Have you de-lidded an Ivy Bridge CPU?

[leeleatherwood]

lower GFlops in IBT is 2 fold.
1. Hyperthreading does have a negative impact... The real cores are overloaded and can't handle all the instructions.
2. Like this guy, he only tested with 1024MB ram. You have to test _ALL_ memory in order to get highest performance AND truly test stability.

We've been over this before Sonda!!!

This^^

Other things I have noticed that affect GFLops in IBT.

Dual channel memory vs single channel (about 10 GFlops for me)
Ram speed, by going from 1600mhz to 2133, I gained another ~5 GFlops.

 

[SonDa5]

lower GFlops in IBT is 2 fold.
1. Hyperthreading does have a negative impact... The real cores are overloaded and can't handle all the instructions.
2. Like this guy, he only tested with 1024MB ram. You have to test _ALL_ memory in order to get highest performance AND truly test stability.

We've been over this before Sonda!!!

Funny you show up and see my post which I used standard in. I haven't used standard in a long time. The only reason I used standard was to match ccityinstaller settings.

I normally use the MSconfig boot to debug/ 4 cores / maximum memory for testing.

 

[CoreStoffer]

Killed my IB trying to de-lid the sucker. I thought I was careful enough, but no.

http://hardforum.com/showpost.php?p=1039231213&postcount=77

 

[SonDa5]

Killed my IB trying to de-lid the sucker. I thought I was careful enough, but no.

http://hardforum.com/showpost.php?p=1039231213&postcount=77

That sucks man. Thanks for sharing the info that killed it off. What type of razer did you use? Did you do corners first then sides? Please describe more details on the hand placement of blade/or tool with blade and order of areas cut first. Thanks.





Hope the stepping and IHS improves on these so that de-lidding doesn't become such a desperate action to fix the high temps.

Please post back when you get the replacement IB to see if temps are better with newer batch.

 

[chronicfx]

I saw that one earlier. Sorry coreStoffer

I believe he said it was done trying to get the corner of the razor blade in first instead of using the middle of the razor blade.

 

[CoreStoffer]

I saw that one earlier. Sorry coreStoffer

I believe he said it was done trying to get the corner of the razor blade in first instead of using the middle of the razor blade.

Well, I cut into the corner of the IHS with the middle of the razor blade, as I should, but as I wiggled the blade towards the middle part of the IHS, I think I allowed the corner of the blade to cut into the glue. Not good. The corner of the blade can scratch into the surface if it goes a bit "anarchist" on you.

Keep those corners of the razor blade out in the open, and I think that you will be good.

 

[ccityinstaller]

Well, I cut into the corner of the IHS with the middle of the razor blade, as I should, but as I wiggled the blade towards the middle part of the IHS, I think I allowed the corner of the blade to cut into the glue. Not good. The corner of the blade can scratch into the surface if it goes a bit "anarchist" on you.

Keep those corners of the razor blade out in the open, and I think that you will be good.

I really wish you had followed the very well written guide using a small hammer to control things that IDC posted @ Anandtech..I linked to it for a reason, and that was to prevent this from happening as much as possible to my [H] brothers...I am sorry you killed the chip..

IC Diamond is designed to be placed in the center of IHS like a pea size ball or like a grain of rice then pushed down. IC Diamond is works best with direct pressure according to IC testing. Mine is probably a little too much but with proper pressure the die is completely covered and good contact pressure results in good temps. I really think my application is working out. I've read by many that spreading IC Diamond isn't optimum because it can lead to air bubbles in TIM while using a drop or grain rice application will spread evenly with pressure when mounting and works the best.

While the method you mentioned is ideal when using a large surface (IHS) and a heatsink/waterblock, spreading it was what IC recommended I try since I told them I was using it directly on the die of the cpu...

I have a question for you, since your sig states that you are WC'ing both the 3570K and your 7950 but you are only listing a single 120mm rad?? Am I correct or incorrect to assume the data you have posted in this thread is the result of the cpu and gpu only being cooled by that single red 120mm rad? Or did you get your new 360mm rad in and not update the sig?

If you are honestly only using that single 120, I am going to want to toss this damn system out the window, since I am using 420mm of rad space with a Swifttech 655B set on high with Gentle Typhoon AP-15s running @ 12V...

 

[schmuckley]

I voted #1 because I got 6c lower
Granted I was using AS5 and didn't give it cure time;I'm sure results would have been better with
that Coollaboratory liquid pro.
To anyone considering using IC diamond..That's not a good idea..It has tiny bits of diamond in it and has been known to scratch IHSes..It will scratch bare silicon,too.

 

[ITSTHINKING]

If this bullshit continues with Haswell I'm going to buy AMD. Congrats on fixing Intel's retarded change.

Laughed hard QFT

 

[plur]

I guess this is to be avoided on Bloomfield?

 

[CoreStoffer]

Please post back when you get the replacement IB to see if temps are better with newer batch.

Got the replacement IB installed and up and running no problems. Phew, I was kinda worried that some sort of short circiut might have been done on the mobo with the faulty CPU.

My new CPU is a batch from week 25 and seems to be pretty fine. My overclock is a mild 4,2 Ghz (plain x 42 Multiplier) and max temps with Prime95 27.7 hits up to 63/66/65/64 Celsius. IntelBurn Stress test hits 64/68/67/66. The old one would hit 76 Celsius on the hottest core.

Good enough for me!

Addition info:All temps is with the case fans in the FT02 running at low and only one CPU fan. I like it nice and quiet. I'm running a Vcore offset of -0.050 and that gets me a Vcore of 1.176-1.184 with IntelBurn and 1.192-1,2 with Prime95.

 

[SonDa5]

I have a question for you, since your sig states that you are WC'ing both the 3570K and your 7950 but you are only listing a single 120mm rad?? Am I correct or incorrect to assume the data you have posted in this thread is the result of the cpu and gpu only being cooled by that single red 120mm rad? Or did you get your new 360mm rad in and not update the sig?

If you are honestly only using that single 120, I am going to want to toss this damn system out the window, since I am using 420mm of rad space with a Swifttech 655B set on high with Gentle Typhoon AP-15s running @ 12V...

My sig is correct and all of my testing with my de-lidding experiments so far have been done using the single HWLABS slim stealth 120mm radiator with a Triebwerk 123 fan blowing full RPM when testing.


The thing is that I really didn't plan on running my loop like this but I sold my HWlabs stealth 360mm radiator that I was running in my loop to get some money to buy a Alphacool UT 60 420. I have been waiting almost 3 weeks to get the UT 60 420 since it was out of stock at Frozen.CPU. Good news is that it finally shipped so I should have the 120mm radiator out of my loop and the new UT60 420 in by this coming weekend.



This is my loop flow right now with the 120mm radiator:

Pump to 120 rAD t0 GPU block to cpu block to reservoir back to pump.


I did have to run default speeds and under volt my HD7950 as a precaution for nice temps during gaming.


Once I get my loop upgrades fixed I'm hoping to lower my CPU temps even more.

 

[SonDa5]

I voted #1 because I got 6c lower
Granted I was using AS5 and didn't give it cure time;I'm sure results would have been better with
that Coollaboratory liquid pro.
To anyone considering using IC diamond..That's not a good idea..It has tiny bits of diamond in it and has been known to scratch IHSes..It will scratch bare silicon,too.

You have to be careful when applying it and removing it. I will post pics of my die condition when I remove the IC Diamond from it and apply Liquid Pro this weekend.

 

[CoreStoffer]

You have nothing to fear by De-Lidding the chip unless you are a total idiot(I kid I kid)..

http://www.youtube.com/watch?v=-IV2UuwHrU0

j/k

(I killed my CPU trying to remove the IHS)

 

[bmoney9]

I really should do this, however I can't risk destroying my CPU during the college semester. Then I wouldn't graduate lol

 

[SonDa5]

Got the replacement IB installed and up and running no problems. Phew, I was kinda worried that some sort of short circiut might have been done on the mobo with the faulty CPU.

My new CPU is a batch from week 25 and seems to be pretty fine. My overclock is a mild 4,2 Ghz (plain x 42 Multiplier) and max temps with Prime95 27.7 hits up to 63/66/65/64 Celsius. IntelBurn Stress test hits 64/68/67/66. The old one would hit 76 Celsius on the hottest core.

Good enough for me!

Addition info:All temps is with the case fans in the FT02 running at low and only one CPU fan. I like it nice and quiet. I'm running a Vcore offset of -0.050 and that gets me a Vcore of 1.176-1.184 with IntelBurn and 1.192-1,2 with Prime95.

Wonder if any changes to IHS were made? Good to see your new CPU is operating much cooler.

 

[mdlam]

I actually scratched mine down to the metal but for some reason, the sucker works!

 

[Blackbird]

If this bullshit continues with Haswell I'm going to buy AMD. Congrats on fixing Intel's retarded change.

I agree 100%, i might also buy AMD in the future if Intel continues with it.
I will never buy a Intel CPU which has TIM under the IHS.

 

[CoreStoffer]

Wonder if any changes to IHS were made? Good to see your new CPU is operating much cooler.

To be honest I just think it is about luck, because my theory is that the IHS mounting on the IB die is a sloppy job. A bit to much glue on the outer rim and the CPU die just doesn't get a proper contact with the IHS. The worlds best thermal paste can hardly fix that.

I just tried to up the speed to 4,4 GHz and lower the VCore offset to -0.020 and I'm hitting max 71 Celcius on Prime 95. So the new CPU it still runs cooler despite a higher clock and a VCore that went up to 1,224 where the old CPU was sweating 76 under 1.192 V.

I really like the IB after getting the new CPU, but it is a bit of a mess for overclockers frankly.

 

[chronicfx]

I agree totally. I bought my 3570k day 1 and it runs very cool for an ivy. I need to be over 1.4v to throttle it with IBT.

 

[ccityinstaller]

My sig is correct and all of my testing with my de-lidding experiments so far have been done using the single HWLABS slim stealth 120mm radiator with a Triebwerk 123 fan blowing full RPM when testing.


The thing is that I really didn't plan on running my loop like this but I sold my HWlabs stealth 360mm radiator that I was running in my loop to get some money to buy a Alphacool UT 60 420. I have been waiting almost 3 weeks to get the UT 60 420 since it was out of stock at Frozen.CPU. Good news is that it finally shipped so I should have the 120mm radiator out of my loop and the new UT60 420 in by this coming weekend.



This is my loop flow right now with the 120mm radiator:

I kinda figured that was what you were doing since I saw that photo the first time, but just wanted to be sure..You are lucky since the 7950s (and newer Nvidia 6XX) cards sip so little power @ idle..even with my GTX 480 running @ 50.Mhz/63.1Mhz core/memory @ idle, it is still dumping a fair amount of heat into my loop..

I finally got some updated pictures of my rig, I am going to edit them and get them up as soon as I finish this post and install Adobe PhotoShop Elements again..Never got around to it after formatting..

Despite my 4.8Ghz o/c being stable with 20+ passes of IBT with all memory @ Extreme mode, it would fail Prime95 within 10-30 minutes..I had to raise my Vcore a bit to 1.272V, and here is a screenshot of an 18 hour run:

Max temps were 70, 81, 80 and 78C, although average temps were 5~7C lower per core when averaging the data from the RealTemp log..

I am now working toward a stable o/c @ 4.9ghz, and will then see how temps look and see if 5Ghz is going to be obtainable with decent temps..

 

[chronicfx]

Nice voltage for 4.8 CC. I think you should have no problem with 4.9, hopefully the chip is willing, but it won't fail from temps.

Just a question, how come you stopped worker 1 & 4 early, is there a story behind that?

 

[scharfshutze009]

What a waste of time because stability is not guaranteed with overclocking according to Jean Andrews A Practical Guide to Managing and Maintaining Your PC 7th Edition. It also decreases the life of your components, which is also mentioned in that book.

 

[chronicfx]

You are more than welcome to your opinion and lower framerate.

 

[Hagrid]

What a waste of time because stability is not guaranteed with overclocking according to Jean Andrews A Practical Guide to Managing and Maintaining Your PC 7th Edition. It also decreases the life of your components, which is also mentioned in that book.

Sorry but I would rather waste time reading a good book then that crap.

 

[beiker44]

bad ass post IMO

 

[SonDa5]

Spent most of the night swapping out IC Diamond 24 TIM for Liquid Pro TIM on my bare die NO IHS 3570k.


I tried using some Thermal pad around the die on the PCB with the Liquid Pro but I couldn't get the right spacing and my CPU would not post.

So I removed the thermal pad and still would not post.

Messed around with my mount for a few hours and I kept getting error 55 on my MB LED display.

I finally got a snug fit and was able to boot up but my temps are still not that great. My 3rd core is hotter than it has ever been. So I am going to have to redo this mount after I get some rest later on this weekend.


So far from what I am seeing Liquid Pro requires a nearly perfect mount to work right while the IC Diamond was more forgiving on the mount.



Screen shots of my temps so far.

Before with IC Diamond.

26C Ambient

With Liquid Pro
25C Ambient

My 3rd core has never been so hot so I think I may have a bubble in that area. I'm going to let it run prime95 for a few hours to see if the temps settle down. Probably going to have to remount.

IC Diamond is leading right now.

 

[SonDa5]

Delidded IB i5-3570k @ 4.5GHZ 1.15v with NO IHS with DT SNIPER Water block
http://valid.canardpc.com/show_oc.php?id=2551498



TIM IC Diamond 24
IBT 4.5GHZ Maximum memory tested 5 passes
Ambient Temps 26C
Best Temps seen over a 2 week period using IC Diamond

TIM Liquid Pro
IBT 4.5GHZ Maximum memory tested 10 passes
Ambient Temps 25C
Fresh Install

With IC Diamond 24 average temp of each core was 73C under IBT load. With Liquid Pro average temp of each core is at 61C!! under IBT load.

Taking a look at my IB 3570k progress since delidding.

Before Delidding.

Ambient Temp 26C

TIM IC Diamond 24

Temps under LinX load 83C 93C 89C 84C average of 87.25C per core.


Since delidding I have been able to lower my average core temps under IBT 100% load at 4.5GHZ by 26.25C per core!!!

 

[Xz CJ zX]

I really feel like a dignified company like Intel that sells thousand dollar processors would be smarter than this. AND they should cover this under warranty so they don't have to muck around with a recall.

 

[chronicfx]

Damn that liquid pro kicked the IC diamond right in the nuts!

 

[SonDa5]

Damn that liquid pro kicked the IC diamond right in the nuts!

It's working great so far. Temps should drop some more with a little burn in period.

I didn't expect Liquid Pro to beat IC Diamond by so much.


Took about 5 mounts of Liquid Pro before I got the right thickness and contact pressure between the DIE and DT SNIPER water block. I'm not using an IHS so I think this is part of the reason the mount was so tricky but after all the effort the results are very satisfying.

 

[Woe]

It's working great so far. Temps should drop some more with a little burn in period.

I didn't expect Liquid Pro to beat IC Diamond by so much.


Took about 5 mounts of Liquid Pro before I got the right thickness and contact pressure between the DIE and DT SNIPER water block. I'm not using an IHS so I think this is part of the reason the mount was so tricky but after all the effort the results are very satisfying.

Great results, thanks a lot for sharing.

 

[SonDa5]

Just got done tweaking my delidded 3570k NO IHS mount.

I added Fujipoly Extreme .5 mm in thickness around the die.

DT SNIPER nice and snug.

Added some lock washers to my bolts on the fron side of the mother board to keep my DT SNIPEr mount bolts nice and firm when I put it back in the case. Also I am using a EK S115X TRUE back plate which works out great for this mount.

Still haven't booted. Once I get my loop back together I will see if it boots. I tried the Fujiploy Extreme last week with 1 mm thickness and the CPU would not post. So hopefully .5mm thickness on the sides will work out better.

 

[scharfshutze009]

Sorry but I would rather waste time reading a good book then that crap.

That book will get you an certification as an A+ certified technician, which could be the start of a great career and is a great certification to start with before moving on to Network+ and CCNA.

 

[leeleatherwood]

That book will get you an certification as an A+ certified technician, which could be the start of a great career and is a great certification to start with before moving on to Network+ and CCNA.

That book is a pile of crap.

 

[scharfshutze009]

That book is a pile of crap.

Yea right you right a better book then. It help me get my A+ certification as well. Your just mad because it doesn't tell you how break your computer by Overclocking it. By the way I have computer old enough to prove that overclocking is bad because they haven't been overclocked and have been running for up to 12 years without any problems. She tells you everything you need to know in that book and doesn't BS about Microsoft's history like Mike Myers. She even has a PHD, which makes here more credible than most if not all other authors. You are in fact wasting your time overclocking because you don't overclock a server or a computer used as a client for a business because it just causes problems no matter how stable you can get the thing to run because it will never be 100% stable. I could go on and on about how it shortens the life of the computer as well and how you waste money buying new hardware because your computer died prematurely due to overclocking. Fact is that the components of the computer were not meant to handle the clock speeds. Do what you want with your hardware, but I know from experience that overclocking can actually decrease performance in some ways and you should at least benchmark before and after overclocking to get a good idea of how much of an increase in performance you gain before overclocking everything right out of the box. Eventually, though you'll get the idea that it not worth the trouble to overclock and that its better just to have a completely stable computer that may only crash due to file corruption in the software do viruses and malware or similar files that corrupt the system, which is also part of the problem.

 

[SonDa5]

Bottom line if you want the best thermal performance possible from an IB CPU over clocked or not delidding will help.


Heat is the problem with the design of the Intel IHS and by lowering the heat the electrical performance will improve. This is a known fact in electrical theory. Electrical resistance is the enemy here and lowering temperature will defeat that enemy.

 

[Taxicab2112]

Does this also work with SB? I've been looking to get better temps out of my 2500k.

 

[SolidBladez]

Does this also work with SB? I've been looking to get better temps out of my 2500k.

Intel uses a fluxless solder with Sandy Bridge. You could try but more or likely will rip the die from the IHS.

 

[psyclist]

Yea right you right a better book then. It help me get my A+ certification as well. Your just mad because it doesn't tell you how break your computer by Overclocking it. By the way I have computer old enough to prove that overclocking is bad because they haven't been overclocked and have been running for up to 12 years without any problems. She tells you everything you need to know in that book and doesn't BS about Microsoft's history like Mike Myers. She even has a PHD, which makes here more credible than most if not all other authors. You are in fact wasting your time overclocking because you don't overclock a server or a computer used as a client for a business because it just causes problems no matter how stable you can get the thing to run because it will never be 100% stable. I could go on and on about how it shortens the life of the computer as well and how you waste money buying new hardware because your computer died prematurely due to overclocking. Fact is that the components of the computer were not meant to handle the clock speeds. Do what you want with your hardware, but I know from experience that overclocking can actually decrease performance in some ways and you should at least benchmark before and after overclocking to get a good idea of how much of an increase in performance you gain before overclocking everything right out of the box. Eventually, though you'll get the idea that it not worth the trouble to overclock and that its better just to have a completely stable computer that may only crash due to file corruption in the software do viruses and malware or similar files that corrupt the system, which is also part of the problem.

You sir are on the wrong site, SoftOCP is I think right up your alley. and why are you trolling these overclocking threads? Most of us only keep a computer 3-5 years anyways, and some switch out every year! Your preaching is falling on deaf ears here son. Take your A+ advice else where