2 x 780Ti 4770k heat Dissapation

Sin_Chase

Sin_Chase

Sin_Chase

Soldato
Joined
13 Jan 2004
Posts
21,261
Sup WC.

Recently added a second 780Ti and a 4770k into my loop. Have undervolted the CPU to 1.15VCORE with a 4.2GHz clock speed. 780Tis are running +200 and +225 on Core and Mem respectively. Once the loop temperature stabilises I am seeing 65-75 Deg C on the CPU and 55 Deg C on the GPUS whilst gaming. GPU temps are fine but not happy with the CPU ones.

Trying to work out if it's a lack of dissipation in the loop or simply Haswell being lame. Tempted to delid but not if it's the loop rather than the IHS/TIM on the CPU. Idles happily at 29Deg C so no mount issues I can see.

Using thinner XSPC EX480 rads x 2 with SP120s (Top and lower mount in a 900D) Whacked fans to full speed while testing but that made little/no difference. The EXs prefer low RPM anyway.

Any ideas? The loop should be plenty for the hardware in my opinion. The water temps seem okay seeing as the GPUs happily sit at 50-55Deg C under full load, perhaps just the CPU IHS/TIM not radiating enough heat into the loop?
 
Last edited:
Afraid it is haswell being lame. The CPU if it could transfer heat to the loop as well as the GPU's would run as cool or more likely, cooler but unfortunately it cant.

Could lap the card and replace the TIM with liquid ultra, thatl knock it down a fair bit without needing to de-lid.
 
Been considering delidding.

Is Liquid Ultra a necessity or can you use 'regular' TIM between the die and IHS? I appreciate Liquid ultra will provide better performance but also with added conductivity risk.
 
If you go that far may as well do it right and get Max you can.

Paint varnish over VRM's and do not apply it like a 5 year old at art class. :p
 
Liquid electrical tape works the same way but is a bit more durable and is especially for this sort of thing.

Reason why i suggest CLU between lapped chip and heatsink is, once lapped it has much bigger surface contact with the block allowing for very thin effective TIM's to shine. CLU has practically the same effective heat transfer as the block itself, so you can count a lapped chip with CLU as only having grease between IHS and chip, as the CLU will give similar temps to it being soldered contact.

Regular TIM can be used but if conductivity is bothering you so much, i would avoid de-lidding all together and see what you can get away with by lapping first. I have previously used a line of MX-2/4 between chips and nearby components to avoid conductive TIM like CLU from spreading that way but now i dont bother and just use liquid electrical tape. This is the same stuff many extreme overclockers use to avoid shorting caused by condensation. It is available in a pot form (like nail varnish or tippex) or as a aerosol spray to cover circuit boards and such. In this case, you obviously want the 'paint on' rather than 'spray on'.
 
Sin_Chase said:
When you say lapped chip, do you mean CPU Block, IHS or both?
Click to expand...
Both lapped is ideal. But on a haswell chip, delidding and putting coollaboratory liquid pro/ultra on the bare die is the most important thing.
 
Last edited:
Waterblocks are flat, so they dont need to be lapped, as you dont have anything to gain.

IHS is concave and lapping will allow more contact and far less paste to be used, also allows the use of a finely painted amount of CLU.

De-lidding and using CLU on a naked die produces the best results but lapping a chip doesn't require de-lidding or anything people would consider particularly risky (not that de-lid is risky). You still can stand to gain 4-10 degrees from lapping the IHS and using CLU over other pastes. If you don't mind chucking the warranty on the chip, i would do this before a de-lid, if you feel uncomfortable with applying TIM to a naked die due to nearby components.
 
Last edited:
I'm not uncomfortable, but I like to weigh up my risk versus reward :p

Going to burn in the CPU and other components for a while then pull the CPU block off, see what kind of spread I am getting then make a decision.
 
Avenged7Fold said:
Waterblocks are flat, so they dont need to be lacked, as you dont have anything to gain.

IHS is concave and lapping will allow more contact and far less paste to be used, also allows the use of a finely painted amount of CLU.

De-lidding and using CLU on a naked die produces the best results but lapping a chip doesn't require de-lidding or anything people would consider particularly risky (not that de-lid is risky). You still can stand to gain 4-10 degrees from lapping the IHS and using CLU over other pastes. If you don't mind chucking the warranty on the chip, i would do this before a de-lid, if you feel uncomfortable with applying TIM to a naked die due to nearby components..
Click to expand...
20c is not uncommon on a deliddec haswell mate, mine at 4.5ghz, 1.2700v will hit 96c in p95 small fft. Ive lappedf cp-u's before, q6600 @3.8ghz under a TRUE cooler, dropped it from 81c down to 68c. Haswell is a different animal
 
Fully aware of the gains by de-lidding but if someone is hesitating over using CLU on a naked die due to nearby components, giving other solutions(even if they are not as effective temp wise) is still useful to OP.

If Chase is weighing up benefits, its useful to have other options. His current temps are not too bad as they are, so a moderate drop might suit him just fine.
 
Avenged7Fold said:
Fully aware of the gains by de-lidding but if someone is hesitating over using CLU on a naked die due to nearby components, giving other solutions(even if they are not as effective temp wise) is still useful to OP.

If Chase is weighing up benefits, its useful to have other options. His current temps are not too bad as they are, so a moderate drop might suit him just fine.
Click to expand...
Tbh mate i was even wary of using clu on an ihs on my previous 3570k and 3770k. And they both had warranty. Current 4770k has no warrenty, so delidding and lapping is a moot point. But tbh im still very wary of delidding. If i do it will be the hair dryer method.:)
 
Was planning to do my own 4770k but until i see a bottleneck, not much point for now.

Will probably lap next time i break my loop down and if i manage to grab a nice clocking spare, will probably choose one of the chips to de-lid for the sake of curiosity.
 
Avenged7Fold said:
Was planning to do my own 4770k but until i see a bottleneck, not much point for now.

Will probably lap next time i break my loop down and if i manage to grab a nice clocking spare, will probably choose one of the chips to de-lid for the sake of curiosity.
Click to expand...
Good clocking 4770k's are as rare as hens teeth, i dont trust auction sites and theyre rare in here.
 
I'm running the same rads as you plus two 240 black ices. My X79 system with CPU 4960 @ 1.4v and 3 GPUs at 1.15v @ 1150/3850 non boost

2X D5 pumps

XSPC case fans @ 1200RPM
4960 @ 59C underload

3x 780Tis: max temps 35-39-41c

Might not be all that helpful, and Avenged IMO offers some of the best W/C advice on here! But food for thought.
 
Check my build log.

I delided, and running with IC Diamond between the CPU and the WB, no more IHS involved.

My CPU at 4.6 gets max 29c or so, playing games.
On Prime I think it went up to around 55c.
 
Running 4.2 at the moment which is plenty, no bottle necks etc. Dropped my VIN to 1.7 and 4.2 is still stable at 1.15VCORE, might be able to lower that more. Helped lower the temps a little.

Will run like this for a while and consider my options when I next look at a drain. The temps are not going to kill the chip but I always prefer to keep them as low as possible.

A little frustrating as I think I have a pretty good clocker. Can boot to Windows and run AIDA64 with 1.2Vcore at 4.6Ghz.
 
Update - Delidded and used CLU. Running a quick video encoding test right now, I estimate I have shaved between 10 and 15 Deg C off temps. Running an x264 encode, all 8 threads loaded 100%, 40-43 Deg C average across all 4 cores.

Was bricking it at every stage doing it though!

Need to fire up BF4 and load both GPUs and CPU and see what temps I get after a few hours compared to before.
 
Avenged7Fold said:
Congrats, you didnt hold out long lol

What did you use in the end?
Click to expand...

Haha nope - ordered CoolLabratory Liquid Ultra while it was in stock with the idea I will have it available when I want to delid.

Arrived yesterday and I thought.... "If I can get CPU block off without draining, I'll just get it done".

I could get block off without draining, so went for it.

Hammer and vice method - very easy and far more comfortable than blade approach.

Used an old plastic gift card to remove the glue and taped some paper over the iVRMs as a mask while I applied the CLU to the die. Considered some liquid electrical tape or plain old nail varnish but decide I would just be extra careful on the CLU application than risk damaging iVRMs.

Stuck with IC Diamond TIM between IHS and block.

Took my time and fortunately I've not killed it :D
 
You must log in or register to reply here.
Back
Top Bottom