Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

[bloodindark]

i just applied nt-h1 , line method , 58-64-64-58 , let's see how it develops , since mx-4 , as5 , CFIII , did the same , but in matter of days , temps went nuts...

 

[GreenChile]

I've been running delidded with CLU (not bare die) for nearly 1 year now with no drop in temps. This stuff works great!

 

[bloodindark]

I've been running delidded with CLU (not bare die) for nearly 1 year now with no drop in temps. This stuff works great!

you have clu on die and clu on ihs?

 

[Idontcare]

you have clu on die and clu on ihs?

Most folks who delid on these forums and on others put the CLU between the die and the underside of the IHS. Then use a traditional non-marring TIM such as NT-H1 or MX4 between the IHS and their HSF.

You can use CLU between the IHS and the HSF, and the thermal conductivity would be superior, but it tends to be a pain to clean off your HSF should you ever need to unmount and remount down the road.

 

[GreenChile]

you have clu on die and clu on ihs?

What IDC said. I use CLU between the die and IHS but MX4 between the IHS and waterblock.

 

[PliotronX]

Most folks who delid on these forums and on others put the CLU between the die and the underside of the IHS. Then use a traditional non-marring TIM such as NT-H1 or MX4 between the IHS and their HSF.

You can use CLU between the IHS and the HSF, and the thermal conductivity would be superior, but it tends to be a pain to clean off your HSF should you ever need to unmount and remount down the road.

They should call themselves TIM refurbers then, delidding is baredie These noobshields should be optional (hint, hint Intel please offer a "mobile" K series).

 

[bloodindark]

btw guys , do you think it may be something wrong with the cooler (hs) itself? problem with the pipes or something , from your experience?since at first i only saw 2-3 degrees between hyper 212 and silver arrow ib-extreme?

 

[Idontcare]

btw guys , do you think it may be something wrong with the cooler (hs) itself? problem with the pipes or something , from your experience?since at first i only saw 2-3 degrees between hyper 212 and silver arrow ib-extreme?

Too many variables in play to say you have a busted HSF. Generally speaking, broken HSF is extremely rare. Many more problems are far more likely - including poor mount, bad airflow, non-aggressive OC'ing, etc.

 

[Idontcare]

They should call themselves TIM refurbers then, delidding is baredie These noobshields should be optional (hint, hint Intel please offer a "mobile" K series).

Delidding is merely the process of taking the CPU package back to bare-die condition (sans IHS, sans stock die-to-IHS TIM)...what you do with it thereafter is something entirely unrelated to the term "delidding".

Usually people will be specific to say they "relidded" their CPU if they stuck the IHS back on, but it is not always clear. However it can usually be inferred based on the context of the conversation.

The noobshield doesn't bother me TBH, it keeps me from needing to modify the mount of my HSF. And removing the IHS and running bare-die doesn't improve the thermal conductivity enough to increase OC'ing speed or reduce Vcc by a significant amount anyways.

For my IB systems, I put the IHS back on after delidding simply for practical considerations. Don't have to mess around modifying the socket retention mechanism on the mobo or messing around with the mount for my HSF/WC.

 

[PliotronX]

Delidding is merely the process of taking the CPU package back to bare-die condition (sans IHS, sans stock die-to-IHS TIM)...what you do with it thereafter is something entirely unrelated to the term "delidding".

Usually people will be specific to say they "relidded" their CPU if they stuck the IHS back on, but it is not always clear. However it can usually be inferred based on the context of the conversation.

The noobshield doesn't bother me TBH, it keeps me from needing to modify the mount of my HSF. And removing the IHS and running bare-die doesn't improve the thermal conductivity enough to increase OC'ing speed or reduce Vcc by a significant amount anyways.

For my IB systems, I put the IHS back on after delidding simply for practical considerations. Don't have to mess around modifying the socket retention mechanism on the mobo or messing around with the mount for my HSF/WC.

Relidding is a good one. Baredie did improve conductivity for my Q6600. I had the IHS and TRUE lapped to 2000 grit which dropped temps but only after I blowtorched the soldered IHS off was it able to run stably at 3600 with lower voltage than 3400 with the noobshield. Whether or not it is worth it is up to the user but to get the most from a chip, noobshields impede speeds any way you slice it (aside from HDT heatsinks).

 

[bloodindark]

about my problem , the problem with the thermal paste persisted , so i decided to buy CLU , i've put back the retention socket too , and the ihs , between die-ihs i used CLU , beautifully spreaded , then nt-h1 between ihs and HS. it;s the second day and no increase in temp yet , i'm very satisfied , 58-61-61-56 IBT load, the temps are stable and 2-4 degrees better than fresh nt-h1 (if used on die). did anyone use CLU on a video card (evga acx 780ti)? please share the experience

 

[BonzaiDuck]

Most folks who delid on these forums and on others put the CLU between the die and the underside of the IHS. Then use a traditional non-marring TIM such as NT-H1 or MX4 between the IHS and their HSF.

You can use CLU between the IHS and the HSF, and the thermal conductivity would be superior, but it tends to be a pain to clean off your HSF should you ever need to unmount and remount down the road.

You may remember a couple months ago I had communicated the essentials of my brief dialog with an Innovation Cooling tech. He had told me they had a multitude of customers who applied the nano-diamond paste to a range of bare-die situations. He was adamant that the nano-diamond particles weren't going to damage the CPU die or lead to "copper migration."

You seem similarly inclined now, given your remark a few months earlier about applying CLU or ICD.

I would think the CLU application would be best between IHS and die, or as you say -- the gains from a permanently delidded die are not sufficient to warrant it, and replacing the IHS was convenient for other reasons (no need to modify the retention mechanism.)

If I were to choose a bare-die result, I'd be more inclined to use something like ICD. I'd think that CLU would make it harder to remove the HSF without risk of damage to the die -- per your other remarks about it.

But I'm still just an bystander with this. Apparently, nobody ever complained of any damage using CLU. I'll be more inclined to build a system with an "E" processor next time, to avoid all this fiddling over the TIM factor in CPU fabrication. UNLESS -- Intel stops making the "E" processors with solder . . .

 

[WhoBeDaPlaya]

Too many variables in play to say you have a busted HSF. Generally speaking, broken HSF is extremely rare. Many more problems are far more likely - including poor mount, bad airflow, non-aggressive OC'ing, etc.

Broken heatpipes do happen though. Was refurbing an old Thinkpad Z61T that was constantly overheating on light workloads.
Fan was clean, used fresh TIM, etc.

Did IR temp measurements on a couple of spots and the radiator end was ice cold, while the CPU side was poker hot.

 

[flapjacky]

Did a delid a couple days ago on my 4670K, got a 20c drop!
NT-H1 noctua paste with the nh-u12p heatsink directly on the die.

After first install it wouldnt boot/post but i figured it was because of too little pressure against the socket pins. So i removed a millimetre more from the plastic spacers by sanding it down. So about 2mm total did the trick.

My motherboard is actually a B85M-G from ASUS which recently got a new BIOS update that allows overclocking.

As long as i run constant manual voltage i can achieve 4,8Ghz with 87c temps. Keep in mind that since i dont have fan splitter cables i can only run one cpu fan and one chassi fan. iirc the second cpu fan gives about 25% better cooling.

For some reason tough; When i run offset voltage to take advantage of power saving/automatic undervoltage, the CPU throttles after ~6 prime95 "steps".

There are so many power options im going mad trying to find out why it throttles. With one type of bios settings it would stay at 3,8Ghz when running prime95 but not go to any higher multiplier.

So i decided since i gotta run at constant voltage ill go for a low overclock @
Cores: 1.217V @ 4,4Ghz
Ring/north bridge: 1.252V @ 4,3Ghz (higher ring clock is highly unstable)
VccIN: 1.72

 

[MongGrel]

Have never delided one myself and probably won't in reality.

I've seen advantages here and there over the years people get off it, but probably never will personally.

 

[ehume]

As long as i run constant manual voltage i can achieve 4,8Ghz with 87c temps.
...
For some reason tough; When i run offset voltage to take advantage of power saving/automatic undervoltage, the CPU throttles after ~6 prime95 "steps".

There are so many power options im going mad trying to find out why it throttles. With one type of bios settings it would stay at 3,8Ghz when running prime95 but not go to any higher multiplier.

So i decided since i gotta run at constant voltage ill go for a low overclock @
Cores: 1.217V @ 4,4Ghz
Ring/north bridge: 1.252V @ 4,3Ghz (higher ring clock is highly unstable)
VccIN: 1.72

The advice we get in OC forums is to turn on the constant Voltage, as you are doing. When you engage offset Voltage you automatically engage "adaptive" Voltage, which allows the Voltage to overshoot, getting your chip too hot, which throttles it.

 

[fstime]

The advice we get in OC forums is to turn on the constant Voltage, as you are doing. When you engage offset Voltage you automatically engage "adaptive" Voltage, which allows the Voltage to overshoot, getting your chip too hot, which throttles it.

I have run -0.035 or -0.030 on my 4.5 GHz Sandy for years now. I don't get what you mean by "offshoot." If the voltage is to high at 100% load, you lower the offset some more.

1.312V at 4.5 GHz and 60C max temps gaming, 70C max temp prime95 and all the way back down to 1.00V at 1.6 GHz 35C idle.

 

[crashtech]

I think I ended up running a bit of negative offset on my 4770K. Seemed to me that with certain settings the voltage would just go wild, it was hard for me to believe an automatic setting could be so potentially destructive.

 

[ehume]

I have run -0.035 or -0.030 on my 4.5 GHz Sandy for years now. I don't get what you mean by "offshoot." If the voltage is to high at 100% load, you lower the offset some more.

1.312V at 4.5 GHz and 60C max temps gaming, 70C max temp prime95 and all the way back down to 1.00V at 1.6 GHz 35C idle.

"Adaptive" Voltage was introduced for Haswell. You don't have it on SB. I used the feature when I was doing my trophy run at 5GHz.

 

[bononos]

You may remember a couple months ago I had communicated the essentials of my brief dialog with an Innovation Cooling tech. He had told me they had a multitude of customers who applied the nano-diamond paste to a range of bare-die situations. He was adamant that the nano-diamond particles weren't going to damage the CPU die or lead to "copper migration."

You seem similarly inclined now, given your remark a few months earlier about applying CLU or ICD.
.......

ICD's product causes scratching, bad enough to scrub the markings on the heatspreader making rma's impossible. The company rep on several forums defended the product by (deceptively) saying it couldn't happen with the materials in the product or it was just some staining. This article also reveals that the troublemaker has sued BB and Coolermaster.

I remembered there was some issue with CLU and aluminium but not with copper.

 

[crashtech]

This article also reveals that the troublemaker has sued BB and Coolermaster.

From the article:

We are not Yahoos or hillbilly’s

Could have fooled me! I'm sorry, but if you are going to post in a forum as a professional manufacturer of quality goods, you just might want to have someone proofread your post before revealing your lack of education to the world.

/grammar nazi /spelling nazi

 

[bononos]

From the article:
Could have fooled me! I'm sorry, but if you are going to post in a forum as a professional manufacturer of quality goods, you just might want to have someone proofread your post before revealing your lack of education to the world.

/grammar nazi /spelling nazi

I wouldn't nickpit on the grammer/spelling. Its the attitude and effort to deceive which is far more telling on whether people should buy ICD. I mean the guy paid for a picture analysis to show that pics were 'doctored' and could not be relied on to prove that hs scratching occurred.

 

[crashtech]

I wouldn't nickpit on the grammer/spelling. Its the attitude and effort to deceive which is far more telling on whether people should buy ICD. I mean the guy paid for a picture analysis to show that pics were 'doctored' and could not be relied on to prove that hs scratching occurred.

Diamond is one of the world's premier abrasives, the only question should be how much material it will remove, not IF material will be removed. The post by the manufacturer removed any credibility as to quantifying the damage, i.e, whether such abrading action is even capable of removing the etched ID placed by Intel.

 

[flapjacky]

Warning: using the noctua nh-d12 cooler with nt-h1 has the "pump out effect". After three days of continuous raised temperatures i checked and there was a spot on the die with almost no paste left. I believe this is because the heatsink mounts with only two screws which allows it to "wiggle" back and forth as the paste contracts/expands.

Can anyone recommend a TIM that is good against this type of thing? IC Diamond sounds good except it scratches the surface, what if i need to replace this TIM too?
The cheramicué one sounds interesting if it hardens after time and doesnt pump out. (atleast not in three days)
The liquid ultra is conductive and i got capacitors right next to the die. so thats not a good choice

 

[YBS1]

The liquid ultra is conductive and i got capacitors right next to the die. so thats not a good choice

CLU doesn't suffer from pump out and it's basically "painted on". If your components are subjected to enough force for it to move from the place it is applied, it's electrical conductivity will be the least of your problems. You just have to use a steady and patient hand when applying it near any place you don't want it.