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

[lau808]

Hey idc are you gonna do delid and oc the hell out of haswell by chance? Not interested in the results for myself but i just love reading your "articles"

 

[Idontcare]

Hey idc are you gonna do delid and oc the hell out of haswell by chance? Not interested in the results for myself but i just love reading your "articles"

I would enjoy doing it but it isn't in the cards right now.

I'm sure I would be more motivated to find a way to make it happen if 22nm wasn't such a dud at the high-end versus 32nm (just thinking of my IB vs SB experience)...so for now I have my sights set on 14nm.

And if the current trend is any indication of things, 14nm will likely come with a 1cm gap between CPU and IHS and it will be filled with chewing gum So the hobbyist will really have something to chew on while delidding

 

[MrK6]

I would enjoy doing it but it isn't in the cards right now.

I'm sure I would be more motivated to find a way to make it happen if 22nm wasn't such a dud at the high-end versus 32nm (just thinking of my IB vs SB experience)...so for now I have my sights set on 14nm.

And if the current trend is any indication of things, 14nm will likely come with a 1cm gap between CPU and IHS and it will be filled with chewing gum So the hobbyist will really have something to chew on while delidding

If the TIM remove actually doubles as chewing gum, that's just Intel tipping its hat towards the enthusiasts.

 

[sanurocks]

i delidded my 3570k.

Then used direct die mounting using h80i and nt-h1. the temperature were high and apparently no difference before delidding. also before and after delidding temp difference between hottest and coldest core was 13C.

then i used ihs the temps are much more even and much lowerbut i am not sure that will using ihs without the adhesive space will crush my die.

other detail
4.4 ghz 1.24v
room temp 31C(i live in india)
right now running 4.2 ghs 1.17v
max temp 60C IBT MAX 13379MB H80i ON balanced

 

[crashtech]

^ It seems to me that the mounting of the waterblock to the bare die must have had a slight problem, which was corrected with the inclusion of the IHS.

 

[GPz1100]

^^To the guy in post # 1000, you did though clean the die and the underside of the water block before attachment to the cpu. Also, the water block needs to have firm contact with the die. That is, it shouldn't be gorilla torqued to it, but should not be finger tight either. Finally, you are using something in between the dye and water block?

 

[BUTCH1]

Hmm, over time I've always noticed that paste seems to dry up a bit, whenever I take apart an old computer the thermal grease always looks dry and worn out, if someone ran these CPU's for 8-10 years say would the same thing happen?, would the CPU start to run warmer due to aging grease?, maybe the fact that it's sealed inside the IHS prevents this from happening, just a noob question..

 

[sanurocks]

^^To the guy in post # 1000, you did though clean the die and the underside of the water block before attachment to the cpu. Also, the water block needs to have firm contact with the die. That is, it shouldn't be gorilla torqued to it, but should not be finger tight either. Finally, you are using something in between the dye and water block?

Dear IDC please help..
Any one experienced please help me..
Yes i cleaned it with ipa 99%.. Every time.
Right now using noctua tim between die ihs and ihs h80i.
And please tell that will i risk the die if ihs is used above it as there is some space between it and the pcb..
Also please suggest the best tim to use between die and ihs(i am going to keep ihs on).good tim which are available in my country are
Cm thermal fusion 400,essentials e1,e2,x1
Deepcool z5,z9
Arctic silver 5,cremique 2
Antec formula 7
Noctua nt-h1..
Just the above are available.i am able to import gelid extreme,arctic mx-4 but it will take 2-3 weeks for delivery.
I currently have noctua, cm tf400, deepcool z5.

 

[crashtech]

If you remove all the adhesive residue from the IHS and the CPU circuit board, there will be no gap between the die and the IHS. As long as TIM is properly applied to the die before re-applying the IHS, this is an acceptable method of reducing temperatures, as you have seen.

 

[GPz1100]

FWIW, i did my setup last november (2012). Delidded, with xpsc raystorm (http://www.xs-pc.com/waterblocks-cpu/raystorm-cpu-waterblock-intel) water block directly ontop of the die. My TIM is Coollaboratory Liquid ultra. I have the water block secured rather firmly to the die. After nearly 10 months and one cooling system flush/refill, my temps are in very close vicinity what they were when I put this whole thing together.

i7 3770k @ 4.6 ghz

Ambient temp ~25C

idle temp @mid 30's C, core one upper 30'sC

(now granted, I normally have ~12 things open so not a true idle, but these 12 applications are more of the static type

Full load (prime95, intelburn, etc.), ~54-57C

Now mind you, i'm using a 360x120mm radiator. This is overkill but does keep the temps very cool at all times.

 

[Idontcare]

Dear IDC please help..
Any one experienced please help me..
Yes i cleaned it with ipa 99%.. Every time.
Right now using noctua tim between die ihs and ihs h80i.
And please tell that will i risk the die if ihs is used above it as there is some space between it and the pcb..
Also please suggest the best tim to use between die and ihs(i am going to keep ihs on).good tim which are available in my country are
Cm thermal fusion 400,essentials e1,e2,x1
Deepcool z5,z9
Arctic silver 5,cremique 2
Antec formula 7
Noctua nt-h1..
Just the above are available.i am able to import gelid extreme,arctic mx-4 but it will take 2-3 weeks for delivery.
I currently have noctua, cm tf400, deepcool z5.

Hi sanurocks,

Did you lap the IHS? (the side that faces the H80 waterblock) What about the H80 waterblock, did you lap it as well?

Regardless, using NT-H1 between the IHS and the waterblock should be good, that is what I use.

However, do not use NT-H1 between the CPU silicon die and the underside of the IHS. The NT-H1 is known to stop working and your temperatures will become hot.

Of the TIMs you listed, I recommend you use Ceramique 2 on the CPU silicon die.

 

[sanurocks]

Sorry for long post.. but i am concerned.
Yes i cleaned everything.glue,previous tim everything.
ok i got the point. there is no risk in running Ihs on the die because the pressure from the mounting bracket adjusts it...

Dear IDC i can get ceramique 2..
but are you disregarding gelid gc extreme due to pump out effect?(Delivery time not being and issue)Also now i checked i can get Phobya HeGrease Extreme,prolimatech pk-3,IC diamond too(but as i can see ic diamond scratched your die).Long story short i can get all the tim except coollaboratory one's and any liquid metal tim.
so of all the tim please suggest the tim which will work good for atleast 4-5 months pumping out taken in account. i will re apply it, as i clean my system(except tim) thoroughly every 2 months.
After finalizing the paste i will do lapping.

 

[Idontcare]

Sorry for long post.. but i am concerned.
Yes i cleaned everything.glue,previous tim everything.
ok i got the point. there is no risk in running Ihs on the die because the pressure from the mounting bracket adjusts it...

Dear IDC i can get ceramique 2..
but are you disregarding gelid gc extreme due to pump out effect?(Delivery time not being and issue)Also now i checked i can get Phobya HeGrease Extreme,prolimatech pk-3,IC diamond too(but as i can see ic diamond scratched your die).Long story short i can get all the tim except coollaboratory one's and any liquid metal tim.
so of all the tim please suggest the tim which will work good for atleast 4-5 months pumping out taken in account. i will re apply it, as i clean my system(except tim) thoroughly every 2 months.
After finalizing the paste i will do lapping.

I'm recommending ceramique2 only because I have used it and no one has reported issue with it on the bare die. The other TIMs may be superior, but I can't say for sure so I can't recommend them.

 

[sanurocks]

i was concerned because in your testing post you have mentioned that it is so thin that it may suffer pump-out effect same as as5..
nonetheless i ordered one.
thank you for valuable input.

 

[Idontcare]

i was concerned because in your testing post you have mentioned that it is so thin that it may suffer pump-out effect same as as5..
nonetheless i ordered one.
thank you for valuable input.

A valid concern, true, but no one to date has reported it with Ceramique so I think you are safe.

Worst case scenario is you find your temperatures getting warmer and warmer over the course of about 2-3 months and you have to pull it apart to clean it and use a different TIM. At this point you've done that plenty of times so it won't take you long if you have to it again.

Too bad you can't get any Liquid Ultra or Liquid Pro. What country are you in that they won't allow it to be imported? (If you don't mind me asking, if you do mind me asking then just ignore this request like it never happened )

edit: doh! I see you already updated your Location field, I see you are in India. Surely the electronic shops can get Liquid Ultra there, are you too rural to get it easily in your home town?

 

[sanurocks]

Actually there is no distributor of coollaboratory products in india..
Local shops and online megastore like amazon and ebay have same inventory just price difference..i live in a developed urban area.
newegg says that clp has been discontinued and 0 pieces in stock..
i think i might get one from frozen cpu the total cost would be $27-$28.it will take 3-4 weeks.. for that time period shall i use deepcool z5..wait for clp and then replace it?
what do you suggest?

 

[sanurocks]

This will be probably my last question regarding this..
Do you have any idea regarding antec formula 7.. Though the performance is on par or better than ic diamond(courtesy many non english review sites)I have one concern, being a nano diamond containing compound will it also scratch the die.?

 

[Idontcare]

Actually there is no distributor of coollaboratory products in india..
Local shops and online megastore like amazon and ebay have same inventory just price difference..i live in a developed urban area.
newegg says that clp has been discontinued and 0 pieces in stock..
i think i might get one from frozen cpu the total cost would be $27-$28.it will take 3-4 weeks.. for that time period shall i use deepcool z5..wait for clp and then replace it?
what do you suggest?

That is a good plan. Get CLU if possible since it is the best of the best, but CLP is OK as well.

This will be probably my last question regarding this..
Do you have any idea regarding antec formula 7.. Though the performance is on par or better than ic diamond(courtesy many non english review sites)I have one concern, being a nano diamond containing compound will it also scratch the die.?

Sorry, no experience with Antec formula 7, but if it has abrasives in it like diamond paste then yes I would be worried about putting that in direct contact with the silicon die.

 

[sanurocks]

ok i will look for ultra..
thanks for the help..
And after spending days and nights of researching as well as your test suggest that realistically any high quality tim would suffice to be used with ihs in place..
bare die its CLU/CLP/any liquid metal no questions asked..
thanks for all your help.

 

[Idontcare]

ok i will look for ultra..
thanks for the help..
And after spending days and nights of researching as well as your test suggest that realistically any high quality tim would suffice to be used with ihs in place..
bare die its CLU/CLP/any liquid metal no questions asked..
thanks for all your help.

Yeah the tests show that pretty much any replacement TIM will work provided you can get it to be much thinner than the stock TIM thickness...but we have seen to many forum members report pump-out issues with both AS5 and NT-H1 that it makes it easy to (now) just recommend people spend a little extra and get the Liquid Ultra or Pro.

But I am sure you'd be just fine with the Ceramique 2 as well.

 

[sanurocks]

i tried re mounting to my bare die the results are good but there is something i dont understand..
my coolest core core0 is running at 59C and hottest core core3 at 74C.. avg package 71C
while when i use ihs core0 runs at 66C and hottst core becomes core1 at 77 avg package being 75C.
all ambient 29C.cpu stock voltage 4.3 GHZ
i dont understand why.. please explain.

 

[Idontcare]

Seems rather hot for stock voltages at 4.3GHz. You may just have a CPU that barely made it through the binning process (meaning it burns more power and requires more volts to function at any given clockspeed).

 

[sanurocks]

Seems rather hot for stock voltages at 4.3GHz. You may just have a CPU that barely made it through the binning process (meaning it burns more power and requires more volts to function at any given clockspeed).

Ok.got the point..
So it has nothing to do with the mounting pressure from the water block beacuse i tried increasing pressure nothing changed. Only the inclusion of ihs changed the numbers...
Btw the stock voltage for my board is 1.236V.turbo overclocked 4.3 ghz.. Prime small fft avx.. Core0 57 core3 66.. Avg package 64.. Previously mentioned result was with IBT max.
Not that great temps but atleast i am getting temps close to stock temps at 3.6ghz.
Next thing to do is change the tim and lap the h80i.
Your posts are very informative IDC.. Thank you

 

[Idontcare]

Happy to help

Btw the stock voltage for my board is 1.236V.turbo overclocked 4.3 ghz..

That voltage is the problem, quite high for a 4.3GHz OC.

Have you tried lowering the voltage?

Not directly comparable, but you can see with a 3770K the 22nm process gives a nice low voltage requirement at 4.3GHz (just 1.1V).

Next thing to do is change the tim and lap the h80i.

Exactly.

Have you had much opportunity to do lapping? I posted about lapping my H100 (and NH-D14) in this thread here, it may give you some ideas to try. (and an earlier lapping thread on the 3770k is here, but not as informative IMO)

 

[sanurocks]

i have tinkered with the voltages i can lower them a bit it passes the test but there are some whea crc error.. and as i can see from the table, the temperature delta are not much different.. then again i will lower it again and try.
also i use turbo overclocking if i used fixed multiplier i can reduce it quite a bit but it consumes quite the power almost 30W idle..but now 8-9W idle.
No i have'nt tried lapping before.. i have seen your posts.. i will first try it on some copper blocks and then go to my h80i when i am confident.
(And one thing i have to tell you deep cool Z5 works extremely well on par or better than noctua and it fills the gap quite good.. i will report if it suffers from pump-out)
my current result:
3570k Vcore 1.212 avg core temp 68.5 room temp 28 Delta:40.5 test used:IBT
i think that is right in the ballpark for h80i stock fans max non-lapped.
what do you say?