We all know that both Ivy Bridge and the Z77 PCH run hotter than the P67 and Sandy Bridge. The reasons for this are not completely understood (yet) but there are some explanations out there. One of these appears to be due to the process. At 22nm there is going to be an increase in power density and also a decrease in the surface area that allows for cooling of the CPU core. The other opinion is that Ivy Bridge is hotter due to the use of Thermal Interface Material (TIM) rather than fluxless solder.
The latter explanation was unveiled by overclocker’s.com and was quickly picked up by multiple sites around the internet. This has led to the belief that this is the real answer to the heat issues and why Ivy Bridge does not overclock as well as Sandy Bridge. Of course there were still questions; why were there no de-shelled comparisons? Once you get the cover off why not test Ivy Bridge again to see what the temps are? If the TIM is the cause you will know right away. Of course there are dangers to doing this though. You run the risk of damaging your CPU just with the heat spreader removal and then you can kill it if you do not carefully mount your cooler.
Fortunately for us royalk over at pceva.com.cn (translated link) took these risks to confirm the issue. It is something like an episode of myth busters… The test system that royalk used is listed below;Test Platform:CPU:Intel Core i7-3770KMemory: PCEVA Extreme Kit DDR3-2133 7-10-7Motherboard: MSI Z77A-GD65Video card: MSI R6570 MD1GD3Hard drives: Plextor PX-128M2PPower supply: Enermax Revolution 85+1050WRadiators: Noctua NH-D14Room temperature: 28 ° cGrease: Prolimatech PK-1 (10.2/mK)
After verifying the temps royalk went ahead and removed the IHS. He performed the same tests with the now naked CPU. There were some things that had to be different. Without the IHS the lid for the CPU socket does not hold the CPU in place (the CPU core is actually a few mm below the top at this point). However, royalk made sure there was good contact between the base of the Notcua NH-D14 and the top of the exposed Ivy Bridge CPU Core.
Surprisingly the idle temperatures (at the same 4.5GHz) were 34/33/43/34. Looks very similar doesn’t it? The load tests were run again and the results were very interesting indeed. The high temp was 80c and the average for each core was 67.0/71.4/76.3/68.0. Guys this is nearly identical to the temps with the TIM and the IHS. In the end the higher temps found on Ivy Bridge are not due to the use of thermal interface material. It really seems like the higher power density combined with the new process (a higher leakage rate can mean higher voltages and more heat as well). We are sure that there will be new steppings of Ivy and with each new one the process will mature and with it the overclocking potential of the CPU. We have to remember that by the time Sand Bridge hit the market the 32nm process was very mature. Even the initial ES chips were able to hit 5GHz without much trouble while the later steppings can reach 5.2 and up. We will see this with Ivy Bridge, we just have to give it time. After all the first 32nm Westmere CPUs were not that great at overclocking either (remember Clarkdale?).
Discuss this in our Forum