Quote:
|
Originally Posted by nigelyuen
with the german micro channel, it has more surface area on top of the cpu die, is that why they are optimize for the low water flow rate?
|
Let me first state that the following is just my personal understanding of the merits of various design approaches, and can always be debated. It is my opinion, but it is also my opinion that drives the work that I do. I'll also keep the explanations fairly broad and simple.
Waterblock designs can be broken down into 3 main categories:
1) The micro-channel/micro-pin approach allows for greater thermal transfer efficiency with respect to the heat capacity of the coolant flowing through the block. If one subtracts the inherent thermal resistance of the waterblock material (copper or whatever) and the thermal interface (thermal goop layer), the performance of a micro-channel design can very closely track the thermal capacitance of the coolant for the given flow rate. This efficiency does tend to drop away fairly rapidly though at higher flow rates/pumping pressures, so while the micro-channel designs have quite amazing efficiency at very, very low flow rates (<0.5 litres per minute), this doesn't tend to translate that well to amazing performance at higher pumping pressures. In fact, the inherent flow resistance of true micro-channels is such that it is extremely hard to get the design that is very efficient at super low flow rates to achieve what can even be called "low flow rates" even with quite powerful pumps. In essence, I believe that the micro-channel design is somewhat self-defeating when matched with stronger pumps - but that is just my opinion. Such designs don't need high flow-rates. My opinion, "micro-channels" in the sense of the cooling effect I describe above, pretty much applies to channels/structures of 0.2mm or less in size. Channel widths between 0.5-1.5mm are really "mini-channels", and form a bit of a performance hybrid between micro-channels, and category 3) below. Channel widths of 2.0mm and above pretty much fall wholly into category 3) below.
2) Impingement based cooling is less efficient than micro-channels in a purist sense. If we subtract the thermal resistance of the thermal goop layer and the block material from the thermal resistance of the block, and compare that to the thermal capacity of the coolant, impingement based cooling is horribly inefficient next to micro-channel cooling (this is somewhat contradicted by true micro-jet impingement cooling, but such approaches typically use evaporation to help their performance, which is not really the focus here). However, for all of impingement cooling's inefficiencies, it can be more
effective at dealing with heat, provided that sufficient pumping pressure is applied. The actual flow rates that people see really are a side-effect of the per-jet pressure drop required to accelerate the liquid to effective velocities, and the number of jets. At least in my mind, jet impingement cooling responds better to stronger/higher pressure pumps than micro-channel cooling does. Jet impingement does rely more on higher pumping pressure more it relies on high flow rates. The velocity of the jets is usually more important than the sheer volume of the water that's flowing. Due to the restrictive nature of jet impingement cooling that primarily uses the pumping pressure to create high water velocity jets rather than allout flow rates like 3) below, it is quite possibile to use smaller ID tubing with a medium-strong pump to achieve excellent cooling.
3) Then we get to pure turbulence based cooling. Such designs can be typified by the older-style USA Maze designs, and indeed some of the older German designs, however there are some more modern designs that use it to good effect (the DangerDen RBX/TDX being a good example). Such designs pretty much just rely on massive flow rates and pumping pressures to get the water flow into a highly turbulent state to improve cooling efficiency. They may also include various lump, bumps, and "turbulence enhancers" to improve cooling efficiency. These designs, more than anything else, spawned the need for full 1/2" ID systems. Don't get me wrong on this though, this approach can be highly effective if done properly, but it doesn't fall into the micro-channel/impingement classes above.
All blocks can be broadly categorised based on their varying use of the above three cooling effect approaches, and many of the more successful blocks across a wide range of flow rates are really hybrids of two or three of the above design approaches.
For example I would classify various commercial blocks in the following ways:
AtoTech MC-1: Purely Type 1
1A-HV2 : Almost purely Type 1
HydroCool Hydro-Stream: Almost purely Type 1
Innovatek Rev 3: Somewhat tricky one to classify - largely a Type 1/3 hybrid channelled block
NeXXoS HP : Mostly Type 1 with some Type 2/3 aspects
White Water : Mini-channelled Type 1/3 hybrid, with Type 2 effects
Storm: Wholly Type 2
Swiftech MCW462A: Wholly Type 2
PolarFlo blocks: Largely Type 2, Some Type 3
Swiftech MCW6000 : Hybrid of 2 and 3
Cascade : Roughly half Type 2 and half Type 3
TDX/RBX : Largely Type 3 with varying degrees of Type 2 depending on nozzle plate used
Jaydee's Lumpy Channel : Almost purely Type 3, with some Type 1 characteristics
Swiftech MCW5000: Wholly Type 3
The other important thing to consider is base-plate thickness. The thicker the base-plate, the greater the effective convectional surface area that comes into play when flow rates and cooling efficiency drops away. However the thicker the base-plate, the greater the copper's thermal resistance, which can hinder performance improvements as cooling efficiency increases with higher flow rates. Some waterblock engineers have made deliberate trade-offs to match their design's strengths with a certain base-plate thickness to target certain flow rates/pump usage.
[Edit: cursory fix of spelling and major grammatical errors, added some more information]