i7-3740QM

I received my 3740QM yesterday, was benching it, and it it's hot somehow and throttle down to 3.1-3.2GHz:



Are you guys seeing similar temperature when XTU is off(Including 3720/3820/3840 users)?

 

those temps are a bit high, you should probably repaste. my 3720qm, OC'd to 3.8Ghz on all 4 cores (via XTU with TDP raised to 55w) just about touches 90c after 5 mins of prime, - this is with the back of the laptop raised a little bit for better airflow and gelid GC xtreme compound. Your temps should be even lower.

 

Yea it's really weird, I have repasted 3 times, all with similar result... You have a P150EM, you flashed P170EM bios like Meaker too?

 

Is it an OEM cpu? Somtimes ES / QS editions run a bit hotter / less efficient than OEM releases.

 

It's an ES. I hope that's not the case .

 

What CPU did you have before and waht temps did you get with it? There have been lots of reports of differences between heatsink performance of various EM series clevos. Some need to be foil modded to imrpove cooling performance, I myself ordered some foil tape to do mine.

 

I did foil tape mod was 83C with 3610QM at 3.1GHz with the same benching I did for the 3740QM. But the temperature gap between each core isn't that big like the 3740QM.

By the way core 4 to 7 always report a higher temperature than core 0 to 3, happened with my 3610QM as well, but again the temp gap between each core isn't that big like the 3740QM I have...

 

As you rack up the speeds your thermal paste application becomes increasingly important. I also added some copper ramsinks to my heatsink.

This means I can run CPU tests without throttling with the power max around 65-70W.

 

do you know mate if i can flash my bios to the P170HM(XTU),as you did it for P150EM?
im interest for 2920xm.
thanks budy.

 

I have no idea i'm afraid, my first inkling would be yes, but it's uncharted and potentially dangerous waters.

Also something like the 3740qm would make more sense.

Push an XM in the 15x series and you may blow your mobo, PSU or both.

 

LOL
so i dont even try....I need my lap....
thank you meaker

 

Hackness I'm seeing exactly the same results with my 3740QM QS.

Hmmm not good.

 

Unfortunately not. In the HM series is not that simple.

 

Ohhhh lol.

You are running a linpack based test.

You are hitting your TDP limits and therefore your CPUs are downclocking.

It's perfectly normal and only by controlling the TDP limits in the XTU can you stop it.

Here is my SS:

http://i4.photobucket.com/albums/y143/Meaks/TSB.jpg

You can see I get a little bit of temp throttle at the end, but my settings are designed for gaming rather than benching and such extreme load is rare.

 

Is this something wrong with the Ivy itself? I cannot believe that 4 cores running at 3.1ghz only drawing 43 watts can clock up 90oC? Even my 920xm sucking up 90watts only scraped 90oC and that was an ES as well.

 

Ivy is a hot chip but the fan profiles are quite lax.....

 

I swear 28 and 32nm just want me to skip a generation even though I really do not want to

 

Well the 680M is certainly shifting its weight, and the more conservative 3740qm means I can use it all under a 180W brick unless I want to go silly on the 680M.

 

Ahh I didn't realise the throttlestop bench was linpack based.

Are you sure about not going over 180W though as HTWingnut was hitting 180W in some games with only a 3610QM and 680M Saltius OC which is not even as high as svl7 OC?

 

Was that from the wall? You wont be taking losses into account there.

 

Yes I believe it was.

 

Then those numbers were much higher than mine

EDIT: Also it's a pi calc, no linpack, whoops. But they all go under the same category and kind of load they generate.

 

Sometimes when playing BF3 it can spike to 90C, now I pretty much just set the multiplier to x33 when using without Fn+1

 

true, but my 3720qm sits at a constant 3.8ghz in prime, so its definitely able to run cool enough for p150em's thermal system. those temps are either because your heatsink does not make proper contact (has been noted on a few EM series laptops in one of large threads here) or that ES cpu runs hotter than OEM sample (my previous 2 ES CPUs all ran hotter than OEM replacements, both had around 10c difference (QX9200 and 2920XM ES).

 

Prime95 drops me down to 3.0GHz after a minute or two

 

83c with 3610qm is very high. when i had 3610qm i would max out at 68c doing throttle stop 1024m bench. Variation in temps between cores looks the same as mine, I think its because certain part of the CPU will be closer to the fins of the heatsink so temps will be lower in that corner. Maybe your heatsink isnt making full contact here? this thread might be of use: http://forum.notebookreview.com/sag...0em-7970m-gpu-cooling-performance-foiled.html

 

Sadly normally the CPU matches up from the radiator to the fan quite nicely it is normally the GPU side that has the issue.

Just a theory but with a DIE shrink surely it would be harder to cool anyway? A smaller area of effect for the heatsink to make contact along with several components generating heat in close proximity of each other is only going to increase the heat output surely?

Or I may just be a drunk brit making conversation

 

Try running the intel burn test..... Then post a ss. 3.8ghz is geater than your 4 core turbo so i think you are only testing one core.

 

check the screenshot on page 1 of this thread. 3.8ghz is with XTU OC at 55w tdp, definitely on all 4 cores at full load.

 

I confirm the ES/QS Ivy Bridge runs hotter than the OEM's (just reverted back to a 3610qm) - but Articjoe temps (68c with 3610qm and 90c with [email protected] after a 1024m TS) are mental to me.

Also, on a side note, I reverted back to a 3610qm and now my 680m doesn't get hot like hell as before (obviously because MSI has got only 1 fan) so I can more easily crank the GPU and that's what matters

 

It only has one fan but the heatsinks are separate, so a hot CPU should "overcool" the GPU....

Also I would need a graph to tell whats going on because a single snapshot in time is pointless.

 

Not when you're playing demanding games (such as BF3) for several hours with a 680M overvolted and heavily overclocked and a 3820ES @ 3.6ghz - both get really hot - the fan has a limited rated airflow so if both components get really really hot, the fan can't cope with such power dissipation.

 

But the airflow goes 50/50 regardless of how each component is loaded. The fan will sit on the highest speed either the CPU or GPU requires, so there should be no issue.

Also your 3820 ES, which stepping was it?

 

Here are the rated power consumption figures as I up my CPU while running the test:

3.1ghz: 34W
3.2ghz: 36.9W
3.3ghz: 39.8W
3.4ghz: 43W
3.5ghz: 46.4W
3.6ghz: 51.0W
3.7ghz: 55.1W
3.8ghz: 57.0W
3.9ghz: 59.0W

 

I have a 3720qm capped with the +400 MHz and I only get up to 78 in IBT. I put 5 heatsinks on it and have my notepal U3 blowing right in those vents above the cpu.

 

This is how quickly it completes without browsing the web while running it.

 

Meaker what speed does your CPU drop down to during the test?

You also hit 100 degrees as I do :O

 

Basically the ES/QS chips have a worse power efficiency than the OEM's - of course we should check and test more in details with the ambient temps etc. but from a first view, the 3720oem's from fenryr and articjoe do run really cool.

I can't remember which stepping it is, but must be an early chip and my 680m temps just drop by at least 6 degrees going from 3820 to 3610oem while playing BF3.

 

It reaches 100C in that test because the speed does not drop, that's 3.9ghz the whole way through because I have unclamped the turbo limits.

I think I can forgive a 3.9ghz 59W fully stressed CPU reaching 100C in a notebook.

 

God help you, 100 C is bloody hot!

Surely hope this is something you've done for the sake of benchmarking

 

Well I don't run prime or superpi all day so i'm good.

 

I limited the multiplier to x31 to see if the temp is like running a 3610QM, however the 3rd core (thread 4 & 5) is always running 2C higher than my 3610QM.

 

Oops double posted.

 

2C is well within the error margins of these inbuilt sensors lol.

They are designed to be accurate mostly at the point of throttling to ensure that happens at the right time.

 

screenshot of 5 mins of intel xtu burn test attached.

 

Yeah but you are still power throttling, look at your frequency graph, hardly flat, and temps reported in XTU are lower than throttlestop and you have not moused over the higher parts.

 

actually that was one of the higher points in temps, max it hit was 87c on one of the cores. I dont know why frequency keeps dipping, its definitely not thermal as it happens even when temps are bellow 70c, but if i run prime its a near constant 3.8-3.75ghz all the way to the end. I will try upping the TDP and see if it smoothes out the graph, though throughout the run it didnt seem to hit 55w limit at all.

 

ok, even when I upped TDP to 60w my XTU stress test is still full of dips. Does eveyrone elses look the same? In contrast throttlestop ran much closer to 3.8ghz through the whole bench.

 

Up the TDP limits more (both short and long) to 65W and then see.... pay attention to throttlestop's power consumption reading.

 

Don't feel bad.. i've been playing with my 3820qm in my P170EM and I can't get it to hold anything higher than like 3.4ghz after loading all 4 cores. that's LOWER than Intel specifies which should be 3.5ghz maintained with 4 loaded cores. And my temps are hardly bad, i'm 78-79C the entire time when running Linpack, Intel burn, Pi, etc. with stock clocks yet it bounces my speed around for no reason.