Originally Posted by PeL4
I have a x6 1075t. OCed to 4ghz right now, used to have it @ 4.2 but with these temps...
is too dangerous to run it to that speed.
You should read the thread, as I have posted some nice pics of my loop and temps readings. I tried to post all the data so you can help me the best you can.
For now, I'm going to try reseat the cpu block. But I don't know what to do different from last time. Maybe I tighten the screws too much?
Right on, I went back and read most this thread, and it looks like 1 of a couple things... but this has already been addressed so your mostly just going to get it again.
1) Your ambient temps in and around 30c air, is too high, if your air is 30c than your water will be 30c at the lowest possible temp, which wont last long.
Quick refresher: Thermal conductivity works on 4 levels (from my concern with h20 cooling standpoint)
A) mass vs. mass, thermal energy has a tendency to travel from the least amount of mass (a small chip) to a greater mass (a big radiator/block), you have plenty of mass so its not that. I bet theres almost a half gallon in that loop.
B) contact area, most objects in this reality make at and around 2% contact from surface to surface at the microscopic level. This is improved in our case by very polished blocks (and lapped processors) which have the material/crystals "worked" into a single grain direction (the more flat a surface is the more luster/photon reflectivity it "tends" to have) so when we mate our thermal object die to thermal conductive block we use both pressure from a mounting solution, and correct TIM/method for that
Forget the argument about which method to use, because the method is flexible to the application. Although too much will tend to resist thermal conductivity. That all said, this bond
translates "in my experience" to a rate, or "faster" temp surge and drop in a good loop. ie. the core goes under load, you get a temp spike, and subsequently that temp drops equally as fast, thus the contact area could be said to be very good as temps are "equalibriated" quickly in the thermal system. Poor contact = poor conductivity and core temp will tend to graph the cpu as "holding" onto the heat once the load is taken off it. I really do not think this is your problem for'se although arctic silver 5 takes around 200 hours to properly set, at which time your temps will be erratic. MX-2 sets much quicker, but is also a little thicker.
3) flow rate/turbulence dictates the amount of material and material exposure that the atoms of Cu and molecules of H2O come into contact with and than are taken away from the block/heated area in a given unit of time. Both of which are able to be graphed and both of which dictate to some degree loop performance, I dont see this being your issue either.
4) Delta T, ahh the big one... temperature difference, I like to visualize temperature difference as being 2 big fish tanks connected by a pipe. If I put 30 liters into one tank and 30 liters into the other tank, there is little to no movement, as they are the same amount. reduce one of the tanks to 20, and 5 liters from one tank will move to the other tank, resulting in a equal state of 25. One thing to keep in mind though is that the greater the difference between the two systems the greater/faster the flow will be between the two systems, and as these systems equalibriate between each other the slower the movement becomes. Its a curve relationship rather than linear. This is why when we "test" our water cooling systems it is generally a good practice to "heat up" the loop before taking measurements. Idle for 15 is normally good enough.
In h20 systems we have 2 deltas, the one between the die and the block and the one between the h20 in the loop and the outside air.
Knowing these things we are able to analyze your loop.
I think your build and block installs look decent... its not "that" big a deal. Your setup and build are good, looks like you took a lot of care, I very seriously doubt your dropped the ball on something that critical. TIM/mounting.
You have plenty of material, 2 rads lots of metal, lots of h20... its not that...
the OAT (outside air temp) is high, so your difference between systems is poor, this will result in very little "headroom" when trying to cool something. Its always a split, with diminished returns as we approach equalibrium.
You seem to be "holding onto" heat in the rads... so lets see, I mentioned that in 4, I figure this to be the biggest offender in your setup. A radiator does just that, radiates heat like my fishtank example above, if the case temp is in the 30's and the rad is in the 30's, thats it... theres nothing to radiate, all the air molecules and h20/metal molecules are at equal vibration.
A fan, just like I mentioned in 3 adds to turbulence/flow of air which assist in the rate at temperature change, but there has to be a temperature difference or there is no way to exchange vibrating molecules to less vibrating molecules.
Couple options here...
modify the case to have push/pull fans that blow the air out and away from any other part of the computer, so that you are not "recycling" heated air
move the radiators away from the computer, hell move the radiators to the outlet vent of your home A/C system and watch the temps plummet
modify the case to have 200+ mil fans in the side wall to draw air in and "flush" the system with some "not as heated" air... keep in mind any "fan" that sucks air into a case adds wattage to the air as the fans cool themselves with the very air they move
lower the oat to below 30c Every degree lower in OAT should translate to a .25->.5->1>1.75 lower temp as the difference in temps expand between the h20 and oat, assuming your not running into a stuffy case setup.
Even at that your temps are not that bad
considering the environment, but the environment is poor for oc'ing a pc.
Could also think about a pelt/tec/water tower-evaporator... or even window A/C unit, and running a rad to it... worse case scenario... back that oc off...
in your case, a water evaporator giving the environment, could be a good solution...