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

Page 28 - 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
Where are you guys getting your CL Ultra?

PCWarGamer, I do intend to test a handful of TIM's once I get through my current set of tests. But I have a limited selection - MX4, NTH1, AS5, Ceramique, IC Diamond, IC Perihelion, TX2.

I have no idea on the timeline for results from those tests as I have been getting tempted lately to drag out my phase unit and see what happens at -50C with bare-die mounting :hmm:
 

oceanside

Member
Oct 10, 2011
50
0
0
Lapped and running CL Ultra on my 2500k... dropped over 10C at load. Great stuff. I think I over-applied, it didn't flow very well from the syringe. No issues so far. The syringe has markings and looks like you can get 3-4 mounts. CL Ultra supposedly doesn't degrade: If true, then I'm very pleased. I should be able to keep the HSF/Processor together until the MB/CPU degrades. If you use CL Ultra and remove your HSF, supposedly, you must relap as the CL Ultra infuses into the copper and/or nickel plating.

I'm not sure what the effects would be, long term, for use on a bare die. Would like to know though, especially if I'm going to be faced with delidding down the road (i.e. Intel sticks with their IHS paste & inconsistent gaps on upcoming processors)

Got my CL Ultra at Amazon as well.
 

PCWargamer

Junior Member
Oct 16, 2012
10
0
0
Where are you guys getting your CL Ultra?

PCWarGamer, I do intend to test a handful of TIM's once I get through my current set of tests. But I have a limited selection - MX4, NTH1, AS5, Ceramique, IC Diamond, IC Perihelion, TX2.

I have no idea on the timeline for results from those tests as I have been getting tempted lately to drag out my phase unit and see what happens at -50C with bare-die mounting :hmm:
Sidewinder - $15.79 shipped USPS 2-3day

Good news about the TIM tests - even if only a small sample - you probably need to include CL PRO and Ultra, as they seem to be two of the best.

Can understand the attraction to see how bare-die will do with your phase unit. :)
 

sun_tzu

Member
Oct 17, 2012
26
0
0
Are they ever going to fix the TIM issues? I'd feel like this kinda thing would be a great PR move for intel (Same chip, 20c COOLER!)
 

C.C.

Member
Aug 21, 2012
28
0
0
I've finally gotten a Prime95 Stable o/c @ 4.8Ghz..While I could run 20+passes of IBT using all memory and with the psychical cores locked through task manger, it would fail Prime95 (27.7 with AVX enabled) until I bumped my Vcore up from 1.256V (IBT Extreme Mode stable) to 1.272V...I ran Prime95 for 18hours without an issue, so I feel it is fully stable @ this point..



While Max temps were 70,81,80,78C, average temps were a good bit lower then peak, anywhere from 5~7C per core according to the log..

I can boot into Windows @ 4.9Ghz, and even run 10 passes of IBT @ "Standard", but Pime95 causes a hardlock..Do you guys think this is simply a sign I need more Vcore, or do you think it has something to do with memory etc?

Also, what are you guy's running your PLL voltage @? Mine has always been set to "Auto", which is 1.81V..Do you think this can be lowered without effecting stability? I would really love to see 4.9Ghz-5Ghz if possible, but am unsure where to proceed from here..
 

Ed1

Senior member
Jan 8, 2001
453
18
81
I think you would have to see what size is causing issues in prime95, run small size then large and see which gives problems .
I read over at HardOCP they raised cpu PLL to something like 1.9 (check there ivy bridge MB review)to get high clocks speeds .

http://www.hardocp.com/article/2012/04/23/intel_ivy_bridge_processor_ipc_overclocking/5

I am just throwing this out there as I did some reading before buying P8Z77v-pro, not sure if there results were just that cpu or mb but worth mentioning I think .

IMO I would try bumping the vcore a notch first and see how it goes, if it gets better .
 
Last edited:

C.C.

Member
Aug 21, 2012
28
0
0
I think you would have to see what size is causing issues in prime95, run small size then large and see which gives problems .
I read over at HardOCP they raised cpu PLL to something like 1.9 (check there ivy bridge MB review)to get high clocks speeds .

http://www.hardocp.com/article/2012/04/23/intel_ivy_bridge_processor_ipc_overclocking/5

I am just throwing this out there as I did some reading before buying P8Z77v-pro, not sure if there results were just that cpu or mb but worth mentioning I think .

IMO I would try bumping the vcore a notch first and see how it goes, if it gets better .

Thanks for the link Ed, I am going to check it out..I always run Prime95 on Large FFTs, since that mode draws the most power, and produces the most heat..I always test IBT first, because it will produce stability errors very quickly if your ram is @ fault..
 

rge2

Member
Apr 3, 2009
63
0
0
@CC, prime 18 hrs screenshot, worker 1 and worker 2 failed at 14hrs. When running prime make sure under advanced and check... round off/error checking.

But nice cpu/overclock with low vcore! just need to bump up vcore 1 notch to run prime 18hrs and probably 2 notches to run it without whea 19 errors...errors the cpu corrects before prime sees them (viewable in event viewer>>warnings>>whea 19 errors, you probably have some during that prime run).

I wouldnt bother with cpu pll unless doing LN2 clocks. And typically it is ~.05v more for each 100mhz more with IVY in 4.5ghz range, until you get to nearer to 5, then scaling gets worse .07v+ per 100 mhz, though some cherry cpus keep lower scaling even at higher end.

For mine prime 12hrs without whea erros, 4.5ghz need 1.2v load, 4.7 need 1.31v. Then 4.8 just to run prime couple hours need 1.39v...your cpu is much better than mine with lower vcore.

If you need 1.28v for 4.8 error free, then 1.34-1.35+ for 4.9, 1.42ish possibly little higher for 5.0...you might make 5 with reasonable vcore.
 

PCWargamer

Junior Member
Oct 16, 2012
10
0
0
Welcome to the forums PCWargamer! :thumbsup:

Let us know how it goes, both the delidding as well as the results with CL Ultra. Very curious to hear more about CL Ultra.
IDC, I have delidded my 3770K and am using CL Ultra on all surfaces. My temps are lower, but I have a question on method to use to measure and report that data:

least core change - 11C
greatest core change - 21C
average change all cores - 15C
highest temp to highest temp - 17C

What seems to be everyone's opinion on the best method to use? Also, which load program to use is a question - IBT (default?), prime95 (90% mem?), LinX, etc? OC and vcore are also important. My results above are at 4.5GHz with same vcore before and after.

CL PRO is on the way, plus I hope to lap soon after that and see what effect that has.
 
Last edited:

MrK6

Diamond Member
Aug 9, 2004
4,458
3
81
Glad I finally got to read through this thread. Thanks IDC for the excellent write-ups and these are some impressive results you guys are getting. Makes my 2500K feel boring. :p
 

Ed1

Senior member
Jan 8, 2001
453
18
81
what is this CL pro and ultra , can you give full name I can't find anything so far with CL .

Edit: found it Coollaboratory Liquid Ultra
 
Last edited:

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
IDC, I have delidded my 3770K and am using CL Ultra on all surfaces. My temps are lower, but I have a question on method to use to measure and report that data:

least core change - 11C
greatest core change - 21C
average change all cores - 15C
highest temp to highest temp - 17C

What seems to be everyone's opinion on the best method to use? Also, which load program to use is a question - IBT (default?), prime95 (90% mem?), LinX, etc? OC and vcore are also important. My results above are at 4.5GHz with same vcore before and after.

CL PRO is on the way, plus I hope to lap soon after that and see what effect that has.
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.
 

Ed1

Senior member
Jan 8, 2001
453
18
81
yeah, hottest core and max peak is what I would think matters .

Idontcare : do you know how thick the IHS is approx ?

I wonder if going to thicker IHS would help spread heat on HS with thin bases or with direct heat pipe HS .

I was thinking i wonder how much the outer heat pipes work compared to the inner 2 on a typical 4 pipe HS .
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Interesting update. I checked ASRock's site for a bios update and noticed this:

"10/4/12 - 2.30
1. Update CPU code.
2. Modify CPU load line calibration behavior"

Something is up with this bios because my temps have come down almost 10c at 4.5ghz at the lowest voltage I was stable at before, and I'm currently Prime'ing at lower and lower voltages. I'll report back when I have some more results.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
Interesting update. I checked ASRock's site for a bios update and noticed this:

"10/4/12 - 2.30
1. Update CPU code.
2. Modify CPU load line calibration behavior"

Something is up with this bios because my temps have come down almost 10c at 4.5ghz at the lowest voltage I was stable at before, and I'm currently Prime'ing at lower and lower voltages. I'll report back when I have some more results.
One thing that will markedly reduce temperatures is the CPU voltage.

The problem here is that if you are relying on software like CPUz to tell you the CPU voltage then you are truly left blind by the mobo BIOS makers because it is complete hit and miss.

I had no idea how much the actual voltage applied to the CPU meanders as a function of BIOS rev for a given mobo until I started measuring my Vcc externally with a voltmeter while keeping track of it via CPUz.

Some BIOS revs the Vcc would be under-reported (CPUz would report 1.2V when actual was 1.25V, etc), other BIOS revs would result in Vcc being over-reported (CPUz would report 1.2V when actual was 1.15V, etc), and other BIOS revs were dead-nuts accurate to within CPUz own internal forced quantization/rounding (it rounds down in 0.008V increments).

What I learned from that was if you see big changes in your CPU temperature when you change the BIOS then it is very likely that what you aren't seeing is a huge change in the actual applied voltage to the CPU as well.

Which in the end is a good thing, don't look a gift horse in the mouth ;) :D

yeah, hottest core and max peak is what I would think matters .

Idontcare : do you know how thick the IHS is approx ?

I wonder if going to thicker IHS would help spread heat on HS with thin bases or with direct heat pipe HS .

I was thinking i wonder how much the outer heat pipes work compared to the inner 2 on a typical 4 pipe HS .
Based on the measurements taken and reported in posts 77 and 78, the IHS metal thickness (the part that is sandwiched between the CPU die and the HSF) is 4.15-1.57=2.58mm.

So ~2.6mm thick.

From the heat-spreader angle you want the IHS to be as thin as possible. A thick IHS is no different than having a thick copper plate attached to the bottom of the HSF covering up those direct-contact pipes in the first place.

On the desktop the IHS is there to physically protect the die as well as to enable a very cheap stock HSF while still hitting the thermal targets.

That is why the IHS is not even used on mobile solutions. The environmental exposure is well controlled by the OEM that is assembling the laptop, and the cooling solution is at a premium for space and weight reasons so they don't mess around with an IHS whatsoever.
 

Ed1

Senior member
Jan 8, 2001
453
18
81
One thing that will markedly reduce temperatures is the CPU voltage.

The problem here is that if you are relying on software like CPUz to tell you the CPU voltage then you are truly left blind by the mobo BIOS makers because it is complete hit and miss.

I had no idea how much the actual voltage applied to the CPU meanders as a function of BIOS rev for a given mobo until I started measuring my Vcc externally with a voltmeter while keeping track of it via CPUz.

Some BIOS revs the Vcc would be under-reported (CPUz would report 1.2V when actual was 1.25V, etc), other BIOS revs would result in Vcc being over-reported (CPUz would report 1.2V when actual was 1.15V, etc), and other BIOS revs were dead-nuts accurate to within CPUz own internal forced quantization/rounding (it rounds down in 0.008V increments).

What I learned from that was if you see big changes in your CPU temperature when you change the BIOS then it is very likely that what you aren't seeing is a huge change in the actual applied voltage to the CPU as well.

Which in the end is a good thing, don't look a gift horse in the mouth ;) :D



Based on the measurements taken and reported in posts 77 and 78, the IHS metal thickness (the part that is sandwiched between the CPU die and the HSF) is 4.15-1.57=2.58mm.

So ~2.6mm thick.

From the heat-spreader angle you want the IHS to be as thin as possible. A thick IHS is no different than having a thick copper plate attached to the bottom of the HSF covering up those direct-contact pipes in the first place.

On the desktop the IHS is there to physically protect the die as well as to enable a very cheap stock HSF while still hitting the thermal targets.

That is why the IHS is not even used on mobile solutions. The environmental exposure is well controlled by the OEM that is assembling the laptop, and the cooling solution is at a premium for space and weight reasons so they don't mess around with an IHS whatsoever.
wouldn't that create a stability issue for OC that originally setup vcore to min values for a given clock speed and then bios changes the LLC. It good chance it might not be stable at values he had for older bios and needs retesting and possibly updating vcore values. Changes in bios like that might help users who use auto in most fields as they aren't tweaking the Vcore manually or offset to the best value .

That sounds like decent amount for thickness 0.100" approx . What I was driving at for example the 212evo 4 heatpipes run parallel to the core length . SO the outer ones are not under the die .

I just did experiment, ran prime95 and measure each heatpipe , well I didn't find outer ones to be cooler . the difference was little over 1c from coolest to hottest and it seemed to go from one side to the other, so in my case one side was hotter .

Hmm, looking at that pic of dies I see the cpu cores offset with cache on the other side .
Edit: there hotter on left side looking down at MB, so it seems that is were the cores are off to a side .

Either way, IHS seems fine and I "might" have a mounting issue but it is very small delta so probably just die construction .
 
Last edited:

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
wouldn't that create a stability issue for OC that orginally setup vcore to min values for a given clock speed and then bios chages the LLC. It good chance it might not be stable at values he had for older bios and needs retesting and possibly updating vcore values.

Absolutely it would be a problem, which is why you are expected to clear your CMOS after installed the new BIOS in which case the old OC settings are completely wiped from the system and you have no choice but to go back in and start rebuilding your OC profile.

Even on my MIVEZ which lets you store 9 separate OC profiles, those OC profiles are completely wiped out and removed when I upgrade the BIOS.

That sounds like decent amount for thickness 0.100" approx . What I was driving at for example the 212evo 4 heatpipes run parallel to the core length . SO the outer ones are not under the die .
It is very thick, surprising thick. And the CPU die is just the opposite, surprisingly thin.
 

Ed1

Senior member
Jan 8, 2001
453
18
81
Absolutely it would be a problem, which is why you are expected to clear your CMOS after installed the new BIOS in which case the old OC settings are completely wiped from the system and you have no choice but to go back in and start rebuilding your OC profile.

Even on my MIVEZ which lets you store 9 separate OC profiles, those OC profiles are completely wiped out and removed when I upgrade the BIOS.



It is very thick, surprising thick. And the CPU die is just the opposite, surprisingly thin.
Ahh, right, forgot about that . you should run default to before updating AFAIK .

Didn't know it removes the profiles I always wondered how profile works, is it just values changed that get saved (in which case you might be able to apply to new bios unless new fields/feature are added) .
Or is it like image of bios in which case for sure you couldn't use profile .
 

Idontcare

Elite Member
Oct 10, 1999
21,118
57
81
The profile saves all the bios settings, from fan settings to whether you are using manual or offset voltages to your memory settings, etc. It is actually pretty handy if you happen to be tweaking your bios settings a lot.

Probably a useless feature though for those folks who setup an OC and leave their system at those settings for a year or more at a time.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
I've played with most of the voltage settings in bios quite a bit, I typically don't leave things on "auto". I have had my air conditioner on and thus a stable air temperature all summer so that wasn't a variable. I have reset my bios many times and started over just to see if any of my settings were having a negative impact. I have built a pretty extensive chart of relations between voltage / temperature / stable clocks. The changes in operating temperature are definitely a result of the new bios.

______________

Interestingly, a cold front came through this morning and I opened my windows, and due to a combination of new bios + 5c lower air temperatures, I was able to drop my operating voltage by almost 70mv. I'm beginning to suspect that I'm right on the edge of runaway thermals, hopefully I'll have great results with CL Ultra, I expect it to arrive on Tuesday. Just as I'm starting to figure things out, something changes! Huzzah.
 

ASK THE COMMUNITY