Some more FrankenResults

[FrankenDesign]

The before:

http://h01i.imgup.net/torture2ee1.jpg

4 weeks of efforts to get past where I was, 4.7ghz, I switched from air cooling to water cooling & custom built my own loop out of nothing except for the cpu block.

This week's testing has got me one more notch up on the water loop

http://m27i.imgup.net/moretortur599b.jpg

4.8ghz stable enough to torture it; this was where all my previous efforts fell over on the air cooler; so I got one step further with all the huffing and puffing of building the custom loop.

To get there I made a big compromise; I turned down every bus in the bios to 800mhz; memory, hypertransport link and north bridge; this purely isolates the processor I hope, on the clock multiplier.

The thermals just stay on the edge of what is tolerable for a piledriver.

4.9ghz doesn't fly on the torture test even with a low system bus; yet again, the heat is doing for it.

But all is not lost. I have some more tweaks to make to my loop; first one is, I have got some copper fins that I can solder to my Jardiniere; I'm going to cool that water some more somehow.

I have also made a secondary reservoir for my other pump bought when there was a fear about the first's ones 35c operating limit. That fear was unfounded; the water temp stayed at 35c.

So the second pump is only 8psi or 2-6m head compared to the 2m head of the first one; so I'm going to stuff that into another reservoir I made which will be sealed with glue; and that will be moved up the loop to pull off the rad and back to the rez.

I want to increase the flow of the return feed; so either it will work or else, the pump will actually increase the pressure and reduce the flow;

But I won't know until I try it.

i just used an empty plastic container that is quite small to just fit the pump and added some fixtures and fittings & glue. If I put that on my second or third turn on the rad. it should pull off the cpu block and push out to the res and hopefully increase the flow to further than just over 1l per minute return feed.

I'm hoping it will accelerate the flow; but I might be totally wrong.


http://o57i.imgup.net/frankenfix3664.jpg

If I open up that middle U bend and put the pump there... maybe.

My loop is:

Submersible pump in 5l copper jardiniere with 15mm outlet running to cpu block with 10mm feed inlet and 10mm outlet switching up to 15mm coupler on the rad going through 12mm pipe through the rad back to 15mm return feed to the rez.

That's res. pump , cpu, rad, res

So the aim of the game now is to increase the flow. And cool the water with the heatsink fins I have got here.

 

[FrankenDesign]

I took down the system and ran it by the sink. With only the radiator, short tubing and the two pumps in series the flow was 4 litres per minute with one pump and when I switched on the other the flow rose to ~9 litres per minute.

Based upon a previous discussion on here I had started to realise that there were a few problems in my loop; I thought the cpu block was to blame but actually it was some of my dodgy couplings.

These plumbers fittings; the water channel actually narrows at the coupler for no evident reason there is simply a small channel and a wide fitting; if you get a 10mm there is a lip making the channel 8mm. If you get a 15mm -10mm coupler the channel is only 8mm as well. So I've decided to solve my problem in a couple of ways.

First one: Drill out the coupler with a 10mm or 9mm drill bit. At least the channel will be 20% wider. Second one: Time to get rid of the compression fitting altogether; what I can do is connect the 15mm hose directly to the 10mm side of the coupler without bothering with compressing the join. The screw thread holds it snug. The 15mm tube is 12mm id; but the screw threads lend the coupler another 2mm. Something about it works anyway. The tube cannot be pulled off. It has to be rotated. & I can simply use hose clamps on it.

3 of those to replace around the loop; then I will know more about the cpu block. I'm also considering disassembling it and drilling out the top for a larger aperture; it's only perspex with an 'o' ring & screws to keep it water tight. So all I'd need for that is a couple larger fittings for it.

I might also replace the tubing with 3/4 inch; because according to some recent documentation that diameter tubing has a negligible pressure drop. Why? Because my pumps are relatively weak. They perform well on the rad by itself; but the tubing, fittings and cpu block + maybe gravity has an effect on increasing the pressure and reducing the flow.

So each tweak I hope will improve the flow and therefore the overall cooling capacity of the loop. I shouldn't have to buy too much if it all works; a few fittings, and a few metres of tubing to rework it. Larger tubing will allow me to use larger couplers that will have less of a pressure drop once they have been drilled through.

The plumber's won't open 'til tomorrow. Back to the drawing board. I think there are still things I can do to improve the performance of my loop before shelling out on a cpu block and uber pump.

 

[CountMike]

Just to remind you about bit of physics. Liquids soak up heat much slower than air, they need some time to receive heat energy from heat exchanger ( CPU pickup and radiator) and too fast fluid flow may result in overheating.

 

[FrankenDesign]

At 1 GPM it takes 264W of heat to raise the water temperature by 1 degree. Ever wondered why rivers do not freeze as easily in winter? Because of flow. If you want to read up on the math and some other details go here.
-
1 GPM is about 3.8 LPM, and the low-end of flow we like to see is around 0.5 GPM or 1.8 LPM. Of course, the higher the flow, typically the better delta-T and better cooling ability.

For reference, closed loop coolers typically run just under 1.0 LPM.

-

http://www.tomshardware.co.uk/answers/id-3059444/cpu-water-cooling-suggestions/page-2.html


A couple of quotes from this thread that got me on the trail of flow; So I guess the question there is, precisely how fast is too fast that causes overheating? With the tweaks I've made the water temperature has dropped from 35c to 31c at full load; but the cpu can still go up to 55c on the package and ~75c on the core.

@4.9ghz that is; when stress testing. I passed a 4.8ghz stress test on p95 for 2 hours which is enough for the amd overclock club; 4.9ghz falls over after 15 minutes.

Testing by the sink I got 4l/pm going on the return feed; with the second pump that leapt up to more than double; all in accord with the known hmaxes of the pumps at 2m and 6m head respectively.

With the cpu block attached in situ the flow drops radically on the return feed to 1.5l/pm. The pressure drop must come from having to pump vertically to get to the rad from the cpu; the narrowing of the tubing from 12mm to 8mm; and the narrowing of the coupler channels from 10mm to 8mm as well.

 

[CountMike]

The worst bottleneck in CPU cooling is heat transfer between CPU core and cooling body.

 

[FrankenDesign]

but dealing with the peripheral ones add up to a goodly increase in performance. By annealing the base of the copper jardiniere I'm using for a reservoir and getting the crud and oxidation off the inside the water temperature dropped 4c. I was running at 35c, the edge of the operating temp. of the pump; it doesn't get there any more. 31c on med. games and maybe 33c max. all that shows is there are plenty of tweaks I can still tune into.

I was just reading the WC thread and their flow was at 1.5gpm giving them 650w dissipation depending on the fans and rad. they used.

My loop isn't doing a 5th of that.

I just calculated the volume of my rad at 1107225mm my delta t is 7-9 my oc TDP @ 209w + 45w for the two pumps 254w with a flow for arguments sake of 1.2lpm or 0.26 us gallons per minute;

but some of those figures are only estimates. The other rad. with an area of 1190000 @ a delta t of 10c only dissipates 235 watts.

However on my rad the water turns 5 times through 6 copper 12mm pipes before returning to the reservoir.

I have got a box full of aluminium fins leftover from my arctic a11; I could jam them in between the fins of the rad. to increase the area a lot more. That's something I can do now.

So the math has all gone pear shaped because it's all estimates & I don't know how to correct the variables as they apply to my radiator. Roughly speaking it looks close; but no cigar.


The overclock of a thousand megahertz begins with a single watt.

 

[CountMike]

Some stuff must be done with trial and error method but there are many kits that can provide a starting point to build upon or modify for particulates your system may present.

 

[FrankenDesign]

It will take more than a mere kit. You need encylopedic knowledge to know what to buy; then overclocking your system on a water loop officially qualifies as an arcane art.

Plus I am in the process of inventing. My Jardiniere is set to become a reservoir and a radiator. Tomorrow I'm going to buy some solder and flux; Then I'm going to solder a heatsink to the base & copper fins all around the circumference. I'm going to try to get them edge on so they look finer but if that doesn't work I'll solder them flat on. It's pretty hard to solder a small area of a conductive metal. You have to heat the whole thing and then get a hot iron or else cool the whole thing in a bucket and just heat the area you want to solder. Nothing about it is simple. When you solder the base if it gets too hot the fins fall off; if you solder the fins the base falls off; hopefully between the stove and a pail of water there is a balance point where it will work. As long as I don't try and solder them too close together; or if I stick 2 or 3 together at the same time it will probably be ok.


That will then become the first victorian radiating reservoir with decorative twirls.

 

[CountMike]

That's why I said to take a kit as starting point, Not to actually buy it but just to see what each component does and it's features. Most of those parts could be substituted by parts from something else. There was one guy that used car heater for radiator for instance and another one main rad from a small car. A motorcycle rad can also be used.
Some are using high capacity fish tank pumps etc.
Adding active chiller from water fountain in the circuit or thru heat exchanger is also a possibility. That can bring whole system down under room/ambient temps.

 

[FrankenDesign]

i've done the motorcycle transmission cooler & the aquarium pump which can do 1 g/pm if there is no pressure drop; but in situ the return flow drops to a dribble of 0.26g/pm.

I thought of getting some sort of water chiller but I am somewhat stymied by that one; where do you actually find something like that that will actually fit on your desk, doesn't cost a fortune, that could actually be connected to the water loop without any nonsense?

Your fridge is huge. So it's just a box but you would have to drill holes in it to get pipes through it and other stuff; or if you are going to buy components for something like that; where on earth do you find them? Usually the only places these sorts of things are found are in retail outlets with huge price markups and don't take well to being modded.

Heat exchanger? Well what's that? How is that different from a radiator? Thought they were the same thing. Again, what is it, where do you find it, how much does it cost, how does it fit, who knows? Can't go around asking for doodads and thingymajigs.

The motorcycle transmission cooler is 31.5x18.5x1.9cm giving me a surface area of ~110million mm with 11 fins per inch on 6 tubes with 5 bends for the water to flow around.

The other off the shelf 120mm rad. with an area of 1190000 @ a delta t of 10c only dissipates 235 watts.

But it doesn't have as many turns and only 8 fpi.

Do the turns count for anything? Thought I read something about that. Bits of jargon swimming in my brain

As for people alarming about condensation; I've already said dozens of times I don't want to go below ambient; I'd be happy with maintaining the water temp at 20c; that is good enough.

I've looked around and searched parts, quite a lot but I have not found anything that offers good information, or affordability, that doesn't make too many compromises but it is starting to look like the market is a bit cornered; unless you have that encylcopedic knowledge.

So far all I have been able to think of is increase the surface area of my jardiniere, stuff some more fins into my rad, get more fans to huff on it, solder heatsinks to the base of the jardiniere reservoir, with fans huffing on them, and applying bigger tubing and drilling out the couplers to make the flow aperture 20% larger.

These are tweaks that are only going to cost a few pounds; a few quid for the tubing, a fiver for the solder and flux, that require more effort and mcgyverish ingenuity.

I'm not looking to fork out on high end gear; just seeking the knowledge to make this work for me. It is evident I need more cooling. So I am going to have to try some more. The radiator isn't a disaster. It only cost £5 so replacing that if it simply ins't good enough, is not a bother. It could be relegated to secondary or gpu.

However; I am nowhere near the end of tweaks I can still do without spending money so that's what I'm going to give a go at today; I will post results as I complete the work.

 

[CountMike]

Water fountain chiller is usually quite small part inside and could be easily taken out.

 

[FrankenDesign]

Googled that and the cheapest one I found was £50; that's entering the range of new cpu block, or a power supply for some peltiers & even a bunch of heat sink mounted peltiers. Other sites only offered glossy brochures; usually a sign of expensive: over £1000 to tear down? Yikes. Totally no point in that. I'm going to see if some drilling & tubing will do!

 

[CountMike]

Find an used one !!

 

[FrankenDesign]

It doesn't exist there is nobody who don't want that

a 500l water butt & pump is feasible for £50. But there isn't any documentation saying it could run 24/7; but it has an insane flow & a 12m head. Any thermal benefit to that huge volume of water? At least it would actually fit under my desk.

The heat exchanger is a better idea. Starting at £10 used with copper and stainless steel; that will go on my to do list. Question though; if heat exchangers are THE sol'n why hasn't the pc water cooling industry adopted them? Hard to credit that nobody has researched that.

 

[FrankenDesign]

Right well today's aggro has got me

http://v67i.imgup.net/widerflow5dcc.jpg

one more step. the fitting on the right is the 10mm I drilled out; the one on the left is the 10mm with the 8mm channel; and the one on the top shows that I can simply attach the 15/12mm pcv hose directly to the screw thread rather than using 12/10mm hose with a compression coupler nut and olive; the thread actually cuts a counter thread in the pvc so it doesn't pull off. The jub clip is for extra safety. Water is ultimately runny stuff that gets everywhere in small amounts through any crevice or nook; can't leave it to chance.

So this will widen my flow channel and hopefully counter some of the pressure drop. 8mm to 10mm is some math I can understand: 20% is not insignificant.

I also got some more 3/4 reinforced flexihose to go on the larger 15mm side of the coupler; so I get no pressure drop or very little from the hose.

This for the cause of re -working the loop; I can come off the pump 3/4 inch from the reservoir; from there go to the cpu block; but before the cpu block I'm going to put the second pump. It seems to me that it makes sense to put the most shove at the point where there is the most pressure drop. I know it's not my rad because testing that by the sink with both pumps gets me 10l/pm flow. Huge. So it has to be the couplers and the cpu block causing the pressure drop; since 3/4 inch doesn't register any pressure drop according to some documentation I found online; If I go 3/4 inch right up to the cpu block, to a pump, and then have the coupler step it down to 10mm as close to the cpu fitting as possible;

If you get my idea; it will shove the water through with the most force the second pump can generate being right next to it practically; so hopefully that will get me my 1gallon per minute idealized flow according to the theory I've been provided with;

If the pump just can't do it from that position I will know anyway that the cpu block is the worst pressure drop bottleneck in the loop; therefore I might consider one with wider fittings diameter/modding it with bigger fittings/or an entirely new cpu block designed for high flow.

I'm shutting down a few hours to apply my mods.

Back; I was slowed down seriously by what must've been the king of drips; I'm talking Elvis. I wrapped it in tape and glued it with the hot glue gun; wrapped it in more tape & more glue; and more glue and more tape; still it dripped. More tape more glue. Finally stemmed it. I think the problem was down to the 3/4 inch reinforced hose being of a spiral so I could not get any hose clamps or cable ties to sit tight because it always crossed a wedge of plastic. Got it now anyway. Maybe I will buy a bigger hose clamp as well, and something to wrap it in. Just one of those things, a perfect storm kind of drip. The one that brings you to the edge of the remains of your sanity; the kind of drip that could end your career.

All that bother huffing and puffing replacing the hose with 3/4 inch and the 8mm couplers with 10mm drilled out couplers got me another maybe 300ml flow; again by the sink, with the rad. and both pumps, the return flow was huge; 10l/pm easy. Connected to the cpu block in situ; the return flow is about 1700ml p/m. Still getting an 80% pressure drop from somewhere; the first pump is feeding the second pump which is feeding the cpu block feeding the rad feeding the return to the res.

Something weird has happened to my temps though; according to the therm on my desk; the in temp (ambient) is 19.5c. I am watching the water temp slowly creed up 0.1c every couple of minutes and it's 21c as I type. That is a delta t of 1.5 from a cold start. It will probably warm up a bit more after a while & certainly under a load. But the temp. isn't rising as fast as it was. The cpu package temp is 25c. and the core temp is 25c. I'm now at 4.9ghz.

The temperature has dropped huge and I can't really explain it. My previous cold starts I think warmed up faster than this. I think I have gained something for my efforts; hard to say how much guess I'll just leave it to see what happens. Maybe the pump has to work less hard pumping through 3/4 inch; and puts out less heat.

http://h78i.imgup.net/4917mhz0921.jpg

This is the result so far on a prime95 torture test; looks like it's going to make it this time! That's the 4.9ghz badge; only half an hour to go to make it 2 hours. I'm running it while I browse and type; it's not locking up on me.

I took the jpeg saved & uploaded it and edited here no problems; the delta t at 100% stress + a bit of typing is 10.4c ambient is 21.1c. The water temp is 31.5c.

http://i40i.imgup.net/4917tortur1538.jpg

Tadaaaa!

I knew it was going to make 4.9ghz.

Still got a couple more tweaks in mind to finally turn the corner on 5ghz if that is possible; I have decided to tie clumps of fins together on the rad. and solder them at intervals + the end pieces to give some better contact & also to give it back it's rigidity; that should do more something for it; plus I'm going to solder the heatsink on to the jardiniere; next time I'm going to get the gorilla glue on the drips.

idle water temp has dropped back to 25.7c from a peak load under p95 torture for 2 hours of 31.5c. As ambient is ~20c I get a delta t of between 5-10c; sometimes ambient goes up to 21c guess that is the pc putting out heat into the environment near the radiator & so on.

Got a heatsink stuck to the base but it wasn't the 150w arctic freezer I had left over. That just wouldn't stick I couldn't get it hot enough with the gas stove, heat gun and soldering iron all together; I would need a propylene torch. This presents another problem; if it gets too hot the heatpipes could rupture ejecting superheated steam. That idea is looking like scratch that.

I had a nice copper core heatsink from goodness knows where probably an old pentium 4 and that stuck & I only needed the heat gun. It does not seem to be doing anything. I have got fixtures for it and I attached a 80mm fan to it plus an exhaust funnel; nada observable difference.

So I might try a high volume fan on it but as I don't have an external pwm controller I'm a bit stuck; it is a pretty simple circuit to make you need a resistor only I think but otherwise pwm fans simply don't work on dc unless you can trip the speed regulator circuit.

I didn't mess with sticking the fins on the side of the jardiniere; it is quite doable but I spent so long trying to get the arctic freezer to stick that I just wanted to get my pc up and running again. As I only have the one reservoir. So I'll come back to that another day; maybe later in the weekend. I think it's worth it to polish up the delta T a bit; I'm not happy with a mere 'good' 10c at full load.

 

[4Ryan6]

Question though; if heat exchangers are THE sol'n why hasn't the pc water cooling industry adopted them? Hard to credit that nobody has researched that.

ROFL :lol:
What do you think a radiator is?
It may not be called a heat ex-changer, but that's exactly what it is.
It exchanges heat from the coolant running through it to the air flowing through the fin grid field.

 

[CountMike]

I meant heat exchanger like in boats. A closed loop radiator in environment water flow. That way dirty or salt water doesn't enter engines.

 

[FrankenDesign]

I know that radiators are heat exchangers! But they are commonly called radiators. heat exchangers such as that you find in a combi boiler are commonly called heat exchangers; they are also closed loop fluid heat exchangers. Since they are only a few quid I was thinking well what if I got an air compressor and another radiator and simply pumped the water through it before hitting the cpu? Only thing I don't really know is how powerful an air compressor I would need. I was also just about to look up refridgerant gases; I was reading earlier it is simply the expansion of the pressurised gas that provides the cooling so you have to go through a variety of different size tubing as the fluid changes phase. Not quite sure how it all fits together but I've seen air compressors for a £10, heat exchangers for a £10 and fridge radiators with fins and all for £25 brand new. Don't know if there is some assembly I can get from somewhere maybe a box fridge or something.

So it would essentially become dual loop; one loop to chill the water and a second loop to chill the cpu. but I suppose if the water was chilled enough I could move the rad around for the heat exchanger instead. So that would go -res-exchanger-cpu- and the second loop would go gas compressor-rad-exchanger; where the fluids meet in the heat exchanger would be the join in the figure 8. That would make the heat exchange chamber the evaporator.

In that instance the radiator is the condensor. I think they use narrow gauge tubing for that so the refridgerant fluid does not decompress before it hits the evaporator. & I think you need a couple of valves on it & another interesting question is where to put the thermostat; or even if you could connect it to the cpu pwm monitor; for you could chill the water in the reservoir for example, to 15c but the cpu will heat it up; where if you chill the water to 15c as it goes through the block the actual water temp. will be lower; guess it depends. If you want the water to be a steady 20c or to maintain a delta T of a certain level you will need a variable control.

Looked up refridgerants; there seem to be quite a few problems associated with all of that: a laundry list of all your worst nightmares. Most of the gases are toxic, polluting, flammable/explosive, require too much pressure to operate safely or else are simply too cold. It must be doable somehow since the refridgerator exists; but the precise details of what refridgerant to choose & also what fittings will make it all join up, beats me. The necessary circuit looks a bit clever too.