I know what you mean. Believe it or not, I had the exact same issue as you back in 2013. Got my hands on a 3770k and decided to retire my 2600k, but I quickly saw a huge temperature increase at lower voltages and speeds. In addition I noticed that my water temperature actually got reduced, which was a clear indicator that it wouldn't overclock well at all. I wasn't quite as comfortable with modding at that time as I am today, so I sold the Ivy and rolled back to my dear Sandy 2600k.
If it had been today the Ivy would've gotten a delid without hesitation. The assembly of those chips are quite horrendous, and RMA is not an option when it's keeping below 85°c with stock cooler at stock speeds. The worst part is just as you describe. I got a friend with a 3770k, and that cpu gets to 4.6GHz with a Noctua NH-U12p, while my H100i couldn't manage 4.5GHz even at PnP @ 2500rpm.
Be sure to not damage your motherboard by over tightening the block. With it tightened, can you wiggle the block by hand, or is it solid with the mobo?
I have the same exact pump, and with two rads and four blocks it gets around at the lowest speed setting(Dial set to 1) without any issues. At that speed my 3960x hottest core runs about 7°c/23°c Idle/Load over water temp, and my GPUs are pretty much the same. 7970 is a couple of degrees shorter, and the 7990 is a few higher vs the CPU.
The fact that the GPU is much cooler just enforces my suspicions. The 780 conducts over three times more energy in less than half the surface area vs the 3770k, which would mean that even if the CPU ran at similar temps as the GPUs, it would still raise a lot of suspicion. If the CPU would act as the GPU, the temps would be closer to the water temps than the GPU temps.
I'm sorry, but with your temps and description it's very unlikely to be anything but the TIM in the CPU. Ivy can be a real hassle and disappointment without delid...
It does make sense if you think about it. Conducting heat is much like conducting electricity, where water is much better than air, but outright sucks compared to copper. Chocolate is worse than no paste at all, just like styrofoam would be. In addition you need to view the link between the CPU and the cooler as a bulk, not as a single material. I'll elaborate.
In the scenario of "No thermal paste", there is partial conductance between the heat spreader and the cooler, which is metal to metal, and the rest is air. In the scenario with chocolate as paste, you'll have nothing else other than chocolate to transfer the heat.
Then we'll look at the proportional difference between the materials.
(W/mK)
Copper - 400
Aluminium - 250
Quality TIM - 4-7
Water - 0.5
Wood - 0.1
Air - 0.025
By this you can see that just about anything is better than 100% air, but you'll also see that even partial contact by metal will be significant, even at just a few percent of the surface area.
The conclusion by this should be that the bulk of air and metal is better than chocolate, but far worse than just about anything else
Just use the stock TIM. The difference between MX-4, AS5, Stock and Icy Diamond are minimal at best.
If you're crazy like some of us are, you'll check out Indigo Xtreme. A metal that liquidizes at burn in and acts like soldering tin, increasing the thermal conductivity two to four times vs the best of pastes.
Sandy_Dandy got a reaction from Bab22 in RATE THE PC SPECS ABOVE YOU :)
Sandy_Dandy got a reaction from Denoxx in Custom waterloop- IvyBridge - Temperatures pretty high
Sandy_Dandy got a reaction from Trik'Stari in Best watercooling thermal paste?
Sandy_Dandy got a reaction from MahDays in Best watercooling thermal paste?
