3770k IHS Removals - CPU temp dropped from 79C to 71C

[roger_h99.nor]

Yea my old LGA-775 Danger Den waterblock still going : - )

Very nice. Btw, does the old 775 block fit the 1155 socket without any modifications?

[andressergio]

Got it, I used my fingernail.... thanks.

[Jonny5isalivetm]

hmm I will try different memory I hoped Ivy could at least take 2400.. Yes latest bios..

I used a scalpel then my nails too

the DangerDen block fits because my asrock mobo has combo cooler holes for older blocks :>

[andressergio]

hmm I will try different memory I hoped Ivy could at least take 2400.. Yes latest bios..

I used a scalpel then my nails too

the DangerDen block fits because my asrock mobo has combo cooler holes for older blocks :>

on my Maximus Formula 4 P67 B3 no problems at all with IVY and 2400 or more, maybe its the mobo or the IMC on your chip mate

Good luck !!! oh you are Irish !!! House Of Pain i luv it !!!

[Jonny5isalivetm]

JUMP JUMP JUMP JUMP JUMP

ok thx for that info il try different memory i think first and see what happens

[andressergio]

JUMP JUMP JUMP JUMP JUMP

ok thx for that info il try different memory i think first and see what happens

glad to help !
good luck
Sergio

[techguy2k5]

Dudes my memory is meant to be 2400mhz but any setting I try and it only boots at 2133mhz.. bad memory or CPU ?

Timings are too tight/voltage too low. That's what I've found with my 2133MHz RAM. Had to back CL down to 10 (from 9) but was able to bump clock speed up to 2200 at least.

[Ket]

for some reason i'm not buying that video..

looks like intel set someone up to do it and fail so that people are scared out of removing it themselves. Who would keep filming and upload a vid of them killing a cpu. bleh.. just dont know

Some people just do, not so much because they killed their CPU but because the video is "instructional". I'd hope anyone with half a brain would know not to use the hammer and club approach in that video anyway.

[graysky]

Just read through this thread cover-to-cover and now I have a hard-on to do this to my 3770K. I found a standard razor blade (0.009") did the trick in about 15 min. Running benchmarks now and will update when TIM is past the breakin period.

EDIT: See post #276 for results.

[MikelMolto]

Thank You for this Post! Very interesting.

You wrote: Data collection
I use a bash script I wrote to query lm-sensors every 2 sec. The resulting data are compiled into a log file.

Is there any way, that You would share, how You collected the data?

Thank You and regards.

[Leeghoofd]

I sense a flash back somewhere...

So how about ambient temperature, humidity ? Your 1.288 run was hotter than 1.296Vcore ? Are the variables controlled in your test environment ? Nice approach mate and a nice way of putting things, but this will be a thorn in the eyes to those that have shared their, maybe to simple, data...

Other thing 4500mhz is still around the sweetspot for Ivy, most CPU's can do this easily around 1.25Vcore, there's close to no need to remove the IHS, if that is your targeted daily speed... it's above that IHS removal and reappliance of the TIM can be the difference between maxing out at 4600-4700Mhz before removal. And finally gaining 100 to 200MHz with lower temps/Vcore afterwards.

Final outcome will depend on test variables, amount of TIM, cooler pressure on IHS, amount/type of TIM used for core contact, ambients,... Far too many variables to provide 100% accurate data. However looking forward to your test results... my goal is to gain 100-200MHz and/or remain below throttling temps. If it's 90 or 91°C during Prime95 AVX enabled is not that important...

Last but not least : each CPU is unique... you can tell me you gained 5°C, good for you but I might have gained 10°C...

[ZoLKoRn]

Finally i have time to try it with this subject today

Here we go !


Done !


i5 3570K retails Malaysia


GELID EXTREME instead (Thanks Gelid Solution for this)


No comment for behind the scenes with my camera stand xD


Tests results will coming up next, it on testing now

[graysky]

You wrote: Data collection
I use a bash script I wrote to query lm-sensors every 2 sec. The resulting data are compiled into a log file.

Is there any way, that You would share, how You collected the data?

Remember, I am using linux. Here is a link to the simple script. Note that you need to have lm-sensors installed and configured. Further, you have to tweak the script to match the output of your system. It can serve as a bare-bones example of what you need though.

So how about ambient temperature, humidity ? Your 1.288 run was hotter than 1.296Vcore ?

Right you are. I cannot control ambient temp beyond allowing the HVAC in my office to work. As I stated, ambient is 74-76 F. In practice, without an internal control to allow for a floating baseline, there is no easy way to control for changes in ambient temp. I think we can keep this in mind when looking at the before/after table I have yet to post. In other words, small changes of say, +/- 1C are probably not real.

To your question about the 1.288 vs. 1.296 vcore: remember that I am not actively controlling the vcore. The board is allowing it to fluctuate based on usage; they are not discrete runs manually defining a value. The reason why there is this apparent disconnect is due to linpack. There are times where is pushes the vcore up without having a full load on the processors. This will be obvious when I show the data split out by method, either mprime or linpack.

Other thing 4500mhz is still around the sweetspot for Ivy, most CPU's can do this easily around 1.25Vcore, there's close to no need to remove the IHS, if that is your targeted daily speed... it's above that IHS removal and reappliance of the TIM can be the difference between maxing out at 4600-4700Mhz before removal. And finally gaining 100 to 200MHz with lower temps/Vcore afterwards.

Final outcome will depend on test variables, amount of TIM, cooler pressure on IHS, amount/type of TIM used for core contact, ambients,... Far too many variables to provide 100% accurate data. However looking forward to your test results... my goal is to gain 100-200MHz and/or remain below throttling temps. If it's 90 or 91°C during Prime95 AVX enabled is not that important...

Last but not least : each CPU is unique... you can tell me you gained 5°C, good for you but I might have gained 10°C...

Agreed to all of your points here. The results are not definitive from a practical stand point, but they should be from a statistically relevant point.

[techguy2k5]

I've got a few tens of hours of Prime 95 and IBT runs under my belt on 2 different overclocked Ivy Bridges that have been de-lidded and had Coollaboratory Liquid Metal applied to both CPU die & IHS. This is over the course of the last 2 weeks. 1 3570k w/H100 @ 4.7GHz, another 3770k w/custom water @ 4.9GHz. Results are holding constant. How many hours would you need to see to believe the results? 23C improvement in one case, 32C in the other. First with same cooling, second with improved water cooling.

It's a simple matter of physics. The stock TIM is absolute garbage. Intel's fluxless solder previously used in SB and now in SB-E chips has a thermal conductivity rating something like 20-30x higher than the stock TIM used on IB. Liquid metal compounds are the next best thing with a thermal conductivity approximately half that of the aforementioned fluxless solder, though still somewhere in the range of 10x higher than competing TIMs.

People can disbelieve all they want to, fine with me. I'll run my Ivy @ 4.9-5.0GHz 24x7 and be happy, you do what you like.

Here's a couple shots on my 3770k, first @ 4.9, second @ 5.0:
82C Max temp on hottest core @ 4.9 in IBT maximum stress test

92C @ 5.0 I have Prime95 open in the background in the 2nd pic just to get the load clock speed in view in CPUZ & Core Temp

[graysky]

How many hours would you need to see to believe the results?
...
People can disbelieve all they want to, fine with me. I'll run my Ivy @ 4.9-5.0GHz 24x7 and be happy, you do what you like.

Excellent results. I don't mean to be condescending. I'm sure the diff you report is real. Think I might order a tube of Liquid Pro after seeing your results. I am only waiting because I know that AS5 has a breakin period and I wanted to give a fair comparison based on how long my stuff sat on the first time.

Do you have any advice as to the application of Liquid Pro?

[OldChap]

Got any screenshots of the 3770K with HT on? Curious to see what difference that makes

[techguy2k5]

Got any screenshots of the 3770K with HT on? Curious to see what difference that makes

That I do not. I'll see what I can do, though. I'm not terribly interested in HT since it isn't useful for my purposes. I run Microsoft Flight Simulator X and do video transcoding. FSX needs only raw CPU clockspeed * IPC, performance doesn't scale with thread count and transcoding speed simply is unmatched with QuickSync @ 1.6GHz on Ivy Bridge.

Excellent results. I don't mean to be condescending. I'm sure the diff you report is real. Think I might order a tube of Liquid Pro after seeing your results. I am only waiting because I know that AS5 has a breakin period and I wanted to give a fair comparison based on how long my stuff sat on the first time.

Do you have any advice as to the application of Liquid Pro?

Sorry if my tone was offensive. I went with Liquid Ultra, it's easier to spread and provides the same level of temp drops (from what I've heard). You can apply with the included paint brushes, though I went out and bought a few more of different sizes to make sure I could apply to the top of the IHS easily also, the included brushes are quite small. Temp drops with Liquid Ultra were immediate for myself and my friend that has used it on his 3570k. No increase yet either.

[graysky]

@techguy - I've read that removing the stuff once it has cured (like over months) is nightmarish. Have you any experience doing this? Made me pause I must say.

[techguy2k5]

@techguy - I've read that removing the stuff once it has cured (like over months) is nightmarish. Have you any experience doing this? Made me pause I must say.

No sir, I have not. I hope not to have to replace the TIM though. If I do, I'll let you know!

[RBS]

Hey guys what would be better?

Die + IHS

1. Liquid Ultra + Liquid Ultra
2. Liquid Pro + PK1
3. Liquid ultra + PK1
4. AS5 + AS5

[babalouj]

Liquid Ultra + Liquid Ultra would be best.

With that said, I am currently using Liquid Ultra + PK1 (with great results) because I like to change my waterblocks relatively frequently.

[Church]

I'd use liquid ultra + indigo xtreme. No problems if lu will diffuse between IHS and core, but indigo xtreme exists only in precut forms for cpus with IHS.

[graysky]

I read that once Liquid Pro or Liquid Ultra set on two metal surfaces for a long period of time, say 6-9 months, the two surfaces are extremely difficult to separate. In some cases, I read about CPUs being permanently affixed to heatsinks. Can you comment? Is this true?

[PcCI2iminal]

Finally i have time to try it with this subject today

Here we go !


Done !


i5 3570K retails Malaysia


GELID EXTREME instead (Thanks Gelid Solution for this)


No comment for behind the scenes with my camera stand xD


Tests results will coming up next, it on testing now

i like the last pic

[st0ned]

I read that once Liquid Pro or Liquid Ultra set on two metal surfaces for a long period of time, say 6-9 months, the two surfaces are extremely difficult to separate. In some cases, I read about CPUs being permanently affixed to heatsinks. Can you comment? Is this true?

I've been using phobya LM ( I'd say it would behave the same as Liquid Ultra/Pro ) thing is Die isn't made out of metal, so neither of this TIMs will even adhere to it, I've remounted my IHS more then enought times to confirm this . IHS it's another thing, it will get stained by these (nothing that a good polishing won't remove ).

Just to sum it up and by my experience I would say, use wtv LM paste you seem fit between IHS and Die, to do the best application you can 1st spread the LM paste on the IHS and then put it over the Die, trying to "paint" the Die with LM pastes it's gonna be a pain because they will "never" adhere to it. Then I would use the best TIM between IHS and CPU block, that being sad I would avoid using any of the before mentioned LM tims just because they will stain and over time "solder" your IHS and Block together, therefore I would go for IX as a "permanent solution" or PK-1 if you intend to do some remounts.

In my system I can pull higher 50's lower 60's with a handfull of waterblocks with my 3770k at 4.8Ghz and 1.36v. Using 1.42v and 4.9Ghz will bring me to high 60's but still verry verry under control. I hope I was usefull