MIKROS BLOCK & white paper

[UNDERBYTE]

Big claims as to performance (did a search and have seen no other posts on it)

Just a FYI

http://www.mikros.net/PDFiles/NFHS%20white%20paper.pdf

[bigben2k]

This is the first serious presentation I've seen that's anywhere near what we're doing... nice!

Note the scope under which this was developped (last page).

[-J-]

now that looks quite intresting.

[Cathar]

Originally Posted by UNDERBYTE

Big claims as to performance (did a search and have seen no other posts on it)

Performance claims appear to be an internal thermal resistance of around 0.10C/W for a 10x10mm heat source. Note, that does not include the TIM layer, i.e. what BillA coined as T/W. With the TIM layer (ala BillA style testing) and at 0.6LPM, they are basically saying up around 0.20C/W for a 10x10mm heat source after perusal of their testbed simulation results. This basically puts their block at about on par with the better blocks on the market today.

What is impressive here is their ability to get such good performance at low flow rates, but their block appears to be extremely restrictive. Pushing >2LPM through the block isn't going to be easy and at that point it appears that the block is starting to give diminishing returns.

Overall the results are about where I would expect them to be for a good pure micro-channel implementation.

For me, it's the age-old trade-off. Can get excellent efficiency with micro-channels, but needs a lot of pumping effort. Or can gets substantially less efficiency, but through boosting flow rates can end up about the same anyway. Really it comes down to what flow-rate range (read: pump) that you intend to optimise for. When the heat load really starts to fly though, higher flow rates (>2LPM) tends to be more attractive IMO.

[UNDERBYTE]

It looks like they did incorporate TIM in the internal resistivity, and not a good paste at that.

A high conductivity thermal paste (4.5 W/m°C) is used to accommodate slight height variations in the components and to provide a stress-free interface between the heat sink and the chip. A chip thickness of 300 m and an interface thickness of 40 m yields a totalinternal resistivity of 0.10 °C/(W/cm2).

What I find interesting is tailoring C/W to the "hot spot"

Another advantage of the NFHS technology is that its resistivity can be tailored over the surface of the heat sink to eliminate “hot spots.” This ability is an intrinsic characteristic of normal flow heat transfer (see Section 4.) As illustrated in Figure 3, tailoring the thermal resistance to the heat dissipation provides the most effective means of eliminating potential hot spots in applications where the heat generation is highly non-uniform. The thermal resistance of the NFHS can be controlled from a low value of about 0.05°C/(W/cm2) to any higher value with a spatial resolution of about 1 mm.

[Cathar]

Originally Posted by UNDERBYTE

It looks like they did incorporate TIM in the internal resistivity, and not a good paste at that.

Uh, by very definition, "internal resistivity" does not include the TIM.

Read section A of Page 5.

In order to dissipate 160W/cm² with a temperature difference of: 55 - 22 = 33°C, the total resistivity must be less than 0.20 °C/(W/cm²).

[UNDERBYTE]

Originally Posted by Cathar

Uh, by very definition, "internal resistivity" does not include the TIM.

Read section A of Page 5.

Looks to me they defined thier own Internal resistivity for the paper. Maybe I am missreading it.


Read section B

[Cathar]

Originally Posted by UNDERBYTE

Looks to me they defined thier own Internal resistivity for the paper. Maybe I am missreading it.


Read section B

Follow the math, not the words (which in that paper are somewhat confusingly worded).

160W/cm² for a 32C rise (over a 1cm² area)

32/160 = 0.2 °C/(W/cm²) (Total Resistivity)

Note their extensive switching between the Total Resistivity and Internal Resistivity terms, often quoting the latter.

The "Internal Resistivity" is the resistivity of the waterblock itself, not accounting for the thermal disjunction between the heat-source and the waterblock. It is, in essence, what BillA once coined as T/W.

The "Internal Resistivity" of a White Water is conjectured to be around 0.08-0.09 °C/(W/cm²) at 10LPM.

[pelle76]

Could someone explain to me how the internals of this block looks like? I did not get that from the paper.

[|kbn|]

Thats probably patented etc.. so doubt they will show anyone...
Are they refering to the "hotspot" as the core, in the area of the whole package - which would be nothing new, or are they talking about hotspot within the core? I found that unclear..