Are there any Socket F/940 evaporator blocks available?

[Shingoshi]

I want to use a phase-change system on my AMD Socket-F server motherboard. The board has four sockets. So I'm looking for something that will work for me. It would be nice if they were low-profile also. But I would be happy with anything that works (and isn't absolutely ugly).

Otherwise, I want to know if it's possible to use my existing Koolance CPU-330 cooling blocks with nozzles and tubing suited for phase-change. Essentially, I would need to make each cooling block my evaporators. So I'm guessing that I would need a small orifice that inserts into the block to allow for expansion in the block itself. If anyone has any experience with this, please let me know.

Shingoshi

[[XC] gomeler]

Just to entertain this idea, is there any particular reason you want to run phase-change on a four socket motherboard? There has been very little if any success overclocking the dual-socket AMD motherboards so I can't imagine a quad-socket board would have any headroom at all.

I don't even know what mounting holes Socket F uses but you can use any evaporator so long as it has a mount for that board. However in an ideal world where you have Shanghai processors with unlocked multipliers and vcore adjustment up to 1.6v you are still only going to hit ~4.5GHz. This isn't an ideal world though so I doubt you have unlocked multipliers and the desire to wire in 4 vCORE mods on that board, so I'd say you are better off with a nice watercooling setup.

[Duniek]

mine evaps fill fit

but waht do you mean about low profile evaporators?

[Shingoshi]

Just to entertain this idea, is there any particular reason you want to run phase-change on a four socket motherboard? There has been very little if any success overclocking the dual-socket AMD motherboards so I can't imagine a quad-socket board would have any headroom at all.

I don't even know what mounting holes Socket F uses but you can use any evaporator so long as it has a mount for that board. However in an ideal world where you have Shanghai processors with unlocked multipliers and vcore adjustment up to 1.6v you are still only going to hit ~4.5GHz. This isn't an ideal world though so I doubt you have unlocked multipliers and the desire to wire in 4 vCORE mods on that board, so I'd say you are better off with a nice watercooling setup.

And I was afraid what you said is true. I was simply looking for another option to consider in my project, besides relying on liquid-cooling. I guess I would have to see if there is a program for Linux that would allow me to control the parameters of my motherboard which may not or don't exist in the BIOS. But like you said, this may be a waste of time.

Thanks for responding!
Shingoshi

[Shingoshi]

However in an ideal world where you have Shanghai processors with unlocked multipliers and vcore adjustment up to 1.6v you are still only going to hit ~4.5GHz. This isn't an ideal world though so I doubt you have unlocked multipliers and the desire to wire in 4 vCORE mods on that board, so I'd say you are better off with a nice watercooling setup.

How exactly would this work? The Shanghais I have are only 2.3GHz. So reaching anywhere near 4.5GHz would be nothing to scoff at. Even if I only got a 50% increase on processor speed, that would still be nice. That would be like adding two or three (for the Istanbuls) more cores per chip. So what exactly would be needed to achieve what you think is a limited benefit? Do the Istanbul processors have the unlocked multipliers that you speak of?

Let me know here...
Shingoshi

[fatguy1992]

Wouldn't a water chiller be a better idea? You could still get negatives and would be easier to build.

[Shingoshi]

Wouldn't a water chiller be a better idea? You could still get negatives and would be easier to build.

I'm involved in a project for which I'm considering the best solutions available. I've already done a water-cooling setup on my quad-socket board. But as I started evaluating the goals of this current project, I started questioning whether any liquid-cooling setup would be appropriate.

For one thing, I have components which may not be suitable to reach the temperatures that I have envisioned. One of the components (my reservoirs) are only capable of 40F before being brittle, according to the company. They were never intended to be used in this manner. They're nothing more than water-filter housings. Albeit, very large housings, at 20" deep and @7" in diameter. When I thought of the possibility of them rupturing, I became somewhat discouraged. Especially since if they are in fact prone to being brittle at 40F, that means I likely couldn't safely go lower than 45F without risk. And while most who water cool would likely be pleased with 45F, I'm not. I am looking for sub-zero Celsius temperatures. I could simply replace them with stainless steel containers. But that raises the price substantially.

So even though I've spent money on these housings, I would rather not use them if they're going to prevent me from reaching my design goals. There's also the possibility that I could keep them in the design, but just mount them outside of the case to provide liquid-cooling for my condensers. So that's what this is all about for me. The other thing is simply their size.

What I'm building must fit inside of a Pelican 1780 Transport case. The case while being large, in fact the largest of it's series with 42x22x15 internal dimensions, is still a limited environment. Both of the reservoirs would take up most of the bottom half of the case, while providing only a limited benefit to dropping my temperatures. And then there's the issue of fragility. I thought about standing this case up vertically. So I would have to be careful that it would never fall over. If it did, chances are the reservoirs could break. Condensers wouldn't have that drawback to be concerned with. And mounting these housings on the outside, only raises the risk of breaking them.

So, I'm inclined for a number of reasons not to use liquid-cooling in my project. But then I'm also concerned about the relative cost of each solution. Most of what I would use for a liquid-cooling setup would likely need to be purchased new. While many of the things for a phase-change system could be used. Although liquid-cooling would be easier to manage, it wouldn't give the kind of results I would expect from phase-change.

So that was why I was asking these questions...

Shingoshi

[[XC] gomeler]

Any particular reason why it must fit within a transport case and why you want the processors to run sub-zero? Sometimes doing things just for the "fun" of it results in a huge headache.

[fatguy1992]

Correct me if I'm wrong, wouldn't a four head SS unit be really hard to tune?

[Shingoshi]

Any particular reason why it must fit within a transport case and why you want the processors to run sub-zero? Sometimes doing things just for the "fun" of it results in a huge headache.

That four-socket board is only one of five going into this project. I am creating a centralized cooling solution for multiple motherboards in a single case. And I already have the case. That's not going to change. The case itself represents a complete delivery system for multiple components. The computer cluster is just one of them.

Correct me if I'm wrong, wouldn't a four head SS unit be really hard to tune?

Why would four heads all coming off of a single line in parallel be any more difficult to tune than a single head? Tuning for the line should tune all four heads, would it not? And then I guess it depends on how narrowly you define your parameters for tuning. But with 12 processors in the entire system, tuning for me may be something completely different than what you're thinking of.

Shingoshi

[apt403]

You'd have to tune to system so you'd get the right amount of flow to each of the heads for whatever load they happen to be under, and I'm not quite sure how that would work with 4 processors which would all be at different loads depending on what you're running, unless you did something like throwing TEVs on every line. Not impossible, but I imagine the system would be a real headache all the way through the build.

Plus you're going to need a a pretty beefy compressor and condenser for the load you're going to be generating. I've seen the Pelican case you're using, assuming this is the system you're talking about, and you're looking at adding a ton of weight to the system.

[Shingoshi]

1.) You'd have to tune to system so you'd get the right amount of flow to each of the heads for whatever load they happen to be under, and I'm not quite sure how that would work with 4 processors which would all be at different loads depending on what you're running, unless you did something like throwing TEVs on every line. Not impossible, but I imagine the system would be a real headache all the way through the build.

2.) Plus you're going to need a a pretty beefy compressor and condenser for the load you're going to be generating. I've seen the Pelican case you're using, assuming this is the system you're talking about, and you're looking at adding a ton of weight to the system.

1.) I don't think I need or care to be concerned about precise tuning. I'd be comfortable with simply providing that all processors are cool at all times. For me, that's enough. There's not going to be so much of a difference in cpu temperatures that anyone of them are going to fail. Neither will the cooling system suffer because of it.

2.) I'm already aware of the size of the components needed. That's not so much of an issue either. I mostly want the system to be enclosed, for reasons I don't need to explain here.

Shingoshi

[apt403]

True, but if it's way off you could damage the system (Compressors generally don't like it when they get flooded with liquid refrigerant.), and it might be overly difficult to build and tune correctly for someone who doesn't have much experience in HVAC.

As far as a low profile evap. goes, I imagine the any possible space savings would be far outclassed by the size of the insulation you'd need around the suction line.

This computer is going to be truly beastly no matter what you go for, so it's all good.

[Shingoshi]

1.) True, but if it's way off you could damage the system (Compressors generally don't like it when they get flooded with liquid refrigerant.), and it might be overly difficult to build and tune correctly for someone who doesn't have much experience in HVAC.

2.) As far as a low profile evap. goes, I imagine the any possible space savings would be far outclassed by the size of the insulation you'd need around the suction line.

3.) This computer is going to be truly beastly no matter what you go for, so it's all good.

1.) Do you really think the temperatures between different processors would be so much, that any one of them wouldn't cause proper evaporation? If you do as a general rule of thumb, "provide evaporation for the lowest temperature cpu, and any processors of higher temperature will automatically evaporate all refrigerants used.
2.) I'm interested in stacking multiple boards in the space of 2U each. That's why the request for low-profile evaporators.
3.) That's an understatement!

Definition: Lowest temperature cpu == Processor with the lowest load

Shingoshi

[apt403]

It depends on how well it's done, how you rig up the heads, and what you use for metering. If done properly, it shouldn't be a problem. There's a thread in this section called Dual Head Singler Stagers, about halfway down FLuxUS posted a diagram of a system with 4 evaps, that how I would probably go about it. A single stage unit would need an insane compressor, and metering would get even more tricky.

Okay, that makes sense for the low profile evaps then. But I'm still not sure how you'd get all the suction lines to fit sandwiched between the boards. Anthing's possible I suppose, just depends on how much work you're looking to put in.

There's also the problem of expelling the hot air from the condensor, and cooling the compressor down properly.

I'm beginning to agree with fatboy about the chilled liquid setup. There are obviously problems inherent with the design, the ones specific to you you've mentioned. However, as far as cost goes, the overall price may be much higher with a phase setup. You're looking at buying the compressor(s), condensor, filter/dryers, TEVs copper tubing, specially designed evaps and possibly suction line (I haven't seen an evap that fits your requirements, and the braided copper ones most people use may not be appropriate for your build.), along with copper piping, cap. tube, refrigerant (which requires an HVAC cert. most of the time for most refrigerants, unless you choose to go with something like R134a), inert gas for purging the system while you braze it up, the electronics for the system, and possibly tools for the job, assuming you don't have MAPP or oxy/ace torches, a vacuum pump, pressure gauges, pipe bender, etc., etc. There's also the cost of RUNNING the system, which will most likely be MUCH higher than a chiller.

So even though you can get the majority of all the parts and tools used, you're still looking at quite a bit of work, research, learning, and above all, cost.

[Shingoshi]

The idea of using a chilled-liquid cooling system seems more and more attractive as I get further into the consideration of this. So say that I do go with a chiller;
1.) How large of a compressor am I going to need?
2.) Just how much less complex is the setup for such a system?
3.) Don't I still need some of the very things that you've mentioned above?

I guess I should just check out the chilled forum for hints of what to do. After reading a few links, I'm wondering if there is a standard definition of a chiller. I was thinking of using something like a water cooler, just larger.

Shingoshi

[s0lid]

The idea of using a chilled-liquid cooling system seems more and more attractive as I get further into the consideration of this. So say that I do go with a chiller;
1.) How large of a compressor am I going to need?
2.) Just how much less complex is the setup for such a system?
3.) Don't I still need some of the very things that you've mentioned above?

I guess I should just check out the chilled forum for hints of what to do. After reading a few links, I'm wondering if there is a standard definition of a chiller. I was thinking of using something like a water cooler, just larger.

Shingoshi

1. Around 1hp rotary should be enought, but that's only a quess.
2. Way more simple, just need some cheap CU blocks, 90 degree fittings and some insulation around tubings.
3. If you mean tools, yes you need some various tools for making a phase change cooler, tho if you use capillary tubing you will get with little cheaper costs :P

[apt403]

Yeah, a 1 hp rotary would be a good estimate, depending on how cold you want to go and what you use for refrigerant. If you don't want something that sounds like a V8 in your case, you could use hermetic or semi-hermetic compressors, but those are more expensive, especially for the kind of power you'd need.

You will need SOME tools that would carry over to phase building, but if you wanted a chiller you could take a standard window AC unit and slightly modify it for your purposes. You'd essentially just have to strip it down, and put it into another housing in such a way that you would have the evaporator sitting in a tank of ethylene glycol or propylene glycol mixed with water.

Or a plate HX with refrigerant running through one side and the water/anti-freeze through the other. That would save on size, but you'd be looking at some minor modifications to the AC unit. Namely vac'ing the refrigerant, brazing on the plate HX in place of the stock evaporator, and recharging the system.

Or you could use a TEC cooler to chill the water, that's even easier. Although you're looking at temps that will be higher than the phase setup, you massively cut down on complexity and cost.

[Shingoshi]

1.) Yeah, a 1 hp rotary would be a good estimate, depending on how cold you want to go and what you use for refrigerant. If you don't want something that sounds like a V8 in your case, you could use hermetic or semi-hermetic compressors, but those are more expensive, especially for the kind of power you'd need.

2.) You will need SOME tools that would carry over to phase building, but if you wanted a chiller you could take a standard window AC unit and slightly modify it for your purposes. You'd essentially just have to strip it down, and put it into another housing in such a way that you would have the evaporator sitting in a tank of ethylene glycol or propylene glycol mixed with water.

3.) Or a plate HX with refrigerant running through one side and the water/anti-freeze through the other. That would save on size, but you'd be looking at some minor modifications to the AC unit. Namely vac'ing the refrigerant, brazing on the plate HX in place of the stock evaporator, and recharging the system.

4.) Or you could use a TEC cooler to chill the water, that's even easier. Although you're looking at temps that will be higher than the phase setup, you massively cut down on complexity and cost.

1.) One of my main reasons for enclosing this system, was to cut down on the sound produced by it. So your suggestion about hermetic compressors "sounds" nice!

2.) Yeah. I've taken notice of Naja002's success with doing something like that.

3&4.) I'm combining my answer to these two comments. I've already thought about using a HX. But I also think a TEC could easily work in conjunction with the HX. The way I'm planning on using my reservoirs, I intend to run cold water into them from the HX to be mixed with the liquid contained there. It would then be drawn out and pumped into the rest of the system.

The only thing that really bothers me about using TEC's, is that they're expensive in terms of power usage. But I may be wrong, having not fully compared the cost. The initial cost of installation would definitely be cheaper. But the operational costs would be significantly greater. The real advantage there being the extremely reduced size of the system.

There's one other form of refrigeration that may be too exotic. But if it were practical for this application, I would consider it. That technology is Thermoacoustics (Wikipedia). I just haven't found out yet what the dimensional or power requirements are. Having the smallest system isn't just preferred, it's a necessity. I know that the performance is somewhat lower than a standard refrigeration system. The primary advantage there being no moving parts.

And then there's something a little more involved to consider. I've had a long fascination with Vortex Tubes (VT). You need to look them up on Wikipedia and the web. Other than needing a compressed gas supply to operate them, once again there are no moving parts. So I've been wondering about using a standard refrigerant compressor, and directing it's output directly to a liquid-cooled HX. I would then power the Vortex Tube with the expanding liquid coming from the HX, using the VT as an evaporator. I would be getting much colder gas from the Vortex Tube than without it, and be able to cool the hot end of the VT with the liquid-coolant leaving the HX. Another interesting thing to note, is the high sound levels produced by VTs. I've been wondering if that sound could drive a thermoacoustic cooling system. As far as the sound goes, it can be muffled by being enclosed. Like a trumpet with a mute.

The last two paragraphs may be nothing more than wishful thinking. So don't pay too much attention to them.

Have any of you seen small axial-flow compressors, strong enough for refrigeration? Ok, this is crazy! I found this doing a search for "axial-flow compressors: http://www.ecrater.com/product.php?pid=2509487 While it's not in the least what I was thinking of, I know that it has uses beyond what was originally intended for it. Just look up "electric superchargers" in any search engine.

Shingoshi

[apt403]

Hermetic compressors are still loud, just not as loud as rotary compressors. Just think a window AC unit, most of them use hermetic or semi-hermetic compressors from what I've seen. There's a thread around here for a single stager (IIRC) that was being built to be as quiet as possible. The end result was the unit inside a massively insulated case, kinda looked like the box that Silent PC Review uses for decibel testing.

[Shingoshi]

Maybe I should just sit the thing outside, so I don't have to hear it. I mean, it will be in a waterproof case after all!

Shingoshi

[apt403]

haha, that is true.

'Course, you'd prolly spend about as much on locks and chain as on the PC itself.. I'd probably have guard dogs surrounding that thing if it was outside.

[quintus]

If noise is a problem maybe you should use 2 or 3 compressors in tandem
Nf11fx or what you guys got there in USA

[[XC] gomeler]

A NF5.5 would be smable to handle a pair of chips in an Xserve(stock speeds so max 280w) all while being easy on the ears. Still think Xserves belong in a rack whirring away

[Shingoshi]

haha, that is true.

'Course, you'd prolly spend about as much on locks and chain as on the PC itself.. I'd probably have guard dogs surrounding that thing if it was outside.

No way in hell would I let this thing out of my sight! After spending all the time and money it will take to build this, I would have to be CRAZY!

Shingoshi