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

Page 35 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Welcome to the forums Jalockin :thumbsup:

awesome article ! Did you try to go over the 4.9ghz mark on the 3770K after delidding? I still have doubt about which one to go for 2600K/2700K or 3770K..... all in all isn't the only benefit of the 3770K a better build-in GPU ?

I went to 5GHz for benchmarking but it wasn't LinX stable.

First you really have to ask yourself would you notice the difference between a 5GHz SB and a 4.7GHz or 4.8GHz IB?

IB gets about 3% higher IPC than SB, so a 4.85GHz 3770k will perform the same as a 5GHz 26/2700k anyways.

But you really pay for that clockspeed, with either CPU, in terms of power consumption. If you can settle for the performance of a 4.5-4.6GHz IB then you'll get considerable power savings at full load.

I don't mind pushing my chips to their limits for the purposes of characterizing those limits, but I'm not so hard-up for that last 5-10% performance that I'd be willing to put up with the extra 100W power usage because of it.

Everyone feels differently about that of course.
 

Jalockin

Junior Member
Mar 20, 2013
5
0
0
Welcome to the forums Jalockin :thumbsup:
Thank you ;)


I went to 5GHz for benchmarking but it wasn't LinX stable.

First you really have to ask yourself would you notice the difference between a 5GHz SB and a 4.7GHz or 4.8GHz IB?

IB gets about 3% higher IPC than SB, so a 4.85GHz 3770k will perform the same as a 5GHz 26/2700k anyways.

But you really pay for that clockspeed, with either CPU, in terms of power consumption. If you can settle for the performance of a 4.5-4.6GHz IB then you'll get considerable power savings at full load.

I don't mind pushing my chips to their limits for the purposes of characterizing those limits, but I'm not so hard-up for that last 5-10% performance that I'd be willing to put up with the extra 100W power usage because of it.

Everyone feels differently about that of course.

But the 26/2700K will perform at 4.5ghz with lower power consumption and lower temps ? Which will you recommend of those 3... they cost about the same - and I'm not expecting to be using stock coolers. The 5ghz mark would be cool, but not at all costs though ;)

Jalockin
 

Jalockin

Junior Member
Mar 20, 2013
5
0
0
So to sum it up for my self...

The best TIM to use under the IHS: Liquid Ultra
and the battle for best use between cooler and IHS is: NH-T1 vs. Indigo Extreme (I feel like its a matter of opinion here) but is the Indigo Extreme a once applied be happy and forget about it? or should it be replaced every few months?
Honestly I have NEVER replaced TIM - i didn't know it was needed.

and IDC, when you cleaned the IHS and the CPU (DIE?) you used a dremel but what type of head where you using, a polishing or a sanding one? I can't really see on the picture.

From when you started delidding til the end when you put the unit back in the motherboard how long does it take you?

As I'm completely new to the terms delidding/lapping/TIM and so on, please disregard any misuse of the terms and correct me, I'm still learning :D

J.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
But the 26/2700K will perform at 4.5ghz with lower power consumption and lower temps ? Which will you recommend of those 3... they cost about the same - and I'm not expecting to be using stock coolers. The 5ghz mark would be cool, but not at all costs though

At any given clockspeed, whatever clockspeed you want to look at, the 3770k will operate at that clockspeed with lower power consumption and deliver higher performance in comparison to a 26/2700k at the same clockspeed with the same cooling.

In this regard the 3770k is to be preferred.

and the battle for best use between cooler and IHS is: NH-T1 vs. Indigo Extreme (I feel like its a matter of opinion here) but is the Indigo Extreme a once applied be happy and forget about it? or should it be replaced every few months?

I tested both NT-H1 and Indigo Xtreme with my 2600k, check here for the entire thread or just look here and here if you want to zero in on the IX results.

5GHzPreliminaryComparisons2.png


IMO the Indigo Xtreme is superior to all other TIMs (even the Liquid Ultra) because it is as close as you can get to basically soldering the two metal surfaces together.

It is very much a "set it and forget it" TIM. You won't want to take the HSF off the CPU after you have set the Indigo Xtreme. You can remove the HSF, it isn't permanent, but once you do remove it you will need to use sandpaper to remove all of the remnants of the IX from the IHS and HSF surfaces.

and IDC, when you cleaned the IHS and the CPU (DIE?) you used a dremel but what type of head where you using, a polishing or a sanding one?


I use the white felt polishing wheels for cleaning up the PCB and the IHS. It is the "compressed felt", rather stiff and hard, not the fluffy felt stuff.

Use these:

Lot10-1-2-font-b-Felt-b-font-Wool-Buffing-font-b-Polishing-b-font-font.jpg


Also I saturate the felt with 90% rubbing alcohol (IPA) and then do a quick spin (low rpm for 2-3 seconds) with the felt wheel inside a coffee cup just to spin out the excess IPA, otherwise you'll make a mess of yourself and the work area.

The felt will get hot, you don't want to scorch the PCB or the felt, but it is easy to know when things are getting warm and need to be cooled down with IPA because it is easy to smell when the felt is heating up too much.

From when you started delidding til the end when you put the unit back in the motherboard how long does it take you?

As I'm completely new to the terms delidding/lapping/TIM and so on, please disregard any misuse of the terms and correct me, I'm still learning

This isn't a quick process :( Lapping the IHS surface takes a good hour once you have all your supplies assembled. Lapping the HSF takes a good hour as well.

Delidding, cleaning, re-lidding with new TIM, and re-assembling into mobo is a lengthy process as well. It takes me a good afternoon (3-4 hours) to get it all done.

I've done three of them so far and have not found any short-cuts. The one time I tried to be quick about it I ended up putting a good sized nick into the PCB but thankfully did not kill the chip.

That said, first time I did a lapping project, and first time I did a delidding project, because I am a "measure twice, cut once" kind of guy I spread out the process over many days.

The delidding project that is captured in the OP of this thread spanned about a week in actual time. But I was intentionally slow and methodical about it. Other people probably do theirs in an hour without issue as well.
 

Chime

Junior Member
Mar 19, 2013
23
0
0
Great thread, and thanks for all the info IDC! I'll more than likely be hitting Microcenter next week to pick up a 3570k, and since I'll be putting it into a tiny ITX rig every degree will count. I'm hoping to be able to hit 4.5GHz, and I think this will help substantially!
 

Jalockin

Junior Member
Mar 20, 2013
5
0
0
Thank you IDC for your long reply to my questions, i will not qoute the reply to avoid a wall of text ;)

You said that the 3770K was to be prefered in regards to power consumption and performance at equal clockspeeds - I might have been off when writing my question had to write it like 6 times because I got logged out or something during my writing and apparently i had to reload the page before it registered my login again, but thats another matter.

What I meant with my question was: if I use the same heatsink for both 26/2700K and 3770K, the 26/2700K will be cooler at same performance level.

I don't really know where they have the same performance but i guess its something like 4.8ghz 2600K and 4.5ghz 3770K - but I assume you have a thread about so, I will look for it :)

Thanks for all your help - and I didn't expect lapping and delidding to be a walk in the park, I am still intrigued and seriously considering to go for it just for the experience - tho my wife would kill me if i killed a €300 CPU ;)
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
What I meant with my question was: if I use the same heatsink for both 26/2700K and 3770K, the 26/2700K will be cooler at same performance level.

It depends whether you delid.

If you delid/replace CPU TIM/relid then your operating temps with the same HSF will better for the 3770k at the same clockspeed (which means performance will be higher for the 3770k too).

ClockspeedversusMaximumTemperature.png
 

kasakka

Senior member
Mar 16, 2013
334
1
81
Does anyone know if there has been any improvement to this in CPU batches made later? Considering Intel claims to have fixed the issue in Haswell, would it be possible that they have moved to better goops for new Ivy Bridges as well? Seems like it would be a pretty trivial fix in the production line.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
I plan on using either Thermalright Chill Factor 3, AS5, or a Prolimatech brand that came with my Megahalems (it doesn't say PK-1, but I think it is the same stuff before they branded their own tim).

Which tim would be best for long term set it and forget it for 3+ years?
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Welcome to the forums kasakka :thumbsup:
Does anyone know if there has been any improvement to this in CPU batches made later? Considering Intel claims to have fixed the issue in Haswell, would it be possible that they have moved to better goops for new Ivy Bridges as well? Seems like it would be a pretty trivial fix in the production line.

Any one who would be in a position of knowing the answer to that question is also not allowed not discuss it if they value their paycheck.

If you want a bunch of unfounded opinion and speculation...well then you've come to the right place :D :whiste:
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
I plan on using either Thermalright Chill Factor 3, AS5, or a Prolimatech brand that came with my Megahalems (it doesn't say PK-1, but I think it is the same stuff before they branded their own tim).

Which tim would be best for long term set it and forget it for 3+ years?

My opinion, for what it is worth, is that if you have already gone to the expense of buying a 3570k ($210) or a 3770k ($320), plus the expense of your mobo (>~$100) on top of the expense of buying a non-stock HSF ($50-$80), and you intend to go to the time and effort of delidding, then you are absolutely being "penny-wise but pound-foolish" short-sighted if you refuse to spend the extra $16 or so to purchase Coollaboratory Liquid Ultra to use under the IHS.

You want to cheap out at the very end of that long list of products you've bought, and your time spent assembling the entire project, just to use some random tube of TIM you already happen to own then you are probably going to get exactly what you have coming in terms of problems down the road.

That will no doubt come across as rather harsh, but it is my honest position on the matter. Don't cut your nose off to spite your face, use Liquid Ultra and do yourself and the CPU justice ;)
 

kasakka

Senior member
Mar 16, 2013
334
1
81
Welcome to the forums kasakka :thumbsup:


Any one who would be in a position of knowing the answer to that question is also not allowed not discuss it if they value their paycheck.

If you want a bunch of unfounded opinion and speculation...well then you've come to the right place :D :whiste:

I was thinking more along the lines of someone having recently bought one, figured out that it's from a recent batch, delidded and found out that the difference was nowhere near as drastic as this thread would suggest.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
Idontcare said:
Quote:

Originally Posted by tigersty1e

I plan on using either Thermalright Chill Factor 3, AS5, or a Prolimatech brand that came with my Megahalems (it doesn't say PK-1, but I think it is the same stuff before they branded their own tim).

Which tim would be best for long term set it and forget it for 3+ years?

My opinion, for what it is worth, is that if you have already gone to the expense of buying a 3570k ($210) or a 3770k ($320), plus the expense of your mobo (>~$100) on top of the expense of buying a non-stock HSF ($50-$80), and you intend to go to the time and effort of delidding, then you are absolutely being "penny-wise but pound-foolish" short-sighted if you refuse to spend the extra $16 or so to purchase Coollaboratory Liquid Ultra to use under the IHS.

You want to cheap out at the very end of that long list of products you've bought, and your time spent assembling the entire project, just to use some random tube of TIM you already happen to own then you are probably going to get exactly what you have coming in terms of problems down the road.

That will no doubt come across as rather harsh, but it is my honest position on the matter. Don't cut your nose off to spite your face, use Liquid Ultra and do yourself and the CPU justice

You're bare die Tim testing is actually inclining me towards using regular Tim. as5 performed within 5c of liquid ultra. and the benefits were mainly seen from removing the glue (gap).

its not that I'm cheaping out. but cleaning out the old liquid ultra is a pain and I like the flexibility to change the Tim when I want.



Posted from Anandtech.com App for Android
 

Sleepingforest

Platinum Member
Nov 18, 2012
2,375
0
76
Actually, unless you are also going bare back, you won't have to ever remove the IHS again; you'll be applying the Liquid Ultra once under the IHS and then forgetting about it.

Also, 5C is a huge difference for cooling. When watercooling, radiator brands compete with 1-2C of each other.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
Sleepingforest said:
Actually, unless you are also going bare back, you won't have to ever remove the IHS again; you'll be applying the Liquid Ultra once under the IHS and then forgetting about it.

Also, 5C is a huge difference for cooling. When watercooling, radiator brands compete with 1-2C of each other.

unless I ever decide to change motherboards or switch cpus. the act of clamping down the CPU on the mobo moves the ihs.

5c is huge. but in context if I'm already dropping 10-15c then leaving 5c on the table for extra added flexibility is worth it.


Posted from Anandtech.com App for Android
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
You're bare die Tim testing is actually inclining me towards using regular Tim. as5 performed within 5c of liquid ultra. and the benefits were mainly seen from removing the glue (gap).

its not that I'm cheaping out. but cleaning out the old liquid ultra is a pain and I like the flexibility to change the Tim when I want.



Posted from Anandtech.com App for Android

Expect to need to replace the TIM under the IHS every 6-12 months if it is not IC Diamond, Liquid Pro, or Liquid Ultra (listed from least preferable to most preferable).

That is why I recommend just going with the Ultra. That is the "set it and forget it" solution.

The other TIMs I tested in the bare-die thread were all showing results of the various TIMs from just an initial installation point of view. But none of the tests I conducted addressed the pump-out effect.

That said, if you want to be able to change out the TIM then definitely do not use IC Diamond or the Liquid Pro/Ultra products because they irreversibly alter the CPU and IHS. In that case I'd recommend you go with the AS5. Just be prepared to have to replace it in less than a year is all.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
Idontcare said:
Quote:

Originally Posted by tigersty1e

You're bare die Tim testing is actually inclining me towards using regular Tim. as5 performed within 5c of liquid ultra. and the benefits were mainly seen from removing the glue (gap).

its not that I'm cheaping out. but cleaning out the old liquid ultra is a pain and I like the flexibility to change the Tim when I want.



Posted from Anandtech.com App for Android

Expect to need to replace the TIM under the IHS every 6-12 months if it is not IC Diamond, Liquid Pro, or Liquid Ultra (listed from least preferable to most preferable).

That is why I recommend just going with the Ultra. That is the "set it and forget it" solution.

The other TIMs I tested in the bare-die thread were all showing results of the various TIMs from just an initial installation point of view. But none of the tests I conducted addressed the pump-out effect.

That said, if you want to be able to change out the TIM then definitely do not use IC Diamond or the Liquid Pro/Ultra products because they irreversibly alter the CPU and IHS. In that case I'd recommend you go with the AS5. Just be prepared to have to replace it in less than a year is all.

how are u coming up with the 6-12 month timeframe? I imagine the Tim would last much longer than that. Tim on bare die gpus last longer and go through much larger temperature ranges.


Posted from Anandtech.com App for Android
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
how are u coming up with the 6-12 month timeframe? I imagine the Tim would last much longer than that. Tim on bare die gpus last longer and go through much larger temperature ranges.


Posted from Anandtech.com App for Android

That's what various people report across the web.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
after delidding, is it normal or common to be able to decrease vcore for a given overclock compared to pre-delidded vcore?
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
after delidding, is it normal or common to be able to decrease vcore for a given overclock compared to pre-delidded vcore?

Min Vcore is a function of clockspeed and operating temperature.

Delidding by itself won't effect Vcore, but the lowering of the operating temps because of delidding will. (the same that can be expected when buying a better HSF)

3770kStockvsDelidded.png


If you look in the chart above you can see for example that Vmin at the higher clockspeeds and temperatures is then reduced once the operating temperatures were reduced.

Taking a 4.2GHz 3770k hitting 101C and needing 1.103V for stability, delidding enabled the operating temps to be lowered to 54C which then enabled the voltage to decrease to 1.071V for stability.
 

tigersty1e

Golden Member
Dec 13, 2004
1,963
0
76
I'm mainly concerned with dropping my Max 24/7 normal operating temps (gaming,etc)

at 4.7ghz and 1.304 core I get 73c Max temps during gaming.

if I delid and use thermal right chill factor 3 Tim inside and outside if the ihs, what kind if temp drop should I expect?

5c or something more like 15c?


Posted from Anandtech.com App for Android
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
I'm mainly concerned with dropping my Max 24/7 normal operating temps (gaming,etc)

at 4.7ghz and 1.304 core I get 73c Max temps during gaming.

if I delid and use thermal right chill factor 3 Tim inside and outside if the ihs, what kind if temp drop should I expect?

5c or something more like 15c?


Posted from Anandtech.com App for Android

Yeah you should see more like a 15C drop in temps, and along with the temperature drop you should be able to reduce your voltage to something around 1.25V.

YMMV of course, silicon lottery and all.
 

kasakka

Senior member
Mar 16, 2013
334
1
81
After the lid is removed and paste changed, do you glue it back on or is the thermal paste and socket bracket holding it in place?
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
After the lid is removed and paste changed, do you glue it back on or is the thermal paste and socket bracket holding it in place?

You can do either.

Personally I have never bothered to glue any of mine back together. The socket retention mechanism locks it in place so tight that it never moves.

It will move around while being locked down, so you kind of have to give it a bit of "planned relocation" by placing the IHS off-center at first and then letting the socket retention mechanism actually push it right into place at the same time as it locks it down and prevents it from moving any more.