De-lidded i7 3770k Test

T

TitanWarrior

n00b
Joined
May 15, 2012
Messages
7
hi all

so as the title says i delided my i7 3770k today :eek:

I used coollabatory liquid ultra then ic diamond between the heatsink and ihs

sorry forgot to take before temps screenshots. But my idle temp before was 40 on core 1, 35 on cores 2 & 3, and 32 on core 4.

my stressing temps were about 85 on linx

inside temp 70 F

these are the after temps






a side note i hate wearing gloves



 
Last edited:
Really no before hand testing on the temperatures...LOL? :p okie-doki then. Not very informative is all I'm saying.
 
I think he meant to say, he didn't take screenshots. He did say his before temps were 85 on LinX, which then became max of 78 afterwards.
 
Thanks for doing this and posting the results. Any plans to re-lid it with the better TIM under there?
 
yes i meant to say forgot to take screen

i did re lid it i used coollabatory liquid ultra on it in between the cpu and ihs
 
Oh, I read that wrong. My bad! Thought you were running it direct HSF to CPU die.
 
What was your vcore voltage under load before and after?


Your hottest core before was 85C and after it was 78? Or did you add up all 4 numbers before and divide by 4 to get your average temp of 85C with linX before doing the TIM mod?
 
You took the IHS off, replaced the TIM, then put it back on? /facepalm

Also it looks like you put too much TIM on there. Hard to tell from the cell phone pics...
 
Well, he delidded it, replaced TIM, put it back on, and got at least a 7 C decrease on the hottest core. Don't see what there is to facepalm about that...
 
Tsumi said:
Well, he delidded it, replaced TIM, put it back on, and got at least a 7 C decrease on the hottest core. Don't see what there is to facepalm about that...
Click to expand...
Well, I guess the facepalm meant that he could have done even better by applying the heatsink directly to the chip. But that is admittedly a quantum leap in effort with a greater risk of breaking things, so it is understandable that someone would only replace the TIM. Actually I am rather amazed that there is a 7C advantage in just replacing the TIM. So not only did Intel fail to use FS (which we all knew), but they must have done a crappy mounting job with the IHS too. I wonder if they use the same guns to apply the TIM which Taco Bell uses to put sour cream in my tacos.
 
Most of the tests with direct die contact have shown less of a temp decrease than his testing, so I would say he might have the right idea after all. That collaboratory liquid is supposed to perform closer to solder than to TIM, so it makes the IHS work more like it would with a soldered die.
 
Indeed. And as other people are saying, it's probably notoriously hard to get proper contact with the direct die, but it is much easier to get good contact between the die and IHS, and then good contact between the IHS and heatsink.
 
ok 85 would be very close to avg with the intel junk.

vcore the same 1.25. if i take it to 4.6 it takes about 1.32 to make it stable which would get in the high 90s. now if i do that i get my old temps

not cell phone pic just a crappy camera at taking up close pics.
i dont know if i got too much. but i only put a little drop on both the cpu and ihs then brushed it. i would rather get a little to much then not enough. plus it pretty much liquid metal

i didnt want to get out the dremel and start modding my mounting bracket
 
Got Brent's new test system with 3770K running 4.8GHz at 1.275vCore and temp is 58C under full load.
 
TIM replaced with high end cooling could do that easily. I can keep my IBT temps at 65C at 1.24V with just an H100 with TIM replaced.
 
tbooth said:
Is that with the IHS TIM replaced? What cooling?
Click to expand...

Stock TIM, Corsair H100. Been running solid 2.5 hours now on another full load CPU/RAM test with the CPU sitting at 61C.
 
ferzerp said:
Well, you've obviously done something.... ;)

Spill the beans.
Click to expand...


Done nothing hardware-wise. What I did to tune mine was to put it under full load with Prime95. Ran a custom profile on it and loaded the RAM to 7300MB, so 100% CPU and about 7.9GB of RAM loaded by the workers. Then I used ASUS Turbo V to to fine tune the vCore. Dropping vCore by .05v at a time and finding a sweet spot that was stable under load. I cold go down to 1.26v but Prime would crash after a few hours. The difference in temp from 1.26 to 1.28 is very significant. Spending time getting the vCore "perfect" under load was critical in getting my temps down.

After getting all my settings tuned in TurboV, I hard set those in the BIOS and have been stress testing again.
 
Maybe a really nice chip? Everywhere else is reporting crappy temps when voltage is given for speeds above 4.5ghz
 
Jorona said:
Pics or it didn't happen. That's right, I'm callin you out.

I still don't know why someone hasn't tried indigo extreme under the lid. That stuff should perform really well in this situation, I'd imagine.
Click to expand...


Indigo Xtreme is pretty much tailored to the exact contact size, liquifies, and the little card confines the liquid to the proper dimensions so that you don't have any run out and short your equipment.

You'd need to develop the proper masking card and volume or risk destroying your chip when the very conductive liquid ran out. You can't seal the IHS down to avoid this happening, because you have to liquify it with a poor mount, and then apply more pressure, so the starting state is a really awful contact (until either your spring mount pulls it down, or you tighten your cooler more).

It's doable, but logistically, it would require some development to get right.
 
Forceman said:
Most of the tests with direct die contact have shown less of a temp decrease than his testing, so I would say he might have the right idea after all. That collaboratory liquid is supposed to perform closer to solder than to TIM, so it makes the IHS work more like it would with a soldered die.
Click to expand...
The IHS attains a flatter, more even high-pressure against the die. That is what is so advantageous -- you have a shell that balances pressure evenly across the surface and provides a large enough outside surface area to apply even more pressure (the other way around would run increased risks of chipping the die and increased wear and tear on the body of the CPU from the outside environment).

It is a lot easier to get closer to an idea surface mount with the IHS than the base of something like a Prolimatech Megahalem, and I think it is because of the shape of the inside shape of the IHS (it covers the die, and then encloses the die with walls that meet a paste around the die on the CPU's circuit board).
 
aronesz, after removing the IHS and seeing the gap, I have to disagree there. There is more space than is needed. It doesn't really sit on the die very well at all (at least mine didn't). It is also a totally unfinished surface, just like the top of the heat spreader.
 
I am not so sure that Real Temp and Core Temp are correctly reporting CPU temperature values. I have seen these deliver some results that are just WRONG.

I just loaded up Real Temp, as I have not used it lately, and it shows one core at 48C and another at 83C under full load. Well, we all know that aint right. Core Temp gives higher results at idle than it does load, so I have been using the TurboV to monitor my CPU temp as it seems to give proper results.

If I am missing something here, let me know.

EDIT: Oops, my bad on that, one thread had just crashed after about 4 hours or so, hence the change in temp on the one core.

Showing differences in ASUS monitoring and Core Temp/Real temp of 15c to 20c.
 
Last edited:
Let me ask ASUS what they think about this.....will let you know.
 
Realtemp is demonstrably accurate at Tjmax.

Below that, it would be very hard to measure its accuracy as you can't really add your own thermistor to the CPU cores.
 
I'm shocked that you don't have a super badass full custom water cooling set-up Kyle.
 
fullvietFX said:
I'm shocked that you don't have a super badass full custom water cooling set-up Kyle.
Click to expand...

I don't need one.

I ran "super badass full custom water cooling setups" for years. The sealed H2O systems now are so good that I am willing to give up 100 or 200MHz for the security that the sealed systems offer. Also, I switch out my hardware enough that having a full loop through my system ends up being a PITA for me.
 
Here is exactly what ASUS had to say, and as I mentioned earlier, my personal testing experiences with Core Temp and its derivatives has given me reason to not trust it.

ASUS sez:

CoreTemp is in serious need of updating at this point for full compatibility with Ivy Bridge as is RealTemp (both latest releases do not include E1 stepping compatibility). Our TurboV is very good at monitoring the correct temps and voltages, but if you need third party verification we highly suggest the latest version of Aida64 Extreme (1938 beta or later) as they fully support Ivy Bridge and all current Z77 boards (not just from us). Besides working with the manufacturers, they also work directly with Intel to ensure proper readings and the stability test is actually quite good also.
Click to expand...

So I will give Aida64 Extreme a run and see if it gives me good numbers compared to TurboV.
 
Last edited:
Kyle, in regards to the low core, core 0 is beside the idle GPU. In mine, it always runs a full 10C cooler under load because it has its own silicon heatsink.

Not 35C cooler, but 10C
 
FWIW, I get concurrent temp reading between CoreTemp, Real Temp/GT, and Turbo-V on my Z77 ST. I think this is the first MoBo I've ever owned that has done that for me. :/
 
I have my stock 3770k running at 4600 1.32v with real temp report 76 max with H100. I did have 4700 but it would crash cinebench and temps were in low 80s running intel burn, so 58 at 4800 is mind boggling to me.

BTW I just assumed Kyle had a LN2 system with a LN2 tanker parked outside.
 
Last edited:
saitei said:
Was this using the 1939 beta or later? The default download is version 1900 which may not support the new chipsets.
Click to expand...


Good point and damn good call. I will edit out the post above so as not to confuse anyone.
 
I built up an Intel DZ77GA-70K MB and Core i7 3770K system several weeks ago, and shortly after firing it up, have noticed a sort of Burned Smell from it ever since.
Have also gotten a couple of "Processor Over Temperature" warnings, from Intel Desktop Utilities, however, that has gone away by itself very shortly each time?
IDU shows a processor Temp of 42 to 44 Degrees C. normally?
Wonder whether Intel might have not properly prepared some, or all, of the Chips when they prepared them?
Hasn't been enough of a deal to think about re-doing the chip, so far?
 
How about SpeedFan? Does the beta of SpeedFan report the proper temps? (Don't have an Ivy Bridge machine to test this out, but I'm curious none-the-less)
 
Looks like AIDA, Realtemp, and ET6 are matching. I'm using the latest ADIA to test.

 
Last edited:
Here is the beta Aida that Asus said to try.

http://www.aida64.com/downloads/aida64extremebuild1947zvrd4ybg3hzip

More ASUS feedback on what is being monitored as well.

"CPU" temperature is read from the onboard sensor chip, so it shows the processor socket and/or diode temperature measured by the on-board sensor chips. Individual core temperatures, "CPU Package" , CPU IA Cores, and CPU GT Cores are all measured by the CPU itself, using its integrated (on-die) temperature diodes. Of course the on-die diodes will show more accurate internal temperatures, since they're the ones that can follow the heating up or cooling down of the processor the quickest (instantly) while the on-board sensor will generally show a more level monitoring path and results based on the polling times and location. Both will end up showing almost identical numbers over a measurement period but the on-die is the most accurate for instant temps and what is occurring as you change voltages real time or cooling changes real time.
Click to expand...
 
It looks like Aida and RealTemp are both showing the same temps in that shot. It doesn't seem like there would be much for RealTemp/CoreTemp to be doing wrong - as I understand it they are just reading out the value of some register, so unless Intel changed the data that is encoded in that register in some way they should be accurate.

Edit: So going by that Asus comment, the individual core temps from Aida should match the RealTemp/CoreTemp temps, since they are reading the same sensor. The "CPU" temp would be something different (and likely lower) - which matches what bottjeremy's screenshot shows.
 
You must log in or register to reply here.
Back
Top