powder wick

- and just in case any wondered, without impurities & in a high vacuum the speed at which the heat is transferred from one end to the other is ............ drum roll.......... 0.67 billion miles per hour (seriously!)

yes it really is that fast.......... the real limiting factor is the speed at which the working fluid can condense and return to the hot side.... this is the engineering challenge I spend many a nights wondering about.......

 

That's also why depending on the materials used heatpipes that are seemingly smaller can outperform larger ones.

 

OK, so I'm not getting my target temp of 80 in Metro Last Light.... seems odd it must just be the game engine I will have to investigate further and compare with Battlefield 3 tonight

I noticed some strange bench results for LastLight, I think it could be in desperate need of a fix/patch - the graphics aren't particularly special or detailed compared to games like BF3 but it really stresses the system out and it does visibly slow down in game, a quick alt tab revealed the CPU was still turbo'ing well but the GPU kept hitting 91 odd, and throttling

 

Calibre41, any chance you would like to make a crash course in how to OC the cpu with Intel xtu? Any help would be greatly appreciated.

 

I would also be really interested in such a xtu guide

 

Well I'm using a specific version of XTU, I have heard that not all version work properly.

IntelXTU from Clevo page

you can get the version I use from the P150SM drivers page (for now at least) link above.

this is a shot of the XTU from my very own P150SM

http://imageshack.us/a/img541/364/gnaz.png

you will notice simple sliders to adjust the turbo multipliers for the 1, 2 3 and 4 cores

you can also adjust voltage, adjust turbo durations and TDP (change it to 37w and it activly underclocks just like a 4702MQ), and disable power saving and other cpu features which improve stability (you'll find these in the "other" section if you look on the left of the image, you can see I'm just viewing the CPU options. sadly although XTU is able to adjust memory timings it seems we have deprived of this feature

These are the following clocks intels unlocked for us on non "extreme" editions... ( I haven't fully confirmed all these but I'm 99% sure)

i7-4700MQ can be given 200Mhz overclock (so max is 3.4Ghz 4 cores to 3.6Ghz 1 core)
i7-4800MQ can be given 400Mhz overclock (so max is 3.9Ghz 4 cores to 4.1Ghz 1 core)
i7-4900MQ can be given 600Mhz overclock (so max is 4.2Ghz 4-cores to 4.4Ghz 1 core)
i7-4920MX can be given any overclock

Interesting, the Desktop Haswell's are totally locked.... so we basically have more overclocking potential on our laptops....... that's a bit odd but "Thanks Intel"

 

thank u for the insight. but isn't the 4700mq's standard speed not already 2.4 to 3.4 ghz? or do I remember it wrong?

 

You don't remember it wrong, your reading it wrong!

the core will hit 3.4ghz standard yes, but only when its running on one core (pretty much, never!)

when running on 2 or 3 cores, it can hit 3.3Ghz, and if all 4 cores are getting "some" use it will only clock to 3.2Ghz

Intel has unlocked 200mhz overclocking ability on the little 4700mq, so you can add 200mhz to all the values above (in bios or via XTU but only if your laptop manufacturer supports this)

I now have my CPU set to 3.4 on 4 cores, and 3.5Ghz on 3,2 & 1 core

 

Calibre41, thank you for the guide, I'll try to mess around with the settings.

 

back on track...

My gpu does hit 93-94 in Metro LL and when it does I lose my 849mhz boost clock and drop 1/2 the way or all the way towards the stock 771mhz and I lost a significant amount of core voltage, I'm unsure if the GPU fan is trying hard enough, I'll try again in a bit with Fn+1 full fans - I'll then try without the bottom cover on (if I can safely - I have some special stand off's I made which will screw into the old screw old screw holes, so I should be able to get some lift while the bottom cover is off using these

I'll also monitor GPUZ as it has a "Perf.Cap" field the records when a boost clock is lost and why - different colours denote different reason (utils is grey)

 

I thought stock clock was 823MHz?

NVIDIA GeForce GTX 780M - NotebookCheck.net Tech

That is reference speed if I recall correct. Anything less means the vbios was modified to lower it... If on Sager/Clevo it is indeed lower that means they lowered the speed of the cards to try to address heat/power concerns and we are not getting true 780m performance... It is suspected that the 780m on stock can suck over 100w which on the 15" would put you over the PSU limit almost... Esp with the 47w CPU when you take into account the screen, chipsets, drives, USB ports used etc.. I really wish they used a different connector on these laptops so could use better PSU

Great hardware limited by a 50 cent part and 100 PSU, which while is a great PSU just really doesn't meet well. It's never good to run things at 100% capacity but that's what it seems to do on these bricks unless you are just web browsing.

FWIW I know nvidia leaves speeds up to the OEM and does not publish true TDP... Just seems odd. I have seem 780M's showing different stock clocks though. Would be nice to run full spec speed though :/

**Update
Upon further reading it looks like others are using the lower clock too. So I guess we can assume this is due to power to keep it at the 100w area. Seems odd givin the 770M technically has a stock clock of 811 which is over the 780m lower set clock... Wish had more to compare or read as to all of this. I guess it's hard to read benchmarks when clocks can be all over. 780M seems like it's downclocked 7% off the bat.

 

GeForce GTX 780M | Specifications | GeForce

Yeah the official page site says it's 823 but there is no mention of the TDP - I suspect all manufacturers have dialed it back to keep it cool - and clearly she prefers 771-800 in my current climate (and without a cooler board under the machine which might make things closer to 849mhz)

so I'm running a little test mod as I mentioned earlier which I made for the P150SM, ........ more details on this later, but basically, It's a set of spacers that allows the back cover to be screwed on and secured, but spaces it out so it's opened up about 8mm wide.

here is how she looks now............... open wide.....:say ahhhhh

 

I don't know that what you did would help much because it causes the vent holes to be somewhat different on the bottom as far as how it gets air. While it still can it is forcing the air to take a different path and "could" possibly make it suck some warm air back in depending the direction it pulled from. Remember it will travel least resisted area which mean prob from back and sides as the back will have the large opening now, which is also where heat exits near.

Better would be no bottom cover but that could be difficult to do as would need to get it to sit properly without putting pressure on any components or blocking anything :/

As for TDP, The 680MX uses 122w and the 780m for all intensive purposes is an overclocked 680MX so it's probably the same at stock clocks. I have seen a few sites mention 122.
The 780m features better GPU boost but it has already been said GPU boost is not HW based but SW (really FW) so basically we have this 122+/- few w GPU in our notebooks with 180w PSU.

CPU takes 47+w (it can use a bit more on the turbo).
GPU, even if you want to say 105w we are at 152w. Now add 3w for ram, 3w for an SSD we are at 158w and have not even took into count the monitor which is probably 10-15w, plus power for mobo, wifi, USB etc.

One thing I have learned with electronics, too little power is just as bad as too much and can also impact performance and heat. Though common myth is less power=less heat, for most part that's true but straining things can create heat.

 

good idea but i think raising the back roughly 1/4 inch would be better.
opening it slightly like that might just let the heat escape into the rest of the innards instead of the fan being flush with the plastic base.

 

as I mentioned earlier those temps are with the back raised already, and over 1/4" closer to 3/8"

Temps are much better with the case mod, Its running 4° cooler and sitting around 89-90 meaning it is holding 800mhz now,

Bryanu, I see what your saying but you'd have to just see the layout - basically It's not going to take hot air from the rear as the cooler iteself blows out this way, and the exhaust is right next to the opening I've created which causes a vacuum pullling air out of the back if anything and not inside- so it would be very difficult to pull hot air back in and over come this natural effect

It grabs fresh air from the sides, and the normal holes in the base, I can really feel it pulling fresh cold air in the sides.

Nice results I think

Interesting it's also holding its 849mhz clock for some time now but it does still break up, and rarther than before where it was becuase of temp's it reports pwr "total design power breeched" so it is failing to boost because it has hit it's max TDP........ I wonder if it's been tweaked down to 100-115w rather than letting it fly at 122w

 

thats good to hear then Calibre. nice one

 

Yeah I like this, it worked a measurable amount more than I was expecting to be honest - to be fair, I was expecting it to be the same - its all looking good then for my other case mods which will give it even more room to breath

I edited that last post, seems I'm hitting a TDP limit now at between 849 and 800mhz just every so often - it's not to bad, I think we need to undervolt these and fine tune them a bit more accurately than the "default" bios offers to really see the potential unlocked. - lower temps, higher/longer turbo's

 

My understanding is the 780m is already undervolted. It's built on the same chip of the 680MX and running at slightly lower voltage and a tad higher clock. (thus how the claim is it runs faster vs. a 680mx but uses same power window.) Except in this case the manufactures have downclocked them also so basically we are running a 680MX. All the specs compare them running at full reference clock but the 780m's are not running at that thus not true to compare.

Undervolting only works well generally if you are not trying to OC at the same time. So running stock speed, no turbo you can get away with some undervolt. You can actually damage equipment though by lack of power just as you can by too much.

I have fried plenty of things by lack of voltage and seen my fair share of it done from others too. While it could run and seem fine for a while one day she just dies.
On the side I do a lot of hardware testing for manufactures, from routers to other computer hardware and some of this is things we play around with, lower voltages etc. That's how I always learned going lower is not always good. IF the manufacture already has it tweaked as good as possible, and these days they are all about making more efficient so chances are it's fairly close, your play room is limited.

Back in the days auto boost for GPU/CPU didn't exist, people learned you could safely push things, and now the manufactures are doing this, but now people are wanting to push what is already pushed when really the manufactures are already eating most of our extra play room.

That said, I don't think ANY modding of the case or opening should be needed for your machine to run right. 90c to me is too high. I like 80-85c. I would love to know how the 157SM runs compared or even the 17" which is basically the same just larger case and PSU.


In all we have great machines but Clevo failed on a few spots I think:

1. Need more fan steps for both noise and better cooling.
2. Need different power plug and higher wattage PSU
3. New case design (P1X7's, old style of course this don't apply) helps cooling with more vent's but hurts by blocking part of the exhaust. probably a wash lol. Why can't they address both?

If number 1 and 2 was done these bad boys could run at full speed and give us much more play... Worst part is the first option is simple for them, unless they have such an cheap EC that it's limited on number of steps...

 

I would reserve judgement on how well Clevo did until I saw how well the likes of the 17" models and the Alienware M17x and MSI GT60/70 did, I suspect that the cooling and ability to maintain turbo frequencies on the Clevo is actually among the best of the bunch!

The EC handles a phenomenal amount of work, you'd be very surprised how much of the machine is wired through that, pick up the service manual and sift through the schematics ands you'll see.

The fan speed could easily be addressed by setting the minimum step to 700-1000 rpm rather than zero, that should keep the temperatures well under control with minimal noise, if any

Regarding under volting, these cards also have over current protection, so this should address the issue you're referring to of under volting causing damage (sort of) I'd say find the lowest voltage it will hold a clock on, and then half the difference although I'm fairly sure the normal rules of lower voltage/higher current don't apply to CPUs/gpus as the voltage directly affects the frequency it can operate at, so it would stop working before currents reached to higher level

 

undervolting will never cause damage, the power draw drops with an undervolt the current will not change much.

 

I will definitely try the modified vBios for the 780m some day so the PowerLimit is increased to 106-120%(?) like i can natively on my 780 i know it's still a notebook but i think it could work when using a proper PSU. I get so magically attracted to these 1006/3000, i still can run ~950/2700 but when e.g. i put the memory to 2750-3000 my Voltage drops and i guess due to Powerlimit but didn't investigated it yet

 

Yeah, if you start wanting to raise power caps replacing the PSU is a top priority.

 

Anyone managed to reduce temps?

I've snapped a screw from my 780m sending it back to my supplier to resolve for me. Having said that, I'm getting 3 degrees less in temps without 1 screw!

 

There was someone using the washer mod.

 

Temps hover around 87-88 at the minute.
Ran 3dMark11 and got 7830 tonight with the latest 326.19 beta driver.

Think my laptop is going in for repair this week, so I'll get them to do a decent repaste job and see what to do from there.

 

Is it safe to play with 80-85 degrees?

Under metro on max settings it gets quite hot. 83*

In Rage - 59*
In Dark Souls - 50*

My worst nightmare would be that my GPu dies cuz of high temperatures Just got my P157SM

GTX 780m

 

could somebody who gets max temps around the 80's do this test for me: From a cool notebook (after startup, with the 780M GTX switched OFF and presumably at room temperature), open GPU-Z and on the sensors tab show the minimum temperature. How high is the temperature just the instant when just powering on the card? I get around 40 C, that is the lowest it will ever be, which is quite warm.. maybe my temperature sensor is not correct?

 

The sensors are not accurate at lower temps sometimes. Also they are subject to the warm air inside the casing

 

Hi dquancey, can you let us know what you ultimately did to correct the temps on your 150SM ?

I'm having similar issues with my 170SM / 780M GPU. Temps are hitting about 92-93C in Hitman Absolution & after 3-4 conseuctive runs of 3dmark 11. I hear the heatsinks are similar between the 150SM and 170SM so am looking forward to hear how you took care of the heat issues. Thanks!.

 

The latest drivers have knocked my temperatures down significantly, and I noticed it holding the full 849mhz turbo - constantly in WoT and BF3 - I have yet to try Metro Last Light although I think it is becoming clear this old game engine is probably not as efficient as others, largely due to it's age - and one (or 10.....) to many buffs - time for a new engine from the guys at 4A I think - this one's been buffed enough, let go....

heatsinks for the GPU are the same over the range - so P150SM, P157SM, P170SM & P177SM all use the same GPU heatsink. The P17* however have a copper finned CPU cooler, whereas the P15*'s have aluminium.

I was supposed to have a couple of the all copper CPU coolers, and a few spare GTX 780 coolers but sadly they still haven't arrived yet

I look forward to giving Metro Last Light a run and comparing - as this seemed to be the killer, really really heavy GPU load like I've never seen on more modern game engines (maybe down to the PhysX?)

EDIT** sorry, I did just forget I have this case mod........ its early in the morning here....

 

I haven't really resolved it, yet.

I've just repasted a couple of times with MX-5.

Still hitting 90 degrees during BF3. So problem still there, as far as I can tell.

 

You see, Last night I was getting 80-85 in BF3 (I do have this case mod obviously), with all settings on ultra and my 60fps cap off - it was running at the full 849mhz all the time!

can you run GPUz in the background and see if you get any coloured bars in the Perf.Cap - and what are they (presumably, throttling due to temperature)

I did forget I have this case mod........ its early in the morning here....

 

Remember with turbo and temperature targets the 780M will run hotter than the 680M did.