Pro/Forums - View Single Post - Shouldn't water cooling Temps = 1 degreee more than air?

Cathar · 03-18-2003, 06:26 PM

Things that affect water-cooling's performance:

1) The thermal paste interface between the CPU and the waterblock. This is proportional to the area of the CPU is is typically around 0.1 C/W for a 100mm^2 die size (and consequently around 0.08C/W for a 125mm^2 die size like an AthlonXP). For every watt of heat that a Palomino AthlonXP CPU dissipates, the CPU will rise by around 0.08C. A really really hot Pally XP can dissipate up to around 85W (real watts), meaning that the CPU will be at least 7C above ambient merely due to the thermal junction between the CPU and the waterblock alone.

2) The waterblock's native cooling performance - this hinges on a few factors such as flow rate and design, with the best blocks offering C/W's of around 0.07 for a 100mm^2 die size, and as bad as 0.25 for a poorly designed block with low flow through it. We can't really relate this to die size in actuality, but for simplicity sakes let's do it anyway. Assuming you used a Pally AthlonXP in a highly overclocked full-load scenario using say a very good block with a decent pump, you might be seeing a C/W of around 0.10 for the block and the AthlonXP sized die. ie. the CPU will climb to around 8.5C hotter due to the waterblock's performance. Note that the waterblock's performance is affected by the temperature of the water flowing through it.

3) Pump heat - the pump will dump heat into the water which needs to be dissipated by the radiator. A typical pump will add around 5-10W of heat to the water. This is handled by 4) below.

4) The radiator then needs to cool the water. A very good large radiator with high speed fans will offer a C/W as low as 0.02, but realistically a 0.05 C/W is a far more common story for a noise bearable radiator that fits inside most people's cases. Given the 85W that the CPU puts off plus the 10W of the pump, and multuply that by 0.05, and we get water temps that will sit at around 5C above ambient.

Add all that up and we get 7 + 8.5 + 5 =~ 20.5C CPU rise above ambient for a very hot CPU.

Just all rough and simplistic of course. There are quite a few other factors operating here (motherboard heat/motherboard heat dissipation - commonly called secondary heat paths), radiator intake temperature which can be higher than ambient especially if the radiator is mounted in-case, etc.

03-18-2003, 06:26 PM	#4
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	Things that affect water-cooling's performance: 1) The thermal paste interface between the CPU and the waterblock. This is proportional to the area of the CPU is is typically around 0.1 C/W for a 100mm^2 die size (and consequently around 0.08C/W for a 125mm^2 die size like an AthlonXP). For every watt of heat that a Palomino AthlonXP CPU dissipates, the CPU will rise by around 0.08C. A really really hot Pally XP can dissipate up to around 85W (real watts), meaning that the CPU will be at least 7C above ambient merely due to the thermal junction between the CPU and the waterblock alone. 2) The waterblock's native cooling performance - this hinges on a few factors such as flow rate and design, with the best blocks offering C/W's of around 0.07 for a 100mm^2 die size, and as bad as 0.25 for a poorly designed block with low flow through it. We can't really relate this to die size in actuality, but for simplicity sakes let's do it anyway. Assuming you used a Pally AthlonXP in a highly overclocked full-load scenario using say a very good block with a decent pump, you might be seeing a C/W of around 0.10 for the block and the AthlonXP sized die. ie. the CPU will climb to around 8.5C hotter due to the waterblock's performance. Note that the waterblock's performance is affected by the temperature of the water flowing through it. 3) Pump heat - the pump will dump heat into the water which needs to be dissipated by the radiator. A typical pump will add around 5-10W of heat to the water. This is handled by 4) below. 4) The radiator then needs to cool the water. A very good large radiator with high speed fans will offer a C/W as low as 0.02, but realistically a 0.05 C/W is a far more common story for a noise bearable radiator that fits inside most people's cases. Given the 85W that the CPU puts off plus the 10W of the pump, and multuply that by 0.05, and we get water temps that will sit at around 5C above ambient. Add all that up and we get 7 + 8.5 + 5 =~ 20.5C CPU rise above ambient for a very hot CPU. Just all rough and simplistic of course. There are quite a few other factors operating here (motherboard heat/motherboard heat dissipation - commonly called secondary heat paths), radiator intake temperature which can be higher than ambient especially if the radiator is mounted in-case, etc.