Storm/G7 Prototype + Piccies

[GlassMan]

Oh, and I agree - this is pretty close to sheer madness.

Maybe, but I must have one. Sometimes form and function meet at a pinnacle. This is my chance to own and use one of these rarities. Impatiently waiting for it to be ready.

Correct on all accounts. I did my best to ensure that the block allowed as free flowing of an exit path for the water as possible, as well as eliminating any and all calls for a possible third barb.

Do you plan on making a plate without the extra exit path to test for effectiveness? If it's not necessary, there seems to be a small stiffness benefit and machining (cost) savings. For those that want a 2nd outlet, it seems it would be easy to arrange , for an exorbitant extra charge of course.

edit: I read ocau post, it seems you are quite confident the extra exits work well.

[Bundles]

Originally Posted by Cathar

Oh, and I agree - this is pretty close to sheer madness.

Yeh, thought you wouldn't disagree on that, lmao!

When it comes to needing aerospace machining methods and what not, things are getting silly.

Maybe you could turn your attention to air cooling and bring that up to date, lol.

[tong]

So the question still remains... when are they getting reelased? i got a bud in australia for a few months on work, and was gonna ahve him pick one up for me and save shipping and customs and stuff, cost.

[Etacovda]

lol, if the 20$ to post it is an issue, then you're going to freak when you see the price of the actual block....

[cotdt]

oh man i want one...

[tong]

hey that's $20 i can spend on somethign else lol.

[Nugit]

Originally Posted by Cathar

Have gathered some results with the G7 vs the G5, and am busy constructing such a post. The G7 is NOT a quantum leap forwards at the higher pumping powered end, nor was I ever expecting it to be such.

A quantum leap, although it sounds pretty big is actually the smallest possible change. Sorry address such a small detail, I just had to read it over again to get what you meant.

[Cathar]

Originally Posted by Nugit

A quantum leap, although it sounds pretty big is actually the smallest possible change. Sorry address such a small detail, I just had to read it over again to get what you meant.

I think the phrase relates to the old show "Quantum Leap", where a "quantum leap" was anything but the "smallest possible change".

Also a quantum leap, even in physics terms, implies a spatial discontinuity. The particle is believed to not physically exist between one location and the next location. It doesn't just move fast from one spot to another, it was at one place before and suddenly it's at a different place. A discrete non-continuous change has occurred such that one cannot track the path from start to finish.

This is how the term "quantum leap" when describing big differences came about. It refers to a point of discontinuity between the origin point and the end point such that one cannot clearly understand how one got from A to B.

Just to nit-pick right back atcha.

[Bundles]

Originally Posted by Cathar

It refers to a point of discontinuity between the origin point and the end point such that one cannot clearly understand how one got from A to B.

So Jack Daniels makes me quantum leap sometimes then

[Etacovda]

Always read the small print, bundles (sorry, i had to) - woops, it appears i quantum leaped the 'ed' on unexpected

[pHaestus]

So my Storm G5 is now obsolete? Just like the Cascade that preceded it on my main rig. Time marches ever onward

[Bundles]

Originally Posted by Etacovda

Always read the small print, bundles (sorry, i had to) - woops, it appears i quantum leaped the 'ed' on unexpected

LOL! nice one mate.

[Cathar]

Originally Posted by pHaestus

So my Storm G5 is now obsolete?

Indeed it is.

I project that the right end of the G7's line would end at about ~7.3-7.4C on the old Procooling testbed charts, at around 1.75-1.80gpm.

I'm in the process of measuring where the left side of the line would sit at flow rates less than that point. The theory tells me that at 0.5gpm that it would project to sit at around 9.7C on the Procooling charts.

[Cathar]

I really am starting to become increasingly convinced that waterblock performance is hitting a wall.

With a ~100mm² CPU die size, and 6W of hydraulic pumping power, my waterblocks have exhibited roughly the following historical progression of effective convectional transfer efficiencies:

"Concept Block" - original WW precursor with plastic slit nozzle Prototype: ~62000 W/m²K
White Water x 1mm channels: ~67000 W/m²K
White Water x 0.8mm channels Prototype: ~71000 W/m²K
Free (un-nozzled) Jet Array Against flat base-plate (1mm jets) Prototype: ~55000 W/m²K
Mini-cupped free jet (un-nozzled) Prototype: ~65000 W/m²K
Cascade: ~72000 W/m²K
Cascade SS: ~74000 W/m²K
Cascade XXX Prototype: ~80000 W/m²K
Storm/G1 Prototype: ~50000 W/m²K
Storm/G3 Prototype: ~65000 W/m²K
Storm/G4: ~77000 W/m²K
Storm/G5: ~85000 W/m²K
Swiftech STORM (G4 Rev2): ~83000 W/m²K
Storm/G5 w/ G4 Rev2 optimisations (theoretical): ~91000 W/m²K (projected estimate)
Storm/G5 w/ G7-level optimisations (theoretical): ~94000 W/m²K (projected estimate)
Storm/G7: ~105000 W/m²K

Now in all that, the difference from the White Water to the Storm/G7 is about exactly 3.0C better for a 100W heat load on a 100mm² CPU die size. Overclocking-wise though the newer blocks do a fair bit better than what such a small temperature difference would otherwise imply.

The trek from 62000 to 105000 has been long and hard. In order to match a Storm/G7 in a copper implementation would take an h(eff) of around 120000 W/m²K, and to be honest I can't see that happening. Even if there was a block done in silver with an h(eff) of 120000 W/m²K, then performance would only improve by around 0.004 C/W (0.4C in 100W) [or by a projected ~0.25C better on the Procooling test charts].

We've seen micro-channel designs, pin grid array designs of various types, WW-variants by the dozen (the 1A-Cooling 1A-HV4 is effectively a WW-variant), and through it all we haven't seen a single design that threatens to hit anything much above 80000 W/m²K with ~6W of hydraulic pumping power on a ~100mm² CPU die.

Call it a magniloquent declaration if you will, and I'm sure plenty of you will , but I'll wait to be proven wrong and I'll believe it when I see it independently tested, verified, and head-to-head against a G7 on a testbed that's measuring actual die temperatures (NOT IHS case-temps which I believe is fundamentally flawed due to uncontrollable variances in IHS-CPU die contact).

[giorgioprimo]

Signed.....

[bobo5195]

surely the blocks especially the storm are not cooling by convection but by conduction (yes i know nitpicky). I would of thought the whole point of the storm is to increase reynolds number levels and provide a nice thick v tubulent ther.mal boundary layer for heat to transfer to. Certainly the dimensions of the storm block inducate that water is mostly movin in the thermal boundary layer region.

Linky about thermal boundary (bring ur text book)
http://scienceworld.wolfram.com/phys...daryLayer.html

[Cathar]

Convection is the transfer of heat energy involving a moving non-solid (eg. a liquid or a gas).

Conduction is the transfer of heat energy through effectively static (immobile) bodies of mass.

The thing that differentiates convection from conduction is that in convection the non-solid is able to move and redistribute the heat away from the point where it is entering the non-solid, thereby typically providing a higher rate of heat transfer than if the non-solid was not moving at all.

Most assuredly all waterblocks work by means of convection.

[Les]

Originally Posted by Cathar

I really am starting to become increasingly convinced that waterblock performance is hitting a wall.

With a ~100mm² CPU die size, and 6W of hydraulic pumping power, my waterblocks have exhibited roughly the following historical progression of effective convectional transfer efficiencies:

"Concept Block" - original WW precursor with plastic slit nozzle Prototype: ~62000 W/m²K
White Water x 1mm channels: ~67000 W/m²K
White Water x 0.8mm channels Prototype: ~71000 W/m²K
Free (un-nozzled) Jet Array Against flat base-plate (1mm jets) Prototype: ~55000 W/m²K
Mini-cupped free jet (un-nozzled) Prototype: ~65000 W/m²K
Cascade: ~72000 W/m²K
Cascade SS: ~74000 W/m²K
Cascade XXX Prototype: ~80000 W/m²K
Storm/G1 Prototype: ~50000 W/m²K
Storm/G3 Prototype: ~65000 W/m²K
Storm/G4: ~77000 W/m²K
Storm/G5: ~85000 W/m²K
Swiftech STORM (G4 Rev2): ~83000 W/m²K
Storm/G5 w/ G4 Rev2 optimisations (theoretical): ~91000 W/m²K (projected estimate)
Storm/G5 w/ G7-level optimisations (theoretical): ~94000 W/m²K (projected estimate)
Storm/G7: ~105000 W/m²K)
.

Not in disagrement with my old "Fantasy Cooling" when C/W(TIM)" is adjusted to the "more accepted 0.06c/w"(from 0.1c/w).(link and link
Updated version(terminology and C/W(TIM)=0.06) is attached.

Originally Posted by Cathar

Now in all that, the difference from the White Water to the Storm/G7 is about exactly 3.0C better for a 100W heat load on a 100mm² CPU die size. ).

Difficult to comment further without access to this data

[Cathar]

Indeed Les, those figures were all derived with a 0.06 TIM C/W assumed. Had postulated much higher h(eff) figures in the past due to using a higher TIM resistance, but went back and recalculated against a 0.06 TIM C/W value.

Originally Posted by Les

Difficult to comment further without access to this data

Yeah, I'm always going to be a little coy on divulging key specifics Les, for the usual reasons. Am pretty much convinced that the White Water and the Cascade both could've done with a thicker bp, especially so since I discovered that one of the thermal probes was faulty at about the same time as I released the Cascade. Oh well.

Without divulging exact details, the G7's bp thickness is far more in line with what works better both in theory and in practise.

[Les]

Originally Posted by Cathar

Indeed Les, those figures were all derived with a 0.06 TIM C/W assumed. Had postulated much higher h(eff) figures in the past due to using a higher TIM resistance, but went back and recalculated against a 0.06 TIM C/W value.

Yeah.
The importance of Incoherent's work in this area cannot be over estimated.(link)

[Cathar]

Originally Posted by Les

Yeah.
The importance of Incoherent's work in this area cannot be over estimated.(link)

Indeed. I followed his work intently and did my best to mimic what I could of it in my own inadequate-in-comparison way, and came to a ~0.065 C/W value but with admitted caveats of my testing's limitations, all of which pointed to the value being a little lower than that. That was close enough for me to put faith in a flat 0.060 C/W TIM value as per Incoherent's work.

[bobo5195]

I think that you might be wrong there. A thermal boundary layer (thin layer of liquid above the surface) , is where conduction is the dominant transfer medium in the fluid. Conduction does occur in fluids but it only dominates over small distances such as the exit of the jets on the storm seris of blocks. What your effectively saying is that the prandtl number is infinite, whihc is not the case.

My estimation is that the exit to the storm is all thermal boundary layer. If it is not boundary layer then it should be by design as by definition the boundary layer is where most of the special luvly heat transfer takes place, or at least the major part of the thermal gradient.

Conduction is heat transfer in a stationary medium; not nescessarly a solid.

[Les]

Probably the most handy definitive view of "fluid to solid" heat transfer is in Antoine Dechaume's article (link)

[bobo5195]

nah i'll stick with imperial college mechanical engineering 2nd year heat transfer course. It has such highly inteligent comments as "mike is a loser" and "I like meat" but it makes the overclockers piece look exceptionally brief and light in places, plus ive got the science museum uber libary for help and my housemate famously declared "convective heat transfer is sexy" so he could answer some his questions.

I'm having a read now and i think i disagree with some of the stuff in there as its obviously dumbed down. One of my lectures loves that book on fluid motion though.

[Cathar]

Originally Posted by bobo5195

I think that you might be wrong there. A thermal boundary layer (thin layer of liquid above the surface) , is where conduction is the dominant transfer medium in the fluid. Conduction does occur in fluids but it only dominates over small distances such as the exit of the jets on the storm seris of blocks. What your effectively saying is that the prandtl number is infinite, whihc is not the case.

No, now you're being nit-picky in the extreme, and taking the discussion into a topic of philosophical classification.

Of course I'm aware of the static layer/film of water molecules that are in contact with the metal's surface and that the primary means of heat transfer from the metal into those molecules is conduction. Never meant to imply that the Prandtl number is infinite.

The issue is how thick is that layer of static conduction? The thickness of the pure-conduction layer is directly affected by the motion of the water. Since the conduction layer is made of a liquid and is affected by means of liquid motion, then it merely becomes part of the overall definition of convection.

We can get philosphical about it if you want. Convection is a thin layer of conduction through a liquid at a thermal boundary followed by heat transfer through means of motion. Given that the thermal boundary layer is affected by fluid motion it is therefore no longer an invariable static body of mass and thus this action collectively falls under the umbrella definition of the "convection process". If you want to argue that the convection process involves some amount of conduction, fair enough, but the whole process is not purely conduction because of the variable nature of the conduction layer.

My estimation is that the exit to the storm is all thermal boundary layer. If it is not boundary layer then it should be by design as by definition the boundary layer is where most of the special luvly heat transfer takes place, or at least the major part of the thermal gradient.

Perhaps we're just quibbling over clarification of terms. You're basically saying that as close to all molecules of water entering as possible are actively involved in the thermal transfer process (what I call convection, what you're calling conduction but that which I believe still falls in the term convection), and yes, I agree with you. That is the singular driving thought paradigm behind all my block designs => "How can I make ALL of the water molecules become directly involved in the transfer of heat from the metal into the liquid".

Of course, that's exactly what micro-channels try to do by literally forcing/constraining all water molecules to act within the thermal boundary layer, but all true micro-channel implementations have tremendous issues with pressure-drop management as well as cloggage (<0.1mm channels clog very easily despite best efforts), and those are the additional knurs that can't be dismissed for enthusiast watercooling use.

Conduction is heat transfer in a stationary medium; not nescessarly a solid.

Indeed. Look up though. At no point did I say that conduction had to be through a solid. I said conduction occurs through a static (immobile) body of mass. This means that "mass" that can be a liquid, gas or solid, which is the same thing you said just above, just worded differently.

I appreciate that you're just trying to flex the theoretical muscle and classify what's going on. I'm just using commonly used terms of thermal transfer performance as a handy means of performance measurement/classification. If you want to argue that it's pure conduction then that's somewhat outside the general broad level discussion that we're trying to achieve here.