WW kicks ***: BillA test results.

I wonder if he was heading in the direction of possibly tapering the tops of the slot walls beneath the inlet so that the coolant slides down them with less deflection as would be encountered on a blunt flat surface?

Hoot

Hoot said:

I wonder if he was heading in the direction of possibly tapering the tops of the slot walls beneath the inlet so that the coolant slides down them with less deflection as would be encountered on a blunt flat surface?

Hoot

Click to expand...

That was my first thought, but I'd seen that asked of Cathar before. This statement of Cathar's is what got me thinking of shorter fins below the nozzle:

Cathar-"Then the impingement takes place on the base. I suspect that the impingement effect on the walls towards the bottom of the channels is acting at close to direct impingement convectional values due to the pressures and turbulence in this region. The water flow hits the base, and then wants to go in all directions but is constrained by the channel walls. This, I believe, is giving a secondary impingement effect acting on the walls close to the base, perhaps affecting the lower 1.5mm portion of the walls (which is the hottest section of the walls - which is exactly what we want)."

It seems to me like if the fins in that area were 1.5-2.5mm high and still flat on top there might be an improvement. To me it seemed like a "blinding flash of the obvious" but I take into account my lack of knowledge in this discussion. I know Cathar did a lot of testing with fin heigt and channel widths and such, but everything I've read when he talked about it left the impression that he was talking across the board, not in a specific area of the fins. Also I've never been on ocau board where the most discussion took place, so he might have already explored this. It wouldn't be hard to test out, but ya might ruin a perfectly good block in the process, which is more than I'm willing to risk. Hopefully he'll comment on this and put my mind to rest.

peace.
unloaded

What I meant was tapering the wall tips only beneath the inlet area. The rest of the wall top area could remain flat. See the accompanying picture. Should be self-explanitory.



Hoot

A good suggestion, Hoot.

I recently went through the exercise of comparing ASTM nozzle specs vs orifice plates. In short, a nice curved entry point will result in an 80% pressure drop (inlet to outlet), where the orifice plate can vary between 60 and 80% (depends on a number of factors).

It might make a difference, but it might not be appreciable

Here's a nice set of pictures by roscal who ran the prototype block through his CFD program. I believe it makes quite a few things clearer for the past few posts.

Further to those CFD pictures, the production block is different.

The fins are lower in height. Basically resulting in boosted velocity through the block. Dropping the fins down much below 4mm in height had detrimental effects as there is now insufficient surface area for the water to act on and the thermal gradients through the metal "hits" the tops of the fins without being adequately cooled.

The impingement on the tops of the fins isn't doing much of anything.

Dropping the fin heights below the prototype block I believe reduces the effect of the water back-eddy swirls that can be seen. I believe the water in general flows more directly from impingement region to the outlets.

Chamfering the tops of the fins like is suggested by Hoot was considered a long time ago. As for boosting cooling effect, we can see that it will have little effect. It will reduce the pressure drop slightly, but we're talking maybe a 10% lower pressure-drop here. This may help the block by a tiny amount (~0.1C) but at the extra expense of machine time due to a tool change now being required. Definitely it was worth considering.

All up, the block as it stands in its basic dimensions may be tweaked in numerous ways that would probably help it pick up maybe 0.5C on a hot CPU. The drawback here is in the added machining cost. Most of the things that could be done all require a tool change, which does all add up.

I was wondering this a while ago...

What if the fins had one more channel cut perpendicular (90*) to the main channels? Sure, it may weaken the structure by a couple tenths of a %, but the rest of the metal can take the slack. Even if there is a reduction of surface area, since that impingement stuff is said to happen at the bottom and the 1/3 of the 'relocated' surface area is at the base...which is the hottest...

And why can't I add my 581 byte attachment? This 'post quota' for enabling functions is starting to irritate me a bit...


Edit: avoids attachments...



BTW - nice swirl/flow pictures...with the colors, they almost verge on artwork.

HiProfile said:

I was wondering this a while ago...

What if the fins had one more channel cut perpendicular (90*) to the main channels? Sure, it may weaken the structure by a couple tenths of a %, but the rest of the metal can take the slack. Even if there is a reduction of surface area, since that impingement stuff is said to happen at the bottom and the 1/3 of the 'relocated' surface area is at the base...which is the hottest...

And why can't I add my 581 byte attachment? This 'post quota' for enabling functions is starting to irritate me a bit...


Edit: avoids attachments...

BTW - nice swirl/flow pictures...with the colors, they almost verge on artwork.

Click to expand...

That might be possible. I think, this is my opinion of course, that for jet impingement to really be effective is having a complete flat base in the center.

SysCrusher said:

That might be possible. I think, this is my opinion of course, that for jet impingement to really be effective is having a complete flat base in the center.

Click to expand...

The thing you have to consider is this. Cathar said that the impingement on the tops of the fins isn't doing hardly anything (about 3 posts above). Now I can see why you want to cut a perpendicular channel, and I don't have any equations to throw out, but I know that impingement depends heavily on the velocity of the water.

By leaving the top of the fins (7 fins and 8 channels) flat (and not cutting them), the water is forced to go through the 8 channels, and hit the base plate. This is only 8/15ths the area the water would have (to hit the base plate) of the area that would be exposed if you cut the perpendicular channel. This would probably almost cut in half the water velocity hitting the base, which would not be a good thing for impingement.

... Ooops, said something that has already been stated

... Delete ...

Cathar said:

Here's a nice set of pictures by roscal who ran the prototype block through his CFD program.

Click to expand...

For those interested a link to the thread where roscal presented the simulations:
http://forums.overclockers.com.au/showthread.php?s=&threadid=133088

Maybe worth noting that roscal's estimate of Temp are similar to that obtained by my "Side-to- Side configuration" Model at the same heatload and flowrate.This is considerably higher than predicted by my "Central Slot Inlet model":


EDIT: Updated/corrected graph - Corrected misinterpretation of Kryotherm's "Base Irregularity Coeff (BI)"( Changed to BI=1.0 from BI=1.2)

Albigger said:

The thing you have to consider is this. Cathar said that the impingement on the tops of the fins isn't doing hardly anything (about 3 posts above). Now I can see why you want to cut a perpendicular channel, and I don't have any equations to throw out, but I know that impingement depends heavily on the velocity of the water.

By leaving the top of the fins (7 fins and 8 channels) flat (and not cutting them), the water is forced to go through the 8 channels, and hit the base plate. This is only 8/15ths the area the water would have (to hit the base plate) of the area that would be exposed if you cut the perpendicular channel. This would probably almost cut in half the water velocity hitting the base, which would not be a good thing for impingement.

Click to expand...

Thats why I said possible(I'v been proven wrong before) but in my mind I don't see cutting a channel through would help any. I'll give the reason why a flat base for jet impingement to really be effective though I'm no scientist or engineer. With the flat top of the fins, the water will hit the top and tend to spread outward, right into the water trying to flow down into the channels. The top of the fins would actually see the best benefit. Why it's not doing much in this case. Even with sharper edges to the fins in the middle I think we would still see the same but not as pronounced. The flow would tend to hit the sides of the fins with some actually hitting the base of the channels. Impingement in my opinion needs a flat base in the center for the impingement to really be effect which will allow the velocity to really hit the base giving it a chance to spread out and thinning that boundary layer from the center out. Even better is to have a small pin either round, square or what ever the shape of the jet is but just slightly smaller that fits in the stagnation area.

I'v been trying to work out the idea through trial and error since I seen Cathar's block. I know it's unscientific and all I got to go on is luck, my visual sense, a dremmil and free Alu. I made two blocks one with channels and another with nothing but a flat base though slightly tapered. All aluminum, which isn't the best. Guess which one where jet impingement actually shows a greater change? The one with the complete flat base.

I'll admit that I haven't the 'formulas' to validate my idea, it's just a suggestion. Its my guess that this could be done with a dremel tool with the 'head on a cable' (extended, flexible head) on one of the beta blocks. All this stuff being talked about now IMO imacts overall effeciency like an exta M80 adds to a TNT blast. Its not so much splitting hairs, its more like splitting atoms. Not to say its all fruitless...most extreme OC'ing IS splitting hairs.

I'm sure someone with a degree, keener knowledge of thermo/hydro-dynamics and a good simulator app could prove this better or worse, but I'm saying the slot could possible add to cooling, nothing more.

As for reducing velocity to the point of being less efficient, the base is very thin, and maybe increased surface area (in a spot that acts like the portion of a HSF directly under the hub - less flow) might help. Fluids aren't know for making perfect right angles, so there should be a spot under that jet where the water lingers a bit more than the rest.

The water will still be traveling fast, albiet not AS fast, when it hits the area where the slot is, but the base slot is MUCH narrower than the slot in the middle plate. Also, the diagrams from roscal that Cathar posted show that the water is still traveling towards the base beyond the slot in the middle plate. The whole idea of the WW is challenging wether velocity/flowrate or surface area is better (basicly reinventing the WB). If there is less velocity [and resulting less pressure in the slot...and don't remind me of presure & liquids] in the slot, it may keep water from dispersing as much and remain in more of a vertical decent...of course I'm the king of educated guesses... The blue spot in roscal's diagram (top most) directly above what should be the base's hot spot is one reason I this might be worth looking into - shouldn't the hottest metal be surounded by the hotest water?

As for extra expense...it's less material to remove for this slot than there is for 1/3 of one channel. When everything is said and done, though, I'm not as much of an expert next to Cathar (or many of the rest of you guys), as well as it's HIS wb design. And it obviously works great already.

BTW - for me, planning & tweeking is the best part of a project...noticed?

HiProfile said:

I was wondering this a while ago...

What if the fins had one more channel cut perpendicular (90*) to the main channels? Sure, it may weaken the structure by a couple tenths of a %, but the rest of the metal can take the slack. Even if there is a reduction of surface area, since that impingement stuff is said to happen at the bottom and the 1/3 of the 'relocated' surface area is at the base...which is the hottest...

And why can't I add my 581 byte attachment? This 'post quota' for enabling functions is starting to irritate me a bit...


Edit: avoids attachments...

BTW - nice swirl/flow pictures...with the colors, they almost verge on artwork.

Click to expand...

I can tell you why you don't want to do that, because I'm faced with the very same issue with my block: the fins over the core are absolutely critical. If you look at the second graph that Cathar posted, you'll see the heat distribution. Without the fin in that area, the baseplate would get hotter, resulting in a higher CPU temp. Like Cathar said: fins can't be lower than 4 mm.

Optimizing the inlet is the only way to go here.

A little question though:
if I had an Eheim 1250 and the CPU block, would I still be able to add a GPU block to the loop without getting to low flow for the cpu block to work its best?

HiProfile said:

-snip-

And why can't I add my 581 byte attachment? This 'post quota' for enabling functions is starting to irritate me a bit...

Edit: avoids attachments...

-snip-

Click to expand...

I have over 3200 posts and I can not post attachments without hosting them on an external site either. Post count has nothing to do with it. That's just the way it is. I do not care for the current provisions for images, but I accept the fact that the administrator felt it was necessary.

Hoot

Les56 said:

For those interested a link to the thread where roscal presented the simulations:
http://forums.overclockers.com.au/showthread.php?s=&threadid=133088

Maybe worth noting that roscal's estimate of Temp are similar to that obtained by my "Side-to- Side configuration" Model at the same heatload and flowrate.This is considerably higher than predicted by my "Central Slot Inlet model":

Click to expand...

But both are lower than Bill's, no?

Ben

T/W = "C/W", minus the TIM joint C/W
we do not have a hard value for the TIM joint's C/W; best est. ~ 0.1 (could be ±0.01)

all of this is why I always put quotes around the term "C/W" when I use it
- to try to remind folks that "C/W" is actually the sum of 2 very different thermal resistances; the TIM joint plus the wb

be cool

and also, his are simulations, theres another thing, It woudl be nice to have something like that to simulate some stuff for me You got a PM les