I am a new 8350 owner, just put the system together last night. I'm using the stock cooler and the stock clock of 4.0ghz and 4.1ghz turbo and running BOINC all 8 cores at 100% for hours, the highest temp is 60C
This is a bit warm but considering the load I think well within the safe range, and a very respectable temp for basic cooling.
Disableing BOINC it quickly drops town to 25-30C
Your results are normal for an FX-8350. The OP's H-100 has installation or operational issues as it too should deliver similar temps around 60C or slightly lower under full load P95 type stress testing, IME. AMD has yet to post the max operating temp for the FX processors released after the FX-8150 which has a 61C max operating temp. IME operating the Vishera FX CPUs above mid 60's C can result in BSOD, reboots, etc.
As previously indicated in a separate thread, AMD should be posting the CPU core max operating temp and P zero state voltages under the individual specifications for each processor model on the AMD website so that PC owners can properly determine the operating conditions of their AMD processor. The vcore voltage supplied on many mobos is poor IME, particularly on 8-core AMD CPUs which draw a lot of power. Excessive vcore voltage has been confirmed to cause CPU overheating conditions even with proper CPU cooling. Socket temp especially on 8-core CPUs can vary drastically depending on the duration of the load and mobo cooling. The CPU core temp responds quickly to changes in processor load and is the best choice for monitoring the CPU's thermal state. 70C core is where internal throttling of most newer AMD desktop type CPUs begins as to prevent destroying the CPU from overheating.
I disagree, I have a Aquacomputer water block with a constant intake temp of 30oc. From what I've seen with my 8350 is the CPU Core will react fast, but the socket/AUX will be much more likely to hit the warning zone and hang around. Where's the CPU Core will bounce around, not suggesting anything getting too hot. Your also looking at the delta of the water flow as well, if the water is flowing twice as fast as idle, the core will notice it first and reduce a fair bit in comparison. It should be pointed out to that anyone using a water block rather than air cooled should have something cooling the VRMs so that area of the board isn't as effected by the highs and lows of other components really. And especially on water where the pins will suffer more damage from overheating well before the chip does over time. Unless you pump stops.
I think CPU Core is just to control the fan PWM signal. As the algorithm works like you stated, if its under load, and or the chip is hot, the fan will work harder. But I wouldn't consider it consistent enough to relay on. I use it to determine the speed of my pump and it does a good job. Plus its a pain explaining to people that its impossible to have a CPU idle at 10oc.
The reason why your socket hits the "warning zone" is because you don't have good processor socket cooling and the socket doesn't have a heatsink or a fan blowing on it like a normal HSF provides. (Most VRMs also need fan cooling which a HSF can and often does provide).
The CPU core temp changes directly with thermal load and that is what is critical not the socket temp. All CPU/APU controls for frequency, power consumption, fan speed, etc. are based on CPU core temp because the core temp is what is critical and all that really matters. Socket temp is just a byproduct of the CPU's operation and can drift all over the place as you have witnessed with liquid cooling vs. a std. HSF. Maintaining a proper max CPU core temp prevents damage to the CPU/APU and a drop in operating frequency. It's also worth noting that the socket temp "danger zone" is much higher than the CPU core temp. That's why it is important to know what the max CPU operating temp is supposed to be and actually is so that you can maintain your CPU below the max core temp to prevent system hangs, BSOD, rebooting, etc. Socket temp won't have any influence on system performance unless it's ridiculously hot.
But this is water cooling. With only the CPU cap and the thickness of the waterblock wall as the convection level. Where's as on air, convection is much greater due to heat pipes or just the fact they are massive. The chip itself will always be cooler on water in comparison due to this.
Providing the water block is well made and has good heat conductivity, the chip will always be less hot during these moments of long heat generation. So with watercooling, its the socket. If you turn your pump off or down to a trickle, say you have a 5oc difference from water in to out. Then the core would make sense as the delta of the water flow will be much less. Much less heat removed. But if you have the pump working at full power, the water passes over it twice in the same amount of time. So CPU Core is much more susceptible for that than the Socket. I don't think anyone cools the socket directly.
Its all relative, the pins will be conductively connected to the clip, where as the chip itself is directly connected to the waterblock. The coolest part of the chip will be the CPU cap. The back of the processor with no cooling is where heat cant escape. If your CPU Core is hitting the warning or danger zone before the socket then there is something wrong with the water block. I can see where you getting at.
You was right. My understanding and calculations where wrong.
Aquacomputer have confirmed it is the Core, its the most important part.
The debate was fun while it lasted, and I've leaned something new