S-TDX water block or What?

[HAL-9000]

I didn't read all of this previous thread, started only a page back. I get the picture. As far as the OCAU thread, from my experience there you do have to register to view the forums. I erased my cookies etc. and it just wants me to register. I will someday...not in the mood to fill out forms right now though.

From the description of your block idea it sounds like a sun-of-a-b to make. Micro-miniature holed channels on the scale your talking (syringe needle?) sounds difficult to say the least. Perhaps there could be a more unorthodox way of doing it, but from what my mind has led me to that way, you would have to be able to melt the thing while performing the process.

However, I have come across a couple of pages from folks out there who have made some home made lasers, with extremely tight beams. A pulsed power Nd-YAG laser with a very tight beam (.25mm) could slowly blow holes akin to your channels in the block, quite precisely. You would need to set up a precise mechanical guide (like a tiny jig) and such to do it...I was thinking. Nd-YAG is also efficient if you get/build the LED diode pumped versions, and your app wouldn't be particularly high power. And Nd lasers beam in the right wavelength for that kind of work. It would take awhile per block, and it would be unorthodox obviously, but you seem to be that kind of a guy. Just a thought.

Anyhoo, whats the dark history you allude to with Silverprop?

[Cathar]

Originally Posted by HAL-9000

I didn't read all of this previous thread, started only a page back. I get the picture.

Actually I think you need to read the last few pages of either thread.

Re: Silverprop - history is history - lessons were learned - that's all that needs be said.

[sevisehda]

If the list was shortened to only those availlable right now what would be the best*.

* In a high flow 1/2in loop, driven by a Mag3 no other blocks.

[pHaestus]

http://www.overclockers.com/articles1027/

I think I'd take the copper one... Same unrounded C/W too means very very slight difference. Looks like no changes in design parameters (to take advantage of silver) to me.

[pauldenton]

Originally Posted by pHaestus

http://www.overclockers.com/articles1027/

I think I'd take the copper one... Same unrounded C/W too means very very slight difference. Looks like no changes in design parameters (to take advantage of silver) to me.

yes - what seems odd though is that he gets over 20% more restriction in the silver version to the copper one????? :shrug:

[pHaestus]

I can only hypothesize Pauldenton... I saw that flow rates were lower immediately after plumbing the TDX w/ nozzles than they were the next evening when I'd test. When I disassembled the block, I noticed that the neoprene on the insert had compressed (as expected) and was hard. At some point I believe the nozzles are letting some water past w/o going through channels. My bet is JoeC tested Cu tdx flow resistance after nozzle insert had "sprung a leak" as it were while on the Ag TDX this hadn't happened at the time of testing.

Only Joe knows for sure though. I DO know that his options are: something funky (like my suggestion or something lodged in the block) w/ the wb or something funky with his test setup. Blocks are otherwise identical.

[AngryAlpaca]

Originally Posted by Bill Adams

did you not know ?
silver is more 'clingy'

Yeah, they did not change a single thing in the silver version. However, these things will still sell like mad. DD did not link to the overclockers review of the silver one... Big surprise there.

[wijdeveld]

I've done some calculations for a phase-change setup (Vapochill PE) with different evaporator materials. There should have been more performance difference based on the heat conductivity difference between silver and copper:

COPPER:
density = 8960. kg/m3 at 293 K
warmth = 387. J/(kg.K)
conduct = 390. W/(m.K)

SILVER:
density = 10500. kg/m3 at 293 K
warmth = 240 J/(kg.K)
conduct = 429 W/(m.K)

I think the lack of a serious performance increase just means we're still facing another bottleneck at the moment; the heat dissipation from the solid surface to the water phase (even with such a good designs as the blocks from Cathar).

[Cathar]

Originally Posted by wijdeveld

I've done some calculations for a phase-change setup (Vapochill PE) with different evaporator materials.<snip>

Density and "warmth" are irrelevant in a system at equilibrium.

The thermal conductivity is all that is important when it comes to a base-plate material.

An evaporator head is a significantly different scenario to a waterblock. An evaporator requires a fairly long "residence" time (in comparison to water) for the coolant to evaporate to give its full effect. Depending on the design the actual area over which the coolant evaporates will be fairly large. This then necessitates a fairly thick base-plate thickness to minimise the spreading resistance for the heat of the CPU to the full area of cooling effect that is provided by the evaporating coolant. Due to the thickness of the bp material, the difference between copper and silver can be fairly significant.

In a waterblock where there is no evaporation and the effect of the cooling is not time/location dependent as per the flow rate of coolant, thus the point of cooling may be much more concentrated. This enables the very thin base-plates that Cascade SS and the S-TDX are using, being well under 1mm on average (excluding the wall sections), as there is less of a need to minimise the spreading resistance of the bp material.

Calculate the spreading resistance difference between silver and copper over that sort of thickness, and apply that as a ratio of the total thermal resistance of the entire "TIM, bp, convectional" heat-flux pathway, and the net C/W difference works out to around 0.003 or so for the die-size that Phaestus is using. Given his predicted heat-load, that's about a 0.2C difference, or something which can easily be "hidden" through even minor variations in lapping quality.

[wijdeveld]

I agree that thinner is better, but the heat capacity of the slid material influenced the steepness of the delta T/ delta X slope to some degree (I'll be back on this subject later). Just an quick plot on the influence of die thickness:

[Cathar]

Originally Posted by wijdeveld

I agree that thinner is better, but the heat capacity of the slid material influenced the steepness of the delta T/ delta X slope to some degree (I'll be back on this subject later). Just an quick plot on the influence of die thickness:

If I'm understanding correctly what you're showing there, what you're seeing there is not the influence of the heat capacity of the base-plate, but the influence of the spreading resistance due to the thickness of the base-plate.

[AngryAlpaca]

The thermal conductivity is all that is important when it comes to a base-plate material.

What about machinability and price? If you can't form the thing into what you want, or if it will cost thousands, what good is it?

[Cathar]

Originally Posted by AngryAlpaca

What about machinability and price? If you can't form the thing into what you want, or if it will cost thousands, what good is it?

AA, the statement was referring to the differences in various bp properties. wijdeveld quoted 3 different properties, "density", "warmth", and "conductance", of which only "conductance" is important when considering performance, which is the scope within which the statement should be viewed.

[wijdeveld]

Originally Posted by Cathar

If I'm understanding correctly what you're showing there, what you're seeing there is not the influence of the heat capacity of the base-plate, but the influence of the spreading resistance due to the thickness of the base-plate.

Yep thickness it was.
I'll get back on the heat capacity now. For that I'll use some more geochemistry oriented examples (since that's my background) and compare heat to a conservative compound (no degradation) that's produced locally (at the die) and has to be transported through a medium (the solid phase = heat spreader + TIM + cooling block) to a cooling medium (water/vapor). We now can translate the heat capacity/heat conductivity of the solid phase to a difference in diffusion speed (also called retardation).

You’re right that under the assumption of steady state the amount of heat removed within a time unit has to be the same for all cooling solutions. But if the solid phase is more restrictive with regard to diffusion, there will be a certain build up of the compound within the solid phase.

Mathematically this looks like:

diffusief transport

fluxi,j = Di,j . Oppi,j . (Ci - Cj) / lengtei,j (2)

with:
fluxi,j mass transport between segment i and segment j (M/t)
Di,j diffusion coefficient between segment i en segment j (O/t)
Ci concentration of compound in segment i (M/V)
Cj concentration of compound in segment j (M/V)
Oppi,j surface area between segment i en j (O)
Lengtei,j distance between center of segment i en j (L)

The diffusion coefficient is normally given in (cm2/s) and is influences by the porosity (read conductivity of the solid phase in case of heat) of the sediment. An increase in the porosity will lead to a higher diffusion speed and therefore a higher flux with the same delta C (or delta T for heat) versus delta X.

So, assuming a linear relation between the heat conductivity and the diffusive flux, silver should have a 10% higher flux. While not completely linear, this should to a decrease in the delta T between the die and cooling medium of roughly 10% to have the same flux. A normal delta T for a water block with a 100 watt load would be around +15 oC (water chiller). Assuming that heat conductivity is the primary limiting factor in transporting heat, using silver would have lowered the delta T with roughly 1.5 oC.

[Cathar]

Originally Posted by wijdeveld

While not completely linear, this should to a decrease in the delta T between the die and cooling medium of roughly 10% to have the same flux. A normal delta T for a water block with a 100 watt load would be around +15 oC (water chiller). .

Highlighted the important statement in all that. The cooling medium effect (convection) is constant (assuming identical design). The TIM interface is constant. What's changes is the deltaT of the base-plate face to the coolant area. That particular delta-T will typically constitute only a very small proportion of the overall delta-T between die and coolant. In thin bp waterblocks, it's estimated to be around 15% of the total delta-T for a waterblock.

So 15% of the 15C is 2.25C, and a 10% gain on that is ~0.2C (simplistically speaking). It again comes back to the spreading resistance of the bp material to the convection area. Yes, the size of the heat flux area engaging the coolant is slightly increased, but this is still only a minor effect.

[BillA]

I do not believe a 'diffusion coefficient' has any relevance to solids and a liquid medium, and would be the same for cu and ag in any case per your definition (varying porosity)

[wijdeveld]

O.K. I've assumed a thickness of 2.5 mm for the reference calculation:

[BillA]

not understood
you are showing the transient effects of heat capacity ?
and at equilibrium ?

[wijdeveld]

Originally Posted by unregistered

I do not believe a 'diffusion coefficient' has any relevance to solids and a liquid medium, and would be the same for cu and ag in any case per your definition (varying porosity)

Yes it does; for a chemist heat can be treated like a chemical compound with regard to most of the relevant properties. Heat doesn't disappear without a reason (it's mass conservative in this regard), the transport behavior is described by diffusion, dispersion and advection and the absence of equilibrium (by lack of transport) will lead to a heat build up with a higher delta T until a new (flux based) equilibrium has been reached.

[wijdeveld]

Originally Posted by unregistered

not understood
you are showing the transient effects of heat capacity ?
and at equilibrium ?

Nope, first calculation was only considering heat conductivity (as discussed here). The second also took the heat capacity into account (as a sorbing 'buffer' phase).Ah well, chemistry isn’t perfect but it can get you a long way in understanding water/phase change cooling

[BillA]

I see we're headed toward inter-molecular and intra-atomic here

granting your definitions above, how do cu and ag differ wrt the heat transfer at their surfaces ?
i.e. what is the material property defining this characteristic ?

back to the graph
ok, you are showing a transient effect
what is its relevance ? (beyond the short term)

[wijdeveld]

Originally Posted by unregistered

granting your definitions above, how do cu and ag differ wrt the heat transfer at their surfaces ?
i.e. what is the material property defining this characteristic ?

back to the graph
ok, you are showing a transient effect
what is its relevance ? (beyond the short term)

All calculations are with perfect heat transport from solid to liquid phase (no double layer constraints), that's the main bottleneck right now I think (as stated a few posts back).

The graph shows the steady state heat profile from the die to the cooling medium (assuming perfect heat dissipation between the transition layers). I've developed this toy model for evaluating the effect of the compostion of TIM on the cooling performance. Seems that for TIM the heat dissipation is also of more importance then the exact compostion with regard to heat conductivity, since the TIM (AS3 in this case) only contributes for 1% to the total delta T.

[Groth]

Originally Posted by wijdeveld

diffusief transport

fluxi,j = Di,j . Oppi,j . (Ci - Cj) / lengtei,j

Okay, you rewrote the standard heat conduction equation (P=k A dT/l) into terms more familiar to you. What are you trying to use to insert specific heat ("warmth" you called it) in your calculations?

[wijdeveld]

Originally Posted by Groth

Okay, you rewrote the standard heat conduction equation (P=k A dT/l) into terms more familiar to you. What are you trying to use to insert specific heat ("warmth" you called it) in your calculations?

"HEAT: Heat is produced analogue to Organic Matter breakdown, which results in a heat production term. Because in normal OM simulations the amount of OM is limited (and for now the heat production must stay constant) a very large amount of potential heat is introduced with a very slow degradation (= production) rate"

What you get is an amount of heat as a compound. By using a Kelvin scale and by calibrating the model on known cooling medium & die temperatures you can make a translation between heat as a concentration in solution and heat as the actual temperature.

[BillA]

I think the analogy is rather strained

"heat as a concentration in solution" is a difficult concept for me in thinking about solid cu and ag
- now you would propose that the "heat as a concentration in solution" of cu is different than in ag ?

please explain
(as I return to inter-molecular and intra-atomic heat transfer)