Delidded my i7-3770K, loaded temperatures drop by 20°C at 4.7GHz

Page 31 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SickBeast

Lifer
Jul 21, 2000
14,377
18
81
What I don't understand is, if the OP has proven that this lowers temps so dramatically, why doesn't Intel release a new revision that fixes this problem?
 

LV3

Member
Nov 30, 2011
34
1
71
Its only a problem for those who want to overclock right? Why would Intel spend money changing the production process for the relatively few who overclock. The return probably wouldn't justify the cost.
 

GPz1100

Senior member
Jun 10, 2001
354
3
81
Chances are, this production process is *cheaper* than the old process. The overclocking/enthusiast community is small compared to the masses/OE's that buy their cpus. So what if temps reach incredible levels during overclock state. Under normal operating conditions, temperatures are well within thermal limits. No clue on how much money this saves intel, but no doubt it does, otherwise the manufacturing process would of been left as it was with SB. All about the bottom line/dollar.

What intel SHOULD do is offer the cpu without the heat spreader altogether. That would save them a few more dollars. Maybe offer it as OE with a shorter warranty.
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
Chances are, this production process is *cheaper* than the old process. The overclocking/enthusiast community is small compared to the masses/OE's that buy their cpus. So what if temps reach incredible levels during overclock state. Under normal operating conditions, temperatures are well within thermal limits. No clue on how much money this saves intel, but no doubt it does, otherwise the manufacturing process would of been left as it was with SB. All about the bottom line/dollar.

What intel SHOULD do is offer the cpu without the heat spreader altogether. That would save them a few more dollars. Maybe offer it as OE with a shorter warranty.
Or charge $100 more, bundled with their CPU warranty.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
What I don't understand is, if the OP has proven that this lowers temps so dramatically, why doesn't Intel release a new revision that fixes this problem?
It doesn't really pay dividends in terms of reducing the loaded temperatures until you start OC'ing the CPU above 4 GHz.



Sure it can reduce temperatures by 10C at nominal clocks (3.5GHz) but the 10C reduction isn't really needed at that clockspeed anyways.

Whether your chip is running at 65C or 55C it isn't a big deal when only clocked at 3.5GHz stock.

But if you are a K-chip owner and are looking to OC above 4.2GHz then it really starts to make a difference in terms of power-consumption and required voltage (cpu lifespan) if you delid and get your temperatures down.

But we are talking about how many people here? A few tens of thousands maybe out of a market that Intel is selling hundreds of millions of CPUs into?

I'm just glad that Intel and AMD bother to even make multiplier unlocked cpu's available to us for less than the $1k extreme CPU pricing tier. So what if we have to put in a little elbow grease and pop the IHS off to squeeze even more out of them. At least we have the option to do so if we so wish.

 
Last edited:

GPz1100

Senior member
Jun 10, 2001
354
3
81
@Yuri, any updates on the condition of your cpu since applying the liquid ultra directly to the die?
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
Coming soon to a thread near you :p



^ bare-die (no IHS) testing with all these different TIMs :twisted:...will probably make a new thread for the bare-die tests though because this one is getting rather long and there will be lots of pics for the new thread :p
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
Coming soon to a thread near you :p



^ bare-die (no IHS) testing with all these different TIMs :twisted:...will probably make a new thread for the bare-die tests though because this one is getting rather long and there will be lots of pics for the new thread :p
Be sure to place a link in this thread to your new thread.
 

congoharrison

Junior Member
Nov 22, 2012
2
0
0
Hi all,

My first post here. So, please excuse me butting into this thread, but there is obviously a lot of talent here and maybe you guys can give me some quick guidance on an embarrasing matter.

I read this great thread with interest after designing/building/installing a gaming system for a friend in another country, monitoring the build remotely from Australia, so, I'm not "hands-on" per se.

I prematurely chose a 3770k for him, not knowing about the chip's heat issue until he tried playing BF3 with us online, and the resultant bios temp alarms alerted us shortly afterwards.

He's not an overclocker, but the GA-Z77 DS3H auto overclocked the CPU at default settings, upping the Vcore slightly in the process, and he immediately had at least one core @ 89*C running prime95, small FTT's, 8 threads running, with Aida64 reporting. Task manager showed all 4 cores @ 100%. Mobos "Turbo" set cpu to 3.8-3.9ghz or so, and the Vcore was then noted 1.1xx odd volts IIRC.
I aborted the test quickly, so no idea how high the temps would rise.

At first I was alarmed by the high temps, I've been out of the building loop for a few years and didn't realise core temps were ok over 70*C for this cpu, so we tried finding a solution and ended up here.

After considering this thread, we turned off the cpu's Turbo mode in bios and locked fan to 100% speed of +-2500rpm. The fan was only reaching 1800rpm at stock bios settings. I'll probably use the slope pwm later to slow it down when it's not under load, but guiding my friend remotely through an unfamiliar bios is painful to say the least.

He then obtained some decent TIM and re-seated the HSF.

Stock supplied Intel HSF used in cold environment with decent case ventilation. Removing the side covers netted a 3-4*C reduction in temps with a desk fan blowing across the opened case.

CPU now maxes @ 3.5ghz, priming for 1 hour stable (test conditions as above except for reduced CPU speed and resultant auto drop in Vcore) with temps only spiking to 79*C, probably averaging 76-78 with minimal core differences, and Aida64 reporting Vcore @ 1.080v and 52watts, well under the nominal 77w.

He's happy to run at that speed with no "Turbo" overclock which was taking it up to 3.8-3.9ghz or so as we originally observed, and the Vcore was then noted 1.1xx odd volts IIRC.

Temps still seem high to me, especially considering:
a) the cold environment the PC is in, (at the moment it's winter in a cold room)
b) Present max setting of 3.5ghz could accidently be reset to stock auto-overclock and he'll also get the slow default fan profile restored.
c) the present low Vcore. (Concerns me because of limited overhead?)
d) dust buildup through neglect

I'm a little worried that as next summer approaches he's going to run into trouble with temps. Also it's a shame he can't just leave the default "turbo" setting switched on in bios for a nice little "gamer's overclock", without risking a BBQ in his PC.

I showed him this thread, and he's not keen to attack his 3770k at all. I can't blame him really, (I want one to dissect!). His supplier at Overclocker's UK offered to replace or refund if his cpu (tested by them), was found to exhibit overheat symptoms, though in my experience, I've seen RMA tech's using platform games to test for overheat and I don't like his chances of them finding a problem, so he'll likely get stuck with a big testing bill if my prediction is correct, and still have the same cpu to take home :p

I'd recommend a big cooler, but if he's got the underlying CPU>IHS issue, it still may not provide a cure.

What do you guys think?
 
Last edited:

GPz1100

Senior member
Jun 10, 2001
354
3
81
^^Unfortunately, no matter how good your cooling solution, if the heat can't make it to the cooler, your temps will remain high.

If his games load up the cpu such at it operates at the upper extremes, then there is only one solution, not unless he wants to go with a sandy bridge cpu. I suppose trying a better cooler might make some difference as anything is more efficient at thermal dissipation than the stock heat sink. However, the underlying issue is still present - getting that heat to the heat spreader.

Going heat spreaderless is not for the faint of heart. It does yield the best results though.

Perhaps it makes sense to better examine the features of the board, disable the automatic overclock, leaving only the default turbo mode as intended by intel - when needed, 1 core will top out at 3.9ghz, not 4 cores. That should keep the temps in check.
 

hokies83

Senior member
Oct 3, 2010
837
2
76
Coming soon to a thread near you :p



^ bare-die (no IHS) testing with all these different TIMs :twisted:...will probably make a new thread for the bare-die tests though because this one is getting rather long and there will be lots of pics for the new thread :p
I have most of those...

Liquid Ultra going to win by a Land slide ;)
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
Welcome to the forums congoharrison :thumbsup:

I'd recommend a big cooler, but if he's got the underlying CPU>IHS issue, it still may not provide a cure.

What do you guys think?
The big cooler will still help even if you don't lap the CPU's IHS or mess around with delidding and all that.



^ these results are for my 3770k when it was still stock (before I started messing with it).

Replacing the stock HSF results in substantial temperature reductions as well as enabling higher OC clockspeeds.

It is true that delidding serves to lower the temperatures even further still, as well as increasing the highest OCs, but you don't need to go to those lengths to benefit from using a 3rd party HSF on your Ivy Bridge CPU.

All the best and hope this helps you resolve you and your friend's conundrum :thumbsup:

(PS - in case you were wondering, the results in the table show what happens if you use a Noctua NH-D14, but if you instead use the more expensive Corsair H100 then the temperatures decrease an additional 3-4°C over that of the NH-D14)
 

congoharrison

Junior Member
Nov 22, 2012
2
0
0
Thanks for the prompt replies, you're champions!

I'll pass this info onto my friend so he can make an informed choice on his options.

I drew this for him last night.... attempting a psuedo explanation in semi-scale .......




Oh, and you can call me congo, (or whatever, I'm thick skinned), the registration applet wouldn't let me use it though :)

Disclaimer: Do NOT cut on the dotted line !
 
Last edited:

chanfo

Senior member
Oct 16, 1999
538
1
76
I delidded my i5-3570k and wanted to see if shimming made a difference. I'm running overclocked at 4.4ghz, 1.18v. My temps running IBT at Very High setting:

Before delidding: 76, 88, 86, 82
After delidding (no shim): 64, 77, 75, 69
with single sheet of 24lb paper shim: 76, 93, 90, 84
after removing the shim: 64, 73, 74, 68

TIM used was Tuniq TX2. Heatsink was Hyper 212+. My tests supports the theory that the gap is the major contributing factor of the poor temps of the CPU.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
I drew this for him last night.... attempting a psuedo explanation in semi-scale .......

Oh, and you can call me congo, (or whatever, I'm thick skinned), the registration applet wouldn't let me use it though :)

Disclaimer: Do NOT cut on the dotted line !
Ha! Good drawing Congo :thumbsup: Do not cut indeed :D

I delidded my i5-3570k and wanted to see if shimming made a difference. I'm running overclocked at 4.4ghz, 1.18v. My temps running IBT at Very High setting:

Before delidding: 76, 88, 86, 82
After delidding (no shim): 64, 77, 75, 69
with single sheet of 24lb paper shim: 76, 93, 90, 84
after removing the shim: 64, 73, 74, 68

TIM used was Tuniq TX2. Heatsink was Hyper 212+. My tests supports the theory that the gap is the major contributing factor of the poor temps of the CPU.
Wow, that is pretty awesome confirmation result, thanks for posting it up. Glad to see your results :thumbsup: Your temps look great too!
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
@Yuri, any updates on the condition of your cpu since applying the liquid ultra directly to the die?
Today marks 1 month since I mounted with CL Ultra.

When I pulled the waterblock off, the CPU didn't even stick to it. However, there was a "residue" on both the die and the base of my waterblock - a sort of lumpy metal "crud". It seems to me that if I worked at it long enough with a cloth I might get the stain off of the die, but it was *not* coming off the block.





I made an oops though. Running on 2 hours of sleep today, I didn't have the presence of mind to try heat and instead pulled out a sharp razor and scraped the crud off. I have a hunch that if I had used a lighter or hot surface it would've melted and wiped off easily.



^ In that image, the base is completely smooth to the touch, that's stain from the TIM. The integrity of the copper doesn't seem to be compromised, and I'd say the image makes it look worse than I feel about it.

Room temperature is lower now than when I applied the CL Ultra, definitely low enough that pure gallium should solidify at room temperature. It looks as though the CL Ultra had started to separate, which might lead to the gallium solidifying in spots. (CL Ultra is most likely an alloy of gallium, indium and tin, which has the interesting property of having a lower melting point than any of the individual elements)

Afaik, none of these elements should react with pure copper.

That aside, temperatures are ~3c lower now that I've remounted, but they were still fantastically below Phobya HeGrease even before my remount.
 
Last edited:

GPz1100

Senior member
Jun 10, 2001
354
3
81
Yuri, thanks for the follow up. Looks like you used alot more than I did, as my residue on the block mirrored the profile of the die only. The second pic above, is that before or after cleaning?

I hope I don't have to remove this contraption for some time now. When I do, I'll probably run the system a while to heat things up, then remove it, just in case.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
In the second picture I had already briefly attempted to remove the TIM by wiping it a few times with a paper towel - I succeeded in smearing it around and little more. The third picture is what it looked like after scraping it with a razor and vigorous rubbing with alcohol.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
Holy batman, Robin :eek:

Now that I see your water-block's mounting flange I can truly appreciate why you are so keen on having springs on your mounting bolts to better distribute the mounting pressure across the die and socket ;)

Your flange, the silver part surrounding the copper block, looks like a hefty solid peice of metal that probably doesn't flex or bend one iota when you you the block.



The H100 is quite a different beast in that regard. It just has these four thin and narrow strips of metal that extend out to the mounting bolts, and they bend and flex (just like a spring) as needed to provided the down-pressure on the socket and die.

Springs wouldn't hurt, but with the bendy flexible metal mounting bracket already present it doesn't seem to be imperative to include additional springs to the mix. But with your water block I can see why there probably is simply no choice in the matter.
 

GPz1100

Senior member
Jun 10, 2001
354
3
81
This is the block i'm using by xspc. It's actually a modular 3 piece deal.

From top to bottom,

1) relatively thin metal X plate which sits in the recess of the acetal (thermoplastic) hold down
2) Hard acetal holdon
3) actual block w/copper base.






Over all, it's a very beefy setup, with no flex in the hold down when tightening down the screws. Not sure how it compares to Yuri's block several posts up, but for this application, it's going nowhere.

@Yuri, what block are you using?
 

dqniel

Senior member
Mar 13, 2004
650
0
76
I haven't tested my load temps in quite some time. I'll need to test and update my results. Possibly tomorrow.
 

aigomorla

Cases and Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
19,421
1,761
126
very nice article IDC.

it shows the raw performance of each tim...

however u missing my favorate which is PK-1.


Just a note to you guys... u can not take this review as a bible...
the tims were very close to each other on a test bed which is difficult to reproduce unless u delid your cpu as well.

Also taking the delta ambient, would give us a better test bed..
Because since the temps were so close to each other, a sneeze moving warm air to your bed would effect the results enough to cause confusion about them.
(keeping ambients the same on all my tests was the biggest PITFA i had to deal with.)
(also in watercooling reviews IDC, you know how much even a fraction of a C can make hence why everyone now just takes delta ambient, because at the same heat source, the temps will scale.
Example.. 200W heat source constant.. will raise cooling block by 5C.... that means at 7C -> 12C and at 10C->15C the delta C ambient is always 5C but in reporting value... its a very big number between 12C and 15C.)

basically u take the ambient ur at... then subtract it by the load temps, to give u a delta ambient.

Finally... when the cpu is pushed at those different clocks, the heat factor can scale up by a multi and not a linear scale...

(sorry this is my many years as a tester speaking for more IDC... dont take offense to it.. )

In all, great review tho.


Last note... those of u guys who are on waterblocks.
99.999% of the waterblocks u buy now are bowed. This means that it has a slight convex to put more mounting pressure in the center of the ihs, and delidding your cpu on a stock block will spell DISASTER.

make sure you do a flat test on your block / heat sink b4 u pop a delided cpu under it.
Use a gridded paper and see if all the lines are absolutely straight.
If you see a slight bow, your better off NOT delidding your IHS.
Sanding a bowed waterblock flat will also spell disaster for the bock..
Once bowed, its very difficult to get it stay flat and sanding it flat could weaken the thin layer and cause block to prematurely fail.

Dont make my many lost cpu's not count in warning you just how fragile these guys are delided.
 
Last edited:

GPz1100

Senior member
Jun 10, 2001
354
3
81
@aigomorla, Please clarify and expand on, either in this thread or elsewhere, what exactly you mean by disaster, and how exactly you're killing your cpu's. If temperatures are within acceptable limits (say < 70C), while under extreme loads, there must be other factors involved.
 

dma0991

Platinum Member
Mar 17, 2011
2,723
1
0
@aigomorla, Please clarify and expand on, either in this thread or elsewhere, what exactly you mean by disaster, and how exactly you're killing your cpu's.
He meant that you're flexing the die with a curved surface. Its like glass, flex it a little and it breaks.
 

ASK THE COMMUNITY