- Joined
- Apr 2, 2011
- Messages
- 2,937 (0.58/day)
Let's logic this one out...because I think there's a ton of things to improve here. I'm going to start by the general layout....and try to understand the logic.
CPU - Radiator (top) - Radiator (back) - Radiator (bottom) - Pump/Res - Radiator (front) - GPU - CPU
1) This is...just wow. So, your reservoir/pump is about midpoint on the height of the loop, meaning that it's got to both pull fluid up and push fluid down... That's a real no-no...you want the pump as low as possible so that it will not run dry period. If you have it both sucking and pushing you invite cavitation...which is its own issue.
2) The amount of radiator here is...silly. Think this through for just a moment. Heat flow is from area of high heat to lower heat. This means that by placing the two heat creating parts together you are going to have less flow (delta in temperatures low, heat flow low). You're going to have problems measuring this because loop temperatures and power limits are low...along with relatively high error margins. I...just don't think that this was thought through....because it seems more like an "add more radiators will make it cooler" type of ad-hoc configuration.
3) Let me math this out for you...using easy numbers. CPU generates 200 Watts and GPU creates 400 watts. Theoretically at stable state you get 600 Watts into loop and 600 watts out of loop...but the problem with both heat sources next to one another the fluid temperature has to rise much higher to support that. The counter is that...because of numbers you don't see it. To raise one liter of water one degree C you need 4180 Joules...1 watt = 1 Joule/Second, so to raise a fluid loop 1 Liter/minute flow you'd need to dump 4180/60 = 69.67 Watts of heat into the system. Yep, that 600 Watt heat source should only have you increasing 8.6 degrees C with a liter per minute flow (and no radiator). Hopefully it's easy to see then that 1500 liters/hour, or 25 liters per minute (25 times faster than our calculation, for a D5 pump) basically yields such low fluid temperatures as to be silly. Yes, that 600 Watt thermal load at a D5 pump rate of 25 Liters per minute, would amount to a fluid temperature of 0.344 C assuming there was no radiator to release energy from the system. Throw in the assumption of radiators existing...and you being at thermal equilibrium, and you've got that 1 liter of fluid heating up with 600 Watts, flowing into a radiator, cooling down 600 Watts, and zero net temperature change in the loop.
4) Does the math make sense? The average water heater is about 30 gallons, and 50000 BTU/hr = 14654 Watts. 14654/600 = 24.3.....so a water heater meant to drive water to about 49 C....a delta of about 25 C, requires about 24x as much energy... The flow rate for a shower head is 7.9 Liters/minute...about 1/3 the rate of a cooling loop, but supplemented by a heated tank that can run out of water and leave you with a cold shower. Yep...the rough acid test makes sense.
5) Let me finish with a construction note on radiators. They're generally a chamber, tubes, chamber, tubes, etc... As long as the flow rate is relatively low those chambers shouldn't have a problem with some off-gassing, and they should absorb flow rate changes. Thing is, if you suddenly start having to pump to fill those chambers you develop a reverse pressure. The pumps designed for PC cooling really don't tolerate this well, so they tend to suck air from anywhere they can to fill what is effectively a temporary gap. What does that even mean? Your 2nd and 3rd pictures show where chambers might be above the level of other radiators...meaning that if you want to push fluid they literally have to reverse fill...which tends to pull bubbles. That reverse pressure is causing the noise...which is basically your pump fighting a vacuum because of poor design.
If I could recommend four things to you, it'd be as follows.
1) Remove two of your radiators. My money would be the front and bottom.
2) Replace the bottom radiator with a large tank. I run a more aggressive loop, but my pump can suck about half a cup from the tank before the loop flow equalizes. Your current tank is woefully undersized for that...and definitely should not be running such a large loop.
3) Place the pump on the bottom of the case. Wet pump is happy pump...and it will make sure that cavitation is much less likely during speed adjustments.
4) Restructure the loop to be something like GPU - Radiator (Top) - CPU - Radiator (Back) - Reservoir - Pump - GPU. It's not going to be a huge difference...but look at the numbers. A 600 Watt load is only 1/3 degree C of delta temperatures to be stable. This is often why people think there's no difference....because the difference of several hundreds of Watts is required to make a number large enough to matter when most tools measure to an accuracy of +/- 0.5 degrees...
If I could ask the obvious...is this your first loop? I'm asking because the first thing everybody thinks is more radiators = lower temperatures...and that's patently wrong. I know my first foray into water cooling saw a double height 360 radiator running a 120 Watt chip and I skipped out on the GPU...and I basically hit the thermal limitations on my CPU because running at maximum pump speed still has to answer to the laws of physics...which were confined to how fast the CPU die could actually conduct the heat into the thermal loop (read: way slower than a double height 360 radiator could reasonably cool). This design seems like that same kind of first time zeal which often costs us too much money...when a couple of 240 radiators would have been more than enough at 1/5th the price. The truth is more radiators = more heat capacity...but with only a few hundred Watts you need higher contact areas and large fin areas far more than additional time to transfer energy at low rates (due to low deltas in temperature).
CPU - Radiator (top) - Radiator (back) - Radiator (bottom) - Pump/Res - Radiator (front) - GPU - CPU
1) This is...just wow. So, your reservoir/pump is about midpoint on the height of the loop, meaning that it's got to both pull fluid up and push fluid down... That's a real no-no...you want the pump as low as possible so that it will not run dry period. If you have it both sucking and pushing you invite cavitation...which is its own issue.
2) The amount of radiator here is...silly. Think this through for just a moment. Heat flow is from area of high heat to lower heat. This means that by placing the two heat creating parts together you are going to have less flow (delta in temperatures low, heat flow low). You're going to have problems measuring this because loop temperatures and power limits are low...along with relatively high error margins. I...just don't think that this was thought through....because it seems more like an "add more radiators will make it cooler" type of ad-hoc configuration.
3) Let me math this out for you...using easy numbers. CPU generates 200 Watts and GPU creates 400 watts. Theoretically at stable state you get 600 Watts into loop and 600 watts out of loop...but the problem with both heat sources next to one another the fluid temperature has to rise much higher to support that. The counter is that...because of numbers you don't see it. To raise one liter of water one degree C you need 4180 Joules...1 watt = 1 Joule/Second, so to raise a fluid loop 1 Liter/minute flow you'd need to dump 4180/60 = 69.67 Watts of heat into the system. Yep, that 600 Watt heat source should only have you increasing 8.6 degrees C with a liter per minute flow (and no radiator). Hopefully it's easy to see then that 1500 liters/hour, or 25 liters per minute (25 times faster than our calculation, for a D5 pump) basically yields such low fluid temperatures as to be silly. Yes, that 600 Watt thermal load at a D5 pump rate of 25 Liters per minute, would amount to a fluid temperature of 0.344 C assuming there was no radiator to release energy from the system. Throw in the assumption of radiators existing...and you being at thermal equilibrium, and you've got that 1 liter of fluid heating up with 600 Watts, flowing into a radiator, cooling down 600 Watts, and zero net temperature change in the loop.
4) Does the math make sense? The average water heater is about 30 gallons, and 50000 BTU/hr = 14654 Watts. 14654/600 = 24.3.....so a water heater meant to drive water to about 49 C....a delta of about 25 C, requires about 24x as much energy... The flow rate for a shower head is 7.9 Liters/minute...about 1/3 the rate of a cooling loop, but supplemented by a heated tank that can run out of water and leave you with a cold shower. Yep...the rough acid test makes sense.
5) Let me finish with a construction note on radiators. They're generally a chamber, tubes, chamber, tubes, etc... As long as the flow rate is relatively low those chambers shouldn't have a problem with some off-gassing, and they should absorb flow rate changes. Thing is, if you suddenly start having to pump to fill those chambers you develop a reverse pressure. The pumps designed for PC cooling really don't tolerate this well, so they tend to suck air from anywhere they can to fill what is effectively a temporary gap. What does that even mean? Your 2nd and 3rd pictures show where chambers might be above the level of other radiators...meaning that if you want to push fluid they literally have to reverse fill...which tends to pull bubbles. That reverse pressure is causing the noise...which is basically your pump fighting a vacuum because of poor design.
If I could recommend four things to you, it'd be as follows.
1) Remove two of your radiators. My money would be the front and bottom.
2) Replace the bottom radiator with a large tank. I run a more aggressive loop, but my pump can suck about half a cup from the tank before the loop flow equalizes. Your current tank is woefully undersized for that...and definitely should not be running such a large loop.
3) Place the pump on the bottom of the case. Wet pump is happy pump...and it will make sure that cavitation is much less likely during speed adjustments.
4) Restructure the loop to be something like GPU - Radiator (Top) - CPU - Radiator (Back) - Reservoir - Pump - GPU. It's not going to be a huge difference...but look at the numbers. A 600 Watt load is only 1/3 degree C of delta temperatures to be stable. This is often why people think there's no difference....because the difference of several hundreds of Watts is required to make a number large enough to matter when most tools measure to an accuracy of +/- 0.5 degrees...
If I could ask the obvious...is this your first loop? I'm asking because the first thing everybody thinks is more radiators = lower temperatures...and that's patently wrong. I know my first foray into water cooling saw a double height 360 radiator running a 120 Watt chip and I skipped out on the GPU...and I basically hit the thermal limitations on my CPU because running at maximum pump speed still has to answer to the laws of physics...which were confined to how fast the CPU die could actually conduct the heat into the thermal loop (read: way slower than a double height 360 radiator could reasonably cool). This design seems like that same kind of first time zeal which often costs us too much money...when a couple of 240 radiators would have been more than enough at 1/5th the price. The truth is more radiators = more heat capacity...but with only a few hundred Watts you need higher contact areas and large fin areas far more than additional time to transfer energy at low rates (due to low deltas in temperature).
.
.
