i7-3770k tested with IHS removed. Results?

[blackened23]

http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs

This incident has ignited two days culprit IVB loaded with hot, foreign websites overclockers.com full of curiosity to open the top cover of the IVB (IHS) and found the inside top cover and the core is connected to the grease, but not before usedthe materials used in soldering process (tin, and other low melting point metal), which is obviously the practices of Intel in order to save costs, because the behavior of the past, Intel in the use of silicone grease on the low-end CPU has always existed, but this time not even 3770K is no exception. In any case, the heat capacity of the metal is more than 10 times of the grease, that way we all agreed that the culprit leading to IVB high temperature is grease, but in theory seems to make sense.

As early as the time of the K8, I have opened a few CPU cover, from time experience tells me openings after cooling is not obvious, at most, 2-3 cooling. Therefore, skeptical, and I intend to put the Fengyun 3770K roof of my hands open, to see how changes in temperature after opening, the culprit that caused the high temperature is not the grease pit father.

Note: The following behavior is very dangerous, first of all to make your CPU lose the warranty if the injury core followed by a little attention, will make your CPU expired, so we can be spectators, not to emulate! I is irresponsible for any problems caused by the CPU top cover is open!


Test platform:
CPU：Intel Core i7-3770K
Memory: PCEVA Extreme Kit DDR3-2133 7-10-7
Motherboard: MSI Z77A-GD65
Video card: MSI R6570 MD1GD3
Hard drives: Plextor PX-128M2P
Power supply: Enermax Revolution 85+1050W
Radiators: Noctua NH-D14
Room temperature: 28 ° c
Grease: Prolimatech PK-1 (10.2/mK)

The results? No difference, IHS removal did not help. High temps are caused by the tightly packed cores.

 

[Avalon]

That's a bummer

 

[Don Karnage]

High temps are caused by the tightly packed cores.

Haven't we been saying that for weeks?

 

[BallaTheFeared]

Good find, thanks!

 

[blackened23]

I'm just sad that i'll have to settle for 4.5ghz-4.6ghz most likely. Thats roughly equal to 4.8 on sandy bridge anyways though, not a big deal.

Maybe this will finally prompt me to get some H2O going....i've been putting it off.

 

[FalconHorse]

I always knew the grease pit father was innocent.

 

[Don Karnage]

I'm just sad that i'll have to settle for 4.5ghz-4.6ghz most likely. Thats roughly equal to 4.8 on sandy bridge anyways though, not a big deal.

Maybe this will finally prompt me to get some H2O going....i've been putting it off.

You wont have to settle for anything. I'll never stray from you again Asus....

 

[blackened23]

Nice! Told you, asus #1

What kind of temps do you get with prime 95?

 

[Don Karnage]

Nice! Told you, asus #1

What kind of temps do you get with prime 95?

I use IBT and temps are in the high 80's

 

[Piano Man]

I"m almost glad it isn't the TIM, cause I don't know if I would have the balls to try this.

 

[TidusZ]

So, removing a layer of tim and copper before the heatsink made temps go up. That actually isn't possible assuming everything else was equal. I'd have to say there is something fishy with those results and I wouldn't be surprised to see the opposite from another source pretty soon.

 

[IntelUser2000]

Transistors have design targets just like every other project. Leakage, On-state current, switching frequency, can all change depending on the physical design of the transistor. Gate length, Fin height, seperators between transistors, contacts, amount of layers, all affect performance.

What works in one area doesn't always work in other. It wouldn't be surprising if the results are because the transistor and the design is tuned to low power(and stock) architectures.

Prescott used transistors with Ioff characteristics of 400nA/um(4x higher than normal). It may have been good if making a really high frequency device, but scaling down to low power wouldn't be ideal.

Intel's claim for Tri-Gate benefits are at 1V, and extended at ~0.7V. 1.2V+ stock for Ivy Bridge chips easily exceed this limit. But maybe its ok at that frequency. It's kinda unfortunate, but with advent of Tablets and slim form factors, this might not change in the future.

 

[BallaTheFeared]

I wish we had more samples to look through with moderately good water.

I dunno if it's worth giving up my 5GHz i5-2500k which is benchable at 5.3 1.55v


The little voltage list from Intel showing 1.55v max for Ivy gave me a bit hope.

The real question is if AMD/Nvidia are going to stop putting out incremental 28nm products anytime soon or if I'll be waiting for haswell on that front anyways.

 

[Khato]

So, removing a layer of tim and copper before the heatsink made temps go up. That actually isn't possible assuming everything else was equal. I'd have to say there is something fishy with those results and I wouldn't be surprised to see the opposite from another source pretty soon.

Eh, I'd both agree and disagree. I'll agree that it wouldn't be possible if all else was equal... but I'll disagree that we're necessarily going to see better results elsewhere because all else isn't necessarily equal. Namely, it's all dependent upon the thickness/consistency of the TIM layer. If Intel's process for attaching the IHS to the core allows for a thinner and perfectly consistent interface layer then it easily could do better than an imperfect direct joint between silicon and heatsink.

 

[LagunaX]

Nice! Told you, asus #1

What kind of temps do you get with prime 95?

IMHO I've had Gigabyte, Foxconn, Biostar, Asus, Asrock.

Foxconn was fine but out of the picture nowadays.

Biostar is snappy but the bios was always a little limited.

Asrock's bios and boards will never be quite the same level as Asus (the parent company) just goes without saying as Asus is the parent company...

Gigabyte's nice with the 2 oz copper and other board features

BUT

As far as overclocking ceiling levels, stability, lower voltage for the same overclock, bios for such, Asus has always been the best for me :thumbsup:

(but if DFI were still around i would give them a chance )

 

[dma0991]

My Core i5 3570K will get to keep its head heatspreader for now.

 

[oceanside]

Temps went up? Something could be wrong with this particular test. We need a bigger sampling.

 

[Arachnotronic]

IMHO I've had Gigabyte, Foxconn, Biostar, Asus, Asrock.

Foxconn was fine but out of the picture nowadays.

Biostar is snappy but the bios was always a little limited.

Asrock's bios and boards will never be quite the same level as Asus (the parent company) just goes without saying as Asus is the parent company...

Gigabyte's nice with the 2 oz copper and other board features

BUT

As far as overclocking ceiling levels, stability, lower voltage for the same overclock, bios for such, Asus has always been the best for me :thumbsup:

(but if DFI were still around i would give them a chance )

ASUS is not ASRock's parent company anymore. ASRock has been spun off and is its own entity now.

 

[thescreensavers]

So, removing a layer of tim and copper before the heatsink made temps go up. That actually isn't possible assuming everything else was equal. I'd have to say there is something fishy with those results and I wouldn't be surprised to see the opposite from another source pretty soon.

This, if the temperature was the same(within an % Error) then we could conclude that the TIM/IHS does not effect temps. Since the Temps went way up (over ~10%) something went wrong

Also the Testing was done quite poorly.

IMO- Get a CPU Temp App such as Core Temp. Have it Data Log CPU Temps and Run a Synthetic Benchmark, or any benchmark that you could repeat so that you could control what type of CPU load there is and repeat it reliably.

Then Simply input the data in Excel and plot the data against each other. You will then see exactly the differences between the two.

Id wait for some better testing. Imo results in that post are useless.

 

[blackened23]

This, if the temperature was the same(within an % Error) then we could conclude that the TIM/IHS does not effect temps. Since the Temps went way up (over ~10%) something went wrong

Also the Testing was done quite poorly.

IMO- Get a CPU Temp App such as Core Temp. Have it Data Log CPU Temps and Run a Synthetic Benchmark, or any benchmark that you could repeat so that you could control what type of CPU load there is and repeat it reliably.

Then Simply input the data in Excel and plot the data against each other. You will then see exactly the differences between the two.

Id wait for some better testing. Imo results in that post are useless.

Test done poorly? If you say so. He has several pics of him destroying his motherboard and CPU warranty, and you can see from his pics that his heatsink is properly applied to the chip. What I also see is that he tested at 4.5ghz with IHS on, and then tested again at 4.5ghz with LHS off. The temps were about the same. I think its about time to accept the reality that IB is just a hot chip because of power density. There's a lot of wishful thinking here that removing the IHS will suddenly allow the chip to oc to 6ghz...don't think its gonna happen. It doesn't really matter because you can still overclock it to the 4.6 and 4.7 ghz range. You need better cooling than SB....who cares?

 

[blackened23]

LHS ON:

LHS OFF:

 

[Absolute0]

Test done poorly? If you say so. He has several pics of him destroying his motherboard and CPU warranty, and you can see from his pics that his heatsink is properly applied to the chip. What I also see is that he tested at 4.5ghz with LHS on, and then tested again at 4.5ghz with LHS off. The temps were about the same. I think its about time to accept the reality that IB is just a hot chip because of power density. There's a lot of wishful thinking here that there's a magic trick will hapen. It doesn't really matter because you can still overclock it to the 4.6 and 4.7 ghz range. You need better cooling than SB....who cares?

By what I can see the temp increased 7 degrees over 4 cores or almost 2 degrees on average from removing a layer of metal and TIM. Surely they should decrease when removing a barrier to the desired thermal flow... would like to see more results.

I'm guessing the lack of mounting pressure caused by the fact that the manufacturer never intended the product to be used in this manner. Also the fact that the person who did the test didn't want to apply as much pressure as he would with the IHS on as he might risk damaging/cracking the die.

@blackened23
LHS or IHS?

Very likely. When I mounted on a naked Opteron 170 I cut off the little squishy pads that you find on the corners of an old AXP and put those on my processor to protect the core. On another processor, I layered duct tape (yep) around the edges of the CPU to prevent too much pressure from hitting the corners. And I saw massive temp decreases mounting my watercooling on a naked core.

I'd like to see how mounting a waterblock onto a naked core goes. There's got to be a way to get the temps down! lol.

 

[dma0991]

This, if the temperature was the same(within an % Error) then we could conclude that the TIM/IHS does not effect temps. Since the Temps went way up (over ~10%) something went wrong

I'm guessing the lack of mounting pressure caused by the fact that the manufacturer never intended the product to be used in this manner. Also the fact that the person who did the test didn't want to apply as much pressure as he would with the IHS on as he might risk damaging/cracking the die.

@blackened23
LHS or IHS?

 

[blackened23]

By what I can see the temp increased 7 degrees over 4 cores or almost 2 degrees on average from removing a layer of metal and TIM. Surely they should decrease when removing a barrier to the desired thermal flow... would like to see more results.

Have you ever run prime 95 with real temp? The temps are NEVER static and constantly fluctuate. The temp is about the same and within the margin of error.

By all means, remove your IHS next week though if you think otherwise. Personally i'll be more than happy if I can get 4.5 at a reasonable temp.

 

[Tempered81]

Seems like it's not worth the risk removing the heat spreader.

Here's a chipped & dead 3570K, heads up.

http://www.xtremesystems.org/forums/showthread.php?280598-Ivy-Bridge-3570K-IHS-removal-youtube