LIVE!!! Calling all XM chip, 3930K/3960X and ultrabook owners! Heat-Spring TIMs for testing!

I am totally in for anything. I pay shipping!

I have GPU which I overclock on 18%. So is it enough

 

Bump for curiosity and future results, gl guys! Me, too lazy to volunteer and can't risk my lappy.. Hoping here will be the future of TIMs if only for not having to reapply all the time and having equal or better results than icd

 

because XM chips have the highest temps and highest TDP. If GPUs could go to 150+ TDP and ran in the 80-100 C i would but the current GPUs in gaming laptops dont have temp restrictions. If i am wrong than we can look into some tests into GPUs with high TDP and high temps but as far as i know there are none. I dont think the 580/675, 680/7970 have temp problems are can overclock high enough to have an issue.

if i have extra i'll post about it.

thx!

 

You can get the 7970m pretty hot, as well as the 680m, and pretty much any other top end GPU with increased voltage. That heats up the GPU fast. 7970m/680m are 100w TDP cards, including the RAM, so probably ~75w TDP for the GPU, still more than any laptop CPU.

 

my 4670 runs at 99c with out fan blowing directly on the back of it. This is because the heat sink was designed for 25 watt tdp.

 

I doubt that you can do the same pressure on GPU as on the processor. Anyone has screws around graphics card? I doubt.

 

My 260m has 4 screws....only soldered on CPUs don't have heatsinks or crappy ones

 

Most GPUs have screws in laptops, actually.

@ OP, so when do you need preliminary tests/graphs? Or will you be reaching us or something?

 

yeah, my mistake then. My heatsink has screws for GPU but they screw in... heatsink itself and only thin to a mobo in a different place.

As I see my GPU needs even lesser size of thermal trans

 

i need to finish instructions and touch up terms and get everyone on here to post that they agree to all terms above before i mail out. I was supposed to finish that today/yesterday but behind a little. got to leave in like 5 mins because my grandma is about to die so not sure if I'll get it done today or tomorrow because of family stuff. What I want is 1-4 weeks of terms with ICD recorded and find your idle and max clock/load temps on your chips. Record and make those into graphs because we need those temps as our basis to test Heat-Spring. svl7 recommended using google docs so everyone can upload their results to an excel file there but i have to look into that. never used google docs before so i got to figure that out.

 

Take your time, HF. Just let us know when you've figured it out already.

 

I still need to touch up instructions and terms a bit more with greater detail and add graphs. touched up clarity and added some info and removed outdated info. Does anyone have any comments on updates? Am I missing anything besides instructions, terms, and graphs?

EDIT: my bother knows how to use google docs so he will show me how to get that to work this week. He does not know if you can make charts and graphs in it. Does anyone know if you can make charts/tables/graphs in google docs? He does not know that one

 

I'm not quite sure if you can do graphs in gdocs, either. You can do charts, though.

What data do you need for graphs? Just clock frequency and max temps?

 

Yea and comparisons of each in line charts. Bar graphs and line graphs would be perfect

 

For normalization and calibration purposes, I would strongly recommend users collect a few datapoints on ambient temperatures and fan RPMs if possible and as accurately as possible. If ambient temperature can't be measured conveniently close to the laptop, then perhaps collect data from other temperature sensors such PCH, HDD/SDD, Memory, which might vary proportionately with ambient temperature.

Sensor logging tools can increase CPU usage (mostly at high sample rates) and therefore could introduce some heat "overhead" in the collected data. So make sure the same data collection protocol is followed for the "before" and "after" measurements. A good tool for logging sensor data at chosen sampling intervals is HWinfo64 which can generate a CSV file containing the collected data.

With turbo boost technologies, an important factor is how fast can the CPU "recharge" it's thermal budget. It would be interesting if users could gather some high-sampling-rate (e.g. 100ms) core temperature measurements under various load conditions (idle, web browsing, benchmarks) to observe if there are differences in the rate of heat flow and overall performance of the cooling system.

Here's a rough example with a Clevo P170HM (i7-2860QM, GTX-580M) @ 30°C ambient (and external cooler):
Sampling set to 100ms, system at idle for first few seconds, then Passmark PerformanceTest 7 "Test > CPU All" was run, then return to idle. Note that "Clevo EC" was turn-on shortly afer the passmark test and hence the 5-10% CPU usage overhead for the last half of the graphs along with a significant decrease in sampling rate (~1 sample per second with EC monitoring enabled).

 

Well, you can collect all the data in a google docs spreadsheet, and then make graphs and stuff with excel later. I have Excel and know how to make the graphs and whatnot if needed.

 

i'll let you know if i'll need your help. I'll probably make the first one and ask you to split the rest so it is even. Maybe some other people wouldn't mind helping.

@arg8 I'll add some parameters in. As long as they maintain testing environment i am not too worried about ambient. It would be cool but I am not going to ask people to spend 20-30 bucks on an external reader. I am going to but I wont ask that of them. Also the turbo boost thing....i am a little sketchy on that for several reasons. I don't think turbo boost is set to heat as more of power draw. I don't think you will see any difference if it is liquid nitrogen cooled or air cooled. Also I am a little concerned on fan speed because the ASUS G51 doesn't spin to 5-6k rpms until 90C, which is retarded. So that would be important info to log depending on laptop. I think most Alienwares and other highend ones dont have such ridiculous settings.

 

Im in and trusted I would hope.. and soon to have an R2.. Ill test ether.. wouldnt mind testing the 3960x or the 3930x

i got all excited and thought you were testing chips.. if this is just for tim im still in.. im down for new tim.

 

most likely got all the desktops covered but keep me posted because i might need you. What laptop chips you got? Also can you get measurements of the die?

 

yea sure.. Ive got right now on hand.. the 2960xm.. m18x-r1 with the 3 pipe heat sink.. and I have the double pipe heat sink..
ill get measurements tomorrow for you..

also im willing to chip in a bit for my sample and a few others.. and i could cover my own shipping..

 

not worried about chipping in right now since i got 9 sample pieces. That gives us a good chunk to test. If I think we need more for testing i'll post but i highly doubt it. Now if this stuff actually works really well and people want to buy some we all can chip in and buy ourselves out fair share. 300 bucks gets 20 2x2 pieces if you buy a 1000 dollars worth you get about 100 pieces, which is way cheaper so if this really works out well we can figure out some way for all of us to buy some

Can you get the die size? Also what CPU are you getting in the R2?

Also that Sager NP6110 does anyone know the temps in that? Or if the GPU is upgradeable




EDIT: finished updating for tonight. Added glossary and other goodies.

 

HONESTLY NOT SURE YET.. ill let you know tho when i know.

 

Turbo Boost on Intel CPU's is purely modeled and based on power draw, it has nothing to do with actual heat of the chip (unless it hits TJmax and throttles)

 

Thanks for info. My new used CPU 840QM unfortunately doesn't show full potential. minus 1 step on 4 cores, minus 1 or 2 steps on 1 core and who knows how many on 2 cores.
It uses 42Watts under stress now. And used 35W when I disabled all but 2 virtual cores (1 core left.)

What could it be...

 

I am not sure what you mean?

 

I do not know the reason of not full TurboBoost.

 

Use throttlestop to see what is limiting it.

 

remember turboboost on 4 cores is different than a single core. check notebookcheck.net for specs. HT mistaken the 3610 for 3.3GHz on 4 cores when that refers to single core.

 

@HopelesslyFaithful I told my familiar (He is on NBR too) with 920XM in Alienware M15x r2 that he has an opportunity to test this TIMs. Perhaps he agrees.

@extide. I know this is out of topic, but I used throttlestop, HWMonitor, AIDA64, disabled cores in MSCONFID etc. Everywhere 1 step is missing.

Maybe later OS reinstall will help.

Otherwise I will risk to update CPU microcodes or bios.

 

The 840QM is not a great processor. They suck a lot of power so when fully loaded you won't get any Turbo Boost. That's normal. These CPUs are locked so you can not adjust the Turbo TDP higher. You need a 920XM or 940XM to do that.

 

I have a Sager NP9150 With the 3610QM that gets rather hot, like 85C with IC Diamond and i would like to participate in this test. If needed I can also test on my GPU which also gets quite warm overclocked, like 93C. My system specs are:
Sager NP9150
Intel 3610 QM
Nvidia GTX680M
16 GB Ram
Sandisk Extreme SSD
500GB hybrid drive with 4GB SSD

 

It is funny to see your post, unclewebb, because today I exactly talked in person with a guy who had 920XM and show me how he uses ThrottleStop with it

I thought about not enough power issue if it is. I have 120W PSU and processor takes up to 42W under full load. At this moment my laptop doesn't support XM processors. I assume because BIOS has Processor Power Limit 35W (don't know how it works though if CPU uses more than 35 anyway). What is your opinion on that and can I do smth to make 840QM work like 840QM instead of being 820QM?

P.S> Remember there are hidden power settings which make influence on TurboBoost.

 

honestly you make no sense. I don't know where to begin

 

I have a 920xm chip I can volunteer, unfortunately, it has no chassis as my old n61jq is dead. Let me see if I can find another housing.
Otherwise, I have a rather hot w110er with an i7-3610m which currently has ICD 7 and I have a couple of tubes of MX-4 to use as a control. I can also volunteer three i5-2500k processors overclocked to 4.2, 4.4 and 4.5ghz which currently uses Noctua NH-1.
Unfortunately, I'm in Australia.

P.S can the stuff be scraped out of the heatsink for warranty purposes?

 

HF, heya. I've figured out how to log and do graphs for the temps and stats I'm getting at HWInfo. I logged it on a 7-hour scale, but it's hard to discern when I put it in a graph. I logged the stats on a 2000ms scan interval. ) looks like a big massive splotch instead of a proper line graph.

What interval do you need the stats to be logged? I can help test what duration works best and stuff.

 

We don't really need help with creating the graphs, we just need people to do the testing and provide data. The reason your stuff is just a big splotch is because you have too many values (logged for too long) for the width of your graph. Make your graph wider or graph a shorter segment of time and it will look more like you are expecting.

 

you want to log every 1-5s i prefer every 1 second for complete data but when you creat a graph you tell the graphing process to only include x number of values or use all values to great a smooth line graph

Anyone want to make a tutorial on how to use HWinfo for logging temps? We need CPU/GPU temps if applicable. Also fan RPMs and if possible external ambient temps. I'll post a link to an item i'll be buying over the weekend if anyone wants to buy it too for testing. I have to use one because my ambient temps fluctuate drastically.

 

I know how to log but as of now the log saves everything HWInfo is monitoring. Is there any way to take that out and just save selected values?

 

not sure yet i haven't used that program...well ever. Speed fan has been the program i have always used I find HWinfo cluttered and useless to my needsbut i will check it out

 

HWinfo logging to file with format .CSV which can be readed with Microsoft Excel, and maybe it can be readed with similar program like Openoffice/Libreoffice. Unclaimed data can be deleted or hided.


By the way i have 920XM in my M15x, and i'd like to take part in testing TIM. And i can pay for sending.


Update: Little tutorial for logging with HWiNFO (maby will save somebody's time)

1. We need to download HWInfo ( HWiNFO, HWiNFO32 & HWiNFO64 - Hardware Information and Analysis Tools)

2. Starting HWInfo and opening Sensors screen:


3. Configuring Scan Interval (changing value in ms / 1000 ms = 1 s / it's up to you what delay to set) and Sensors Display (while disabling/hiding unneded sensors)



4. Start logging and type the name of logfile


5. Press Stop logging (when you finish logging) to save the logfile.

At last we will have some file like HWiNFO_log.CSV or another name which you will chose.
If we will open this file with MS Excel (just for example), we will be able to save sensors table in another one format and than edit.

P.S. Sorry for my English ...

 

nice +rep

I'll add that to main post when i get time over weekend. I'll be busy of the next day or two

 

Oh so there's no way but to disable monitoring for those, huh?

Anyway, good one. I can start doing control tests. Which specific data do we need to log for this?

 

just wait until monday to start control/base test i'll get back to you Saturday on that

 

My understanding is that one of the new features of Turbo Boost 2.0 is that it factors-in thermal performance as well and makes use of thermal "headroom" as much as possible. I agree that power draw and time parameters can be specified by the PC OEMs, but these can be overriden and customized in some cases, as these are limitations of power circuit hardware implementation and not limitations inherent to the CPU.

I thought I noticed a few scenarios in the past where peak turbo boost could be sustained longer at max power draw, and also seemed to hit the top multiplier more consistently for idle tasks when the ambient temperatures were relatively low. However, I may have been fooled by an algorithm governed by power draw/time windows. I'll try to look a bit more closely into this when I have chance.

It would make sense though that with a power supply that has enough margin to sustain maximum CPU power draw indefinitely, the CPU performance should only be thermally limited. Hence turbo boost behavior should be strongly determined by the active cooling capabilities of the system. If anyone has seen any documentation that describes exactly the internal algorithm that governs the i7's Turbo Boost behavior, it would be enlightening to many. (Or is it Intel's "secret sauce"...)

 

uncle webb is the man to answer such a question but i thought it was purely based off of power draw and multipliers that intel set. I don't have the time to "attempt" to explain what i mean so i hope uncle webb shows up again and posts a good explanation.

 

You are mostly correct, it is based on heat, but not measured heat, rather modeled heat.

 

it is modeled to draw and use the amount of power around a 45w TDP but actual heat has no effect to how it works. Right was that phrased decently?

 

Yes. AMD GPU's also work this way where as nVidia Kepler uses actual heat, and the problem with using actual heat is the thermal sensors are a bit slow to respond, where as the model can be just about instantaneous.

 

what I hate about TDP is the fact like iGPUs get screwed. they never give enough TDP for the CPU and GPU to be at max at once. so my netbook could play a couple decent games if the GPU didn't get hammered first. it is the SB ULV that is 1.3GHz i forget the model. The lowest ULV one

 

I understands that it models the thermodynamics of the cooling solution, but I haven't seen documented anywhere whether these model parameters are hard-coded prior to runtime, or dynamically adapted during runtime based on environmental conditions. From the following (assuming no power supply limitations or locks are in place) it would seem that power parameters are estimated, while temperature is continuously monitored during runtime:

www.intel.es/content/dam/www/public/us/en/documents/datasheets/3rd-gen-core-family-mobile-vol-1-datasheet.pdf#page=43

Note: They don't use the word "estimated" in reference to "The package temperature".

So are you suggesting then, that the temperature monitoring is purlely for thermal throttling at critical temperature thresholds and that "The estimated package prior and present power consumption" factor, i.e. power integration = energy, may be a hard-coded parameter into the CPU SKU? Hence, Turbo Boost cannot ever be sustained above "power-integrated" TDP even when cooled with liquid nitrogen?