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

Page 29 - 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
21,118
57
81
IDC, is there any place I might be able to measure vcore from on my board if they haven't specifically made one?

http://www.asrock.com/mb/Intel/Z77 Extreme4/
Unfortunately I don't know enough about it to give you any kind of guidance that could be replied upon. I bet if you asked around in the XS forums you would find some knowledgeable people who could get you pointed in the right direction.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
I guess what you would have to check is how loading of cpu affects vcore with auto, and then each of LLC settings or a least the ones you used .
See if VID or vcore is affected by new bios .
My board has 5 LLC settings, 1-5, and "auto" is the same as picking LLC @ lvl 5 (the lowest), both before and after the new bios. LLC doesn't change the requested VID at a given offset but it does change the voltage received by the CPU as reported by CPU-Z. Any setting above LLC @ 3 looks to overshoot sometimes - that is, a light load on a single core will actually have a lower voltage than a heavy load on all 4 cores. This doesn't appear to have changed with the new bios, but at a given voltage my CPU is running at a lower temperature. As a result, I have been able to lower my voltages because my chip seems very sensitive to operating temperature.

IE, 4.4ghz @ 1.216v in the old bios might run 75c, it now runs at 70c at 1.216v and as a result I'm able to lower my voltage to, say, 1.192v. The difference is more pronounced at higher clocks/voltages.
 
Last edited:

Ed1

Senior member
Jan 8, 2001
453
18
81
My board has 5 LLC settings, 1-5, and "auto" is the same as picking LLC @ lvl 5 (the lowest), both before and after the new bios. LLC doesn't change the requested VID at a given offset but it does change the voltage received by the CPU as reported by CPU-Z. Any setting above LLC @ 3 looks to overshoot sometimes - that is, a light load on a single core will actually have a lower voltage than a heavy load on all 4 cores. This doesn't appear to have changed with the new bios, but at a given voltage my CPU is running at a lower temperature. As a result, I have been able to lower my voltages because my chip seems very sensitive to operating temperature.

IE, 4.4ghz @ 1.216v in the old bios might run 75c, it now runs at 70c at 1.216v and as a result I'm able to lower my voltage to, say, 1.192v. The difference is more pronounced at higher clocks/voltages.
"Any setting above LLC @ 3 looks to overshoot sometimes - that is, a light load on a single core will actually have a lower voltage than a heavy load on all 4 cores. "

That is how my MB works to and seems logical to me .
If your voltage is the same under load I wonder why/were the lower temps are coming from .
Either way its good, I just wonder were change is from .
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Well the way I figure, if a CPU needs 1.300v for stability at a given clock speed, it probably needs it regardless of how light or heavy the load is? So if under a light load you have, say, 1.300v and under a heavy load you're at 1.320v, your heavy load voltage is unnecessarily high. I'm Prime'ing right now with LLC at 3 which seems to have minimal droop, but doesn't overshoot.

As for the lower temps, I'm not sure, but power circuitry is a fairly complex thing so there's a lot that could have changed.
 

Ed1

Senior member
Jan 8, 2001
453
18
81
Well the way I figure, if a CPU needs 1.300v for stability at a given clock speed, it probably needs it regardless of how light or heavy the load is? So if under a light load you have, say, 1.300v and under a heavy load you're at 1.320v, your heavy load voltage is unnecessarily high. I'm Prime'ing right now with LLC at 3 which seems to have minimal droop, but doesn't overshoot.

As for the lower temps, I'm not sure, but power circuitry is a fairly complex thing so there's a lot that could have changed.
I do think you need lower (can get away with lower) with lighter loads , no need to push full load voltages if your cpu loading is below 100% . thats just my opinion and seems how the MB works too .

Yes, you also never know what changes they make and don't list them , so its hard to say for us to figure out .
 

PCWargamer

Junior Member
Oct 16, 2012
10
0
0
All my temperature numbers are for the hottest core, peak temperature for the hottest core, after 5 runs using LinX/IBT with problem size 43122 (~14.2GB ram used).

Peak temperature is what matters in terms of CPU throttling, the PCU doesn't care if three cores are at 40C and one it as 105C, if any one core reaches TJmax then the PCU starts throttling all the cores at the same time.

Peak temperature is also what determines the maximal rate of degradation, and it is what determines the minimum Vcc that you have to apply to all the cores such that it is stable and can pass LinX/IBT at the temperatures it is running at.

In short, average temp means nothing, neither does min temp. Max (peak) temperature for the hottest core IS the rate limiting temperature that plays into all the things that we are wrestling with as OC'ers, so it is the temperature you care most about.
Thanks. Makes great sense. In the OCN IVY Delidders Club we are going to request that everyone who joins do a pre & post delid run of IBT (default 10x) @ 4.5GHz with recorded comparisons of high temps. Should be an easy enough way to gain some useful info on temp gains for delidding and for the different TIMs used.

Best results so far are from CL PRO. ICD did well too, but not as good as PRO. Still seeing how CL Ultra does. Also seeing if temp gains degrade over time due to the possible "pump out" problem found on this thread.

BTW, I gained 17C from delidding and using CL Ultra, but then I lapped the IHS and gained another 12C for a total gain of 27C. Turned out my IHS was concave on the top and convex on the bottom, which became very visible while lapping/sanding. Much flatter now on both sides. :)

Edit: I just noticed 17+12=29 which is not correct total gain - sorry for mistake above - here is the actual test run results data:

1) pre-delidded, IBT (default 10x run) @4.5GHz, high temp = 96C
2) delidded IBT, (default 10x run) @4.5GHz, high temp = 82C (-14C)
3) lapped IHS & delidded IBT, (default 10x run) @4.5GH,z high temp = 70C (-26C)

One reason for the additional 12C gain was that my IHS was concave, which came out during sanding/lapping, so that the HSF never made the best contact with it.
 
Last edited:

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Thanks. Makes great sense. In the OCN IVY Delidders Club we are going to request that everyone who joins do a pre & post delid run of IBT (default 10x) @ 4.5GHz with recorded comparisons of high temps. Should be an easy enough way to gain some useful info on temp gains for delidding and for the different TIMs used.

Best results so far are from CL PRO. ICD did well too, but not as good as PRO. Still seeing how CL Ultra does. Also seeing if temp gains degrade over time due to the possible "pump out" problem found on this thread.

BTW, I gained 17C from delidding and using CL Ultra, but then I lapped the IHS and gained another 12C for a total gain of 27C. Turned out my IHS was concave on the top and convex on the bottom, which became very visible while lapping/sanding. Much flatter now on both sides. :)
My processor ran hot even for Ivy to begin with (not sure if you've followed my posts in this thread). I went UP several celsius delidding and replacing my TIM with Phobya HeGrease (like 78->81), and only dropped a few degrees running bare die with the same paste (78->81->74). This is under water.
 

PCWargamer

Junior Member
Oct 16, 2012
10
0
0
My processor ran hot even for Ivy to begin with (not sure if you've followed my posts in this thread). I went UP several celsius delidding and replacing my TIM with Phobya HeGrease (like 78->81), and only dropped a few degrees running bare die with the same paste (78->81->74). This is under water.
Yes, I recall. You had the unique chip that did not improve temps through a delid. It was suspected of being due to an already low gap clearance for your chip. Even though it appears hot to you, I think it was lower than others at same OC and vcore. I could be wrong about that. Have you tried CL PRO or Ultra yet? Maybe some improvment could result from doing so.

Your experience is very interesting, but it seems that chip is an oddity as so many others need the delid to make the IB what it should have always been. Normally temp drops are ~20C+ which can lead to some lower vcore for some OCs and extended OC through not hitting temp barrier at 105C.

I am glad you did share the results of your chip, though, so that we know that that can happen even with all your efforts to improve its temps. I would be interested if there are others who have ran into what you did on this thread. Thanks for sharing your experience.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
I'm hoping my CL Ultra will come in today. Has anyone reviewed it against Pro? Anyone know what was added to make it more paste-like?
 

PCWargamer

Junior Member
Oct 16, 2012
10
0
0
I am wanting to try CL PRO on my die to compare its effects to the CL Ultra on it now.

But I am wondering if PRO can be cleaned off of the die once applied.
 

TaranScorp

Senior member
Oct 13, 2009
253
0
76
This thread got me interested so I delidded my 3770K and used CL PRO and then Timtronics Grey Ice 4200. I booted up checked my temps and at idle nothing changed and thought I just wasted hours for nothing, but then I decided to try OCCT linpack to stress test and was really happy. I'm running 4.4 at 1.18 and was topping out at 74c at hottest core and there was always about 5c to 8c difference between cores. Now my hottest core tops out at 60c and there is only a 2c difference between all cores and I have HT on and Igpu on.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,236
1,168
126
This thread got me interested so I delidded my 3770K and used CL PRO and then Timtronics Grey Ice 4200. I booted up checked my temps and at idle nothing changed and thought I just wasted hours for nothing, but then I decided to try OCCT linpack to stress test and was really happy. I'm running 4.4 at 1.18 and was topping out at 74c at hottest core and there was always about 5c to 8c difference between cores. Now my hottest core tops out at 60c and there is only a 2c difference between all cores and I have HT on and Igpu on.
These little testimonies are beginning to accumulate, despite our overall conclusion thus far about TIM replacement and "the gap." Never heard of Timtronics Grey Ice and now I'll have to look it up . . .
 

BonzaiDuck

Lifer
Jun 30, 2004
15,236
1,168
126
This thread got me interested so I delidded my 3770K and used CL PRO and then Timtronics Grey Ice 4200. I booted up checked my temps and at idle nothing changed and thought I just wasted hours for nothing, but then I decided to try OCCT linpack to stress test and was really happy. I'm running 4.4 at 1.18 and was topping out at 74c at hottest core and there was always about 5c to 8c difference between cores. Now my hottest core tops out at 60c and there is only a 2c difference between all cores and I have HT on and Igpu on.

Any idea about what the ingredients for Grey Ice are?
 

dqniel

Senior member
Mar 13, 2004
650
0
76
It's been about three weeks since I put IC Diamond between the core and the IHS, and load temperatures are still vastly improved over the stock solution. I will update my old post in a moment to show details-

Thermal Cycling/TIM Failure Test

System Info:

3570K Delidded with IC Diamond between the die and IHS
ZT-10D HSF with Arctic Silver Ceramique between the IHS and the HSF
4.5Ghz @ 1.232v (after vdroop)

Delidded and IC Diamond applied 8/29/2012

After one week:
(9/4/2012)

Ambient temp - 25C
3570K max temps after Prime "blend" for 30 minutes - 67C, 76C, 74C, 73C

After three and a half weeks:
(9/22/2012)

Ambient temp - 22C
3570K max temps after Prime "blend" for 30 minutes - 62C, 72C, 71C, 69C

After about five weeks:
(10/1/2012)

Ambient temp - 22C
3570K max temps after Prime "blend" for 30 minutes - 62C, 72C, 70C, 69C


After lapping IHS and HSF base down to the copper and using Coollaboratory Liquid Ultra in all areas of contact:
(10/2/2012)

Ambient temp - 21C
3570K max temps after Prime "blend" for 30 minutes - 52C, 61C, 56C, 59C :eek:

(10/24/2012)

Ambient temp - 23C
3570K max temps after Prime "blend" for 30 minutes - 55C, 63C, 60C, 61C

More results to come...
Hard to believe it's been three weeks since I've applied the CL Liquid Ultra. I'll be updating these results ASAP to see if temps have degraded. Possibly tonight.

Updated with tonight's results. It appears the CL Ultra is still going strong.
 
Last edited:

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Update! CoolLaboraty Liquid Ultra came in!


Before, with Phobya HeGrease (which multiple reviews show to be among the best non-metal pastes:




3570K with CL Ultra applied:




Results:




I'm mega-impressed with the temperature uniformity I have now!

My new TIM hasn't had any time to set in, though I'm really not sure if liquid metal needs to. I currently have my windows open and 79°F is also my room temperature, which is approximately 26c. Remember that I reapplied my Phobya HeGrease multiple times with no improvement in temperature.

I'm highly impressed with CL Ultra, though I definitely would not attempt to apply it to the bare die of a GPU, as I would worry that if I spilled some, I would never get it off of small components. I haven't worked with Liquid Metal Pro, but Ultra definitely does not behave like a normal "paste", it tends to form beads much like water on (insert your choice hydrophobic surface) but can be spread out if you work it a little.

EDIT: I'm currently running with LLC @ 3 which gives the closest to no vdroop with my new bios, but I'm considering either using a negative offset in bios or dropping LLC down a notch so I can lower the voltage below +0mv, as I'm currently Priming there where it would BSOD quite quickly before.
 
Last edited:

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Here's my extreme upper-end results:



4.7ghz previously ran at approximately 100c under Prime small FTT load. After changing to liquid metal (and with a 2-3c increase in room temperature) I'm down to ~82c on my hottest core.

Also, shame on Intel for making such a power hungry chip, I mean 117w? Ridiculous. (I believe that's calculated off of VID which is almost 1.5v)
 
Last edited:

Ed1

Senior member
Jan 8, 2001
453
18
81
Here's my extreme upper-end results:



4.7ghz previously ran at approximately 100c under Prime small FTT load. After changing to liquid metal (and with a 2-3c increase in room temperature) I'm down to ~82c on my hottest core.

Also, shame on Intel for making such a power hungry chip, I mean 117w? Ridiculous. (I believe that's calculated off of VID which is almost 1.5v)
looks good .

Yes, wonder how accurate that is when you use manual/off-set voltages, does that only apply to VID/Auto values. Since for example I get like 57-60w at 3.8 .

If you have UPS with power usage display that would be good way to compare those values .
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
Also, shame on Intel for making such a power hungry chip, I mean 117w? Ridiculous. (I believe that's calculated off of VID which is almost 1.5v)
At 4.7GHz, 1.368Vcc, and 82C your chip is not consuming 117W, it is more like 165W (I have a new thread coming out soon on the topic of temperature versus voltage and power consumption for Ivy Bridge, nearly have all my data acquired at this point - some 900 data points :eek:).

The power readings that Intel's PCU reports appears to neglect/ignore the static leakage power consumption. As near as I can tell it simply reports the dynamic power consumption.

My 3770k at 4.7GHz and 1.281V peaks at 80C and consumes 143W (28W static power losses, 115W dynamic power usage).
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
At 4.7GHz, 1.368Vcc, and 82C your chip is not consuming 117W, it is more like 165W (I have a new thread coming out soon on the topic of temperature versus voltage and power consumption for Ivy Bridge, nearly have all my data acquired at this point - some 900 data points :eek:).

The power readings that Intel's PCU reports appears to neglect/ignore the static leakage power consumption. As near as I can tell it simply reports the dynamic power consumption.

My 3770k at 4.7GHz and 1.281V peaks at 80C and consumes 143W (28W static power losses, 115W dynamic power usage).
I have a UPS, I can give full system idle and load consumption figures.

At 4.5ghz / 1.30VID / ~1.248v vcore, I have 243w at the wall at idle, 363w while Prime'ing, and 373w under IBT. CoreTemp reports around 85w consumed by the CPU under IBT, and around 12.5w at idle. Delta power draw is thus 132.5w at the wall, and 72.5w as reported via CoreTemp.

Dropping down from 4.5 to 4.4 and reducing the voltage by around 0.30v brings IBT load power at the wall down to 364w, 9w lower than at 4.5ghz.
 
Last edited:

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
I have a UPS, I can give full system idle and load consumption figures.

At 4.5ghz / 1.30VID / ~1.248v vcore, I have 243w at the wall at idle, 363w while Prime'ing, and 373w under IBT. CoreTemp reports around 85w consumed by the CPU under IBT, and around 12.5w at idle. Delta power draw is thus 132.5w at the wall, and 72.5w as reported via CoreTemp.

Dropping down from 4.5 to 4.4 and reducing the voltage by around 0.30v brings IBT load power at the wall down to 364w, 9w lower than at 4.5ghz.
Yup, those are much more realistic numbers. Makes me wonder what value the Intel numbers are that it reports to programs like CoreTemp when the numbers themselves are so clearly in error.
 

Ed1

Senior member
Jan 8, 2001
453
18
81
I have a UPS, I can give full system idle and load consumption figures.

At 4.5ghz / 1.30VID / ~1.248v vcore, I have 243w at the wall at idle, 363w while Prime'ing, and 373w under IBT. CoreTemp reports around 85w consumed by the CPU under IBT, and around 12.5w at idle. Delta power draw is thus 132.5w at the wall, and 72.5w as reported via CoreTemp.

Dropping down from 4.5 to 4.4 and reducing the voltage by around 0.30v brings IBT load power at the wall down to 364w, 9w lower than at 4.5ghz.
is those values from wall only system or do you also have router, monitor connected .

I see many reviewers stating something like 300w approx and that in 3d benching (mainly testing vid card draw ) . many times the system is running OC cpu in test .

The power draw that is shown/displayed in many cpu apps, is that just the CPU or they talking the whole MB+ cpu .

Edit: hmm , maybe it avgs out as in 3d benching or game testing the CPU usage is way down so might avg out close .

look at these from 4.8ghz 2500k
http://www.hardocp.com/article/2012/09/03/msi_gtx_660_ti_power_edition_oc_video_card_review/9

Anand's 3770k

http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review/20

idle power is way down compared to above post . need to find his OC results .

This one matches what IDC was saying (pretty high power usage at high frequencies
http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge
 
Last edited:

dqniel

Senior member
Mar 13, 2004
650
0
76
Glad to hear the Liquid Ultra worked out so well for you Yuriman. I love the temperature change I got from it, as well.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
4.7ghz previously ran at approximately 100c under Prime small FTT load. After changing to liquid metal (and with a 2-3c increase in room temperature) I'm down to ~82c on my hottest core.
Glad to hear the Liquid Ultra worked out so well for you Yuriman. I love the temperature change I got from it, as well.

That is awesome :thumbsup: I am so gonna get me some CL Ultra now.

Seems like we are building a general concensus in the enthusiast community, from multiple forums now, that the CL Ultra is the way to go for delidded IB's in which the owner is reattaching the IHS after replacing the CPU TIM.
 

ASK THE COMMUNITY