Build Log: 100,000 PPD by 1/23/2011

[charlie]

He's talking 100,000BOINC PPD..equals 700,000 WCG PPD

that's kinda' what I thought... then that's quite the goal

[charlie]

that's a lot...

like 20 i7 rigs... or half a dozen top shelf Westie rigs.

[trn]

And I was thinking I set my goals to low (since the date was 1/23/2011 )

I have:
10 - L5640
2 - 980x
2 - 970
2 - 920
1 - dual E5520
1 - 930
1 - W3520

I think I need to get rid of some 45nm's it will be difficult to plug them back in after putting them on the kill-a-watt compared with the 32nm hexas. (250w - 280w for the 45nm quads @4.0GHz vs 175w for the l5640's at 3.7GHz) So I plan to dump the 920's. Even without the 2 920's If I get everything else up and running that should be equal to 100,000 boinc. I know i'm still short on some RAM and mobo's and maybe some other components I'll figure the rest out later once I start to get some loops up and going.

[trn]

I don't know if you ever measured the power usage of your box fan but I measured mine the other day. Surprisingly (to me) high power usage for mine. It has 3 speeds, on speed 1 (low) it was 74W. Speed 2 = 86W. Speed 3 = 106W.

Not good power scaling for the slower speeds, and bad static pressure as someone else mentioned. To bad because the size and cost are attractive.

[Pointhore]

Did you ever think of running a heating circulator pump like this to get your flow problems fixed.
Low power draw, good head pressure, etc.



Link to specs. http://www.grainger.com/Grainger/TAC...C90?Pid=search

This is just an example of what you could use. There are many other pumps available to choose from.

[Movieman]

Did you ever think of running a heating circulator pump like this to get your flow problems fixed.
Low power draw, good head pressure, etc.



Link to specs. http://www.grainger.com/Grainger/TAC...C90?Pid=search

This is just an example of what you could use. There are many other pumps available to choose from.

Max. Working Pressure (PSI) 125
Wouldn't that pop every seal in the waterblocks?

[D_A]

Max working pressure is 125psi but it can only push 10' of head, which is about 4.3psi. The working pressure is to do with the pump seals, not the discharge pressure.

[OldChap]

Deja vu? Might need to consider the potable water versions made of stainless or bronze if you want to resolve the dissimilar metals thing

[Pointhore]

Deja vu?

Sorry if it was mentioned before. I read the whole thread and did not see any mention of it.
Must be my old age, I think my eyes are going.

[onex]

found you this:

this site is like the TFT-Central for water cooling,
you can see in the graph that these huge rad's might be the culprit which takes the pressure down,
the 3 radiators shown differ in the pressure drop when the GPM count goes up,
the author reffers to the fact that adding another one of these MCR220's (or any other for that matter) would lift the drop rate at up to 2 times (or a bit less as the GPM drops the resistance would drop either).
differences between radiators could be due to the curving angle of the coils, the size the fluid has to go through, the material used and it's resistance and (can't remember that word) but the actual quality of friction factor from the inside.
maybe even the substance that is being used and it's thickness.
there are ways and tools to check these drops,
link to the article:
http://martin.skinneelabs.com/Swifte...20-Review.html

that probably should be the main issue burden on these pumps

E:
just seen you actually kind of solved it!

[trn]

Deja vu? Might need to consider the potable water versions made of stainless or bronze if you want to resolve the dissimilar metals thing

Totally Deja Vu and still a good suggestion.

I've always heard not to mix metals such as AL and CU in water cooling loops. What does this cause? A brass pump would keep all my metals brass, CU, nickel (and a small amount of galvanized (zinc plated) steel.) If It slowly causes deposits and ruins a CPU block in 2 years or something minor like that then I don't care. Bronze pumps are $100 more than the cast iron equivalents. If it is worth it then i'll invest in the bronze.

With my reconfiguration and improved flow rates I need less pumping power now from a booster pump; I might even be able to get away without one but that doesn't' seem wise from what I've seen.

Also If i place 2 pumps in a loop in serial and they have different flow rates will that be a problem? MCP665's have 5.3GPM and even the Taco 1/35HP seems to crush that in GPM. http://www.grainger.com/Grainger/TAC...427?Pid=search. Or a 1/25HP also blows away the MCP655 in flow rate. Does anyone think this would create a problem with the 655's not being able to pump enough water to supply the in-line pump and the in-line pumping more than the 655's can flow?

It also seems these pumps have 3/4" interior dimension threaded fittings on them. *edit* I don't trust my research, these look like 3/4" outside diameter *edit.* My rad lines are 1-1/2" and 1-1/4" so stepping up the pipe/tube sizing should increase the head pressure help keep down the flow rate somewhat (though I estimate it would still be 2x more than a 655 )

[D_A]

You can always put some kind of restriction in the line to reduce excessive flow rates. If you put the restriction on the suction side of the pump it's motor will draw less power. If you put it on the discharge you will end up with less flow of course but the pump will be pushing more pressure and draw full power.

[OldChap]

Personally I would use a good quality auto coolant additive (anti-freeze) at a fairly low concentration then run dissimilar metals ...seems to work in cars and that at high temps. I have this just now (but removed most alu after 6 months) running the past year in a whole board under water rig and have seen no deterioration in temps. I will probably take it down before Christmas and flush etc so I won't be absolutely sure till then.

Normal rads are usually copper pipe and brass ends/tanks. I think that for the amount of time that we normally expect rigs to last we should be able to ignore the dissimilar metal argument if using a corrosion inhibitor

[trn]

aw hell... next problem.... FANS!

I temporarily rigged up all my fans for a test run tonight. Good news is they moved a ton of air and created a strong pull from inside my garage and nice breeze outside my exterior window.

Bad news is the power... these fans (despite my earlier testing) undervolt like ! They can't seem to run and draw less than ~70 or 80w or so and they are loud (minor concern.) I had all 6 connected to a Lamptron FC-6 which was overloaded by a factor of about 2! And I had that Lamptron connected with only one PSU line. I shut it down after trying my best to make these fans run. If I left them going for longer wires would burn or the fan controller would die a sparky death.

Pumps have been OK; I let them run for 24 hours, no leaks and 5 pumps run at 110w. Flow still unknown because I didn't hook them up to any blocks yet.

Time to rethink my fan's. Currently I was using:

5 - 14"
1- 12"

Suggestions?

[Sparky]

Try wiring two fans in series with each other, then to the 12V supply. That would give each fan 6 volts.

[D_A]

Remember you're using two car radiators to offload heat here, so you're not going to need a turbine to move enough air, in fact you may not need much air flow at all.

What I've been thinking about using is barrel fans, like you see in domestic room heaters. They can't push any real static pressure but they move a ton of volume, use little power and they're pretty quiet.

[trn]

I'll try the 6v, but I don't think they even start moving till about 9v. Maybe I'll test one with a multimeter first to find the starting voltage.

PPC case fans... seem to weak:
http://www.performance-pcs.com/catal...ducts_id=24345

Newark browsing for HVAC barrel type fans:
http://www.newark.com/jsp/search/bro...lues%3D1001783
smaller diameter than id like but i'm sure the static pressure of the things are great and the watts look nice on these and they are more expensive.

Something like this might do it:
http://www.newark.com/sunon/a1259-mb...vac/dp/20M3858 26.45W and it runs on 110v so I don't need to use a computer PSU which is nice.
http://www.newark.com/adda/ak2582hb-...fan/dp/58P0939 another newark option, these look more power hungry.

[D_A]

http://www.newark.com/sunon/a1259-mb...vac/dp/20M3858

These are good size fans but I'm not sure about the 52dba noise level.

[trn]

http://www.newark.com/sunon/a1259-mb...vac/dp/20M3858

These are good size fans but I'm not sure about the 52dba noise level.

Noise should be OK (in the garage so... whatever ) Those Sunon fans look like the ones to beat so far:

[D_A]

Poor garage is gonna feel like a wind sock!

[trn]

Poor garage is gonna feel like a wind sock!

I can hope Since its Thursday night I got all weekend to browse for fans before ordering monday

[fallwind]

I'd suggest using 120v fans if possible. They're cheap, efficient, move a ton of air and come with a speed switch. One of those big square box fans would be best/easiest if possible. If not, a couple ~12" desk fans would do the trick. I know whenever my car fans die, I'm going to rig up some desk fans. Remember that you don't actually have to mount the fans directly to the rad. If the rads are in an enclosure(as they appear to be), you can make cutouts on the other side of the enclosure and mount the fans there. Have them sucking or blowing air into the enclusure and out through the rads. It's actually more efficient that way and exposes more rad surface area to air flow.

Just for reference, here's some pics of a farm I built a few years back. It held 6 rigs and was cooled by 4 3x120mm rads with a disassembled 12" desk fan for moving air. The fan drew air in through the rads and then pushed it out the top. Worked great!

[Rembulan]

Sorry If I hijack this thread,

can I ask fallwind to see internal of the box I like how you manage the cooler maybe your internal box view will give me more idea for my rendering farm.

Thanks again sorry for hijack

[D_A]

Maybe if you install BOINC on your farm he'll play show and tell with you.

[fallwind]

Rembulan, PM sent.