cooling: optimal airflow ?

[Paddy]

After much reading I have come to the conclusion that you can do it either way.

If you do it all in series, then the water reaching CPU1 will be warmer than CPU0 and therefore CPU1will have a higher resulting temperature. This doesn't matter if the water is flowing fast enough. SCompRacer has adopted this method, and it looks as if he has done it well as the second CPU is only .4'C warmer.

If you opt for the parallel option (with the split) you have to ensure that the resistance down both pathways is identical or you will get a disproportionate flow down one of them.

To be honest, I dont think it really matters. Doing it in serial is much more straight forward, with less connections and therefore less chance of leakage.

I will be going for:
res > pump > cpu0 > cpu1 > rad

[VJ]

Originally posted by Paddy [MU]
If you do it all in series, then the water reaching CPU1 will be warmer than CPU0 and therefore CPU1will have a higher resulting temperature. This doesn't matter if the water is flowing fast enough. SCompRacer has adopted this method, and it looks as if he has done it well as the second CPU is only .4'C warmer.

If you opt for the parallel option (with the split) you have to ensure that the resistance down both pathways is identical or you will get a disproportionate flow down one of them.

Yes, I didn't think it mattered that much, but I couldn't find it anywhere... You're right about the resistance in both pathways (in the parallel options), this is why I added the option with the extra pump. But serial will probabely be the way to go .

And after seeing some layouts: is there an option to mount several radiators (e.g. 2 small ones instead of 1 large one) ? Or will this stress the pump too much ?

Well, seeing these available kits
(www.innovatek.de, www.dangerden.com), makes me far less reluctant of going the water-cooling route. I'm still not sure if I'll go for it though (but then again, I'm still several months away from my system).


Jörg

[Paddy]

One decent 120mm rad will be enough.

If you look at SCompRacer's rig. He has pleanty of room to spare. Even if the CPU were runing at 20'C higher that would be well within normal specs.
I guess it's all down to a choice of overclock or noise. I want my rig to be quiet, so I'd want to w/c as much h/w as possible.

[Greebe]

reservoir -> pump -> split to cpu1/2 -> recombine is correct way of doing it.

[SCompRacer]

There are basically two paths you can go down in water cooling, and there are endless debates as to what approaches to take in each. There are low flow systems and brute force systems. Block design determines that. Innovatec takes the low flow approach while the Danger Den blocks like high flow. They work with low flow but temps go up slightly. In overclocked single CPU systems, I have experimented with flows from 55GPG to 120GPH with the danger den blocks. I tried different pumps, radiators (stack plate aluminum and brass/copper cores) and line size (3/8" and 1/2"). The temp difference ranged 4C with the best results at high flow.

A reservoir is not needed, but it is a convenient way to fill and bleed the system depending on where it is located. Since that may not occur too often, some just use a T with a fill line capped off at the highest point in the system. I wanted a liquid light show instead so I did an acrylic reservoir. My first water cooled case used no res at all. I soldered two 1/8" brass pipe thread bungs to the back tank of a horizontal mounted radiator opposite the inlet/outlet fittings. It was canted slightly so the air would find the highest point. Remove both plugs to fill, just the suction side to top off even while the pump was running.

There are suggested ways of routing components. Depending on where things are mounted in the case, the order of convenience can save tubing and make a neater job. I chose convenient and neat. My system is routed pump out > radiator > reservoir > CPU 1 > CPU 2 > pump intake. I can remove my fill plug while the system is running to top off during initial fill and bleed. I do have to give the case a tip to get the all the air out of radiator upon initial fill since it is mounted with the inlet/outlet on it's side. That leaves a small air pocket in the radiator above the outlet.

I'm no engineer but this is how I understand the differences in flow. When splitting flow, blocks in parallel will see less flow than the overall system. Since block cooling requires convection which thrives off of velocity (especially on the DD blocks), plumbing in series generally yields lower chip temperatures. With a high flow rate through the blocks (I get 95GPH), I may get a higher overall rise in in the fluid temp but the result is less than 1C between the two CPU's. So in my case it's easy and it works.

[SCompRacer]

Originally posted by VJ



And after seeing some layouts: is there an option to mount several radiators (e.g. 2 small ones instead of 1 large one) ? Or will this stress the pump too much ?

I tried two radiators in parallel using an Eheim 1250 pump and only got a 1C decrease. Results depend on the temp of the water to the ambient air. The greater the difference, the more you would benefit from two radiators. I believe you do take a slight hit in overall flow but I never measured it. I have seen some posts where two small radiators were used in series and reported temps seemed good.

Another important factor is the size and design of the heater core. Most US cars use what I call the wrinkle core. Some heater cores are like the BIX with smooth cores and fins. You get higher coolant flow and airflow through them. My results between a 6" x 6" wrinkle core and the smaller BIX is only 1.5C. As expected, the larger core does a better job. You can see a visual comparison here.

[Greebe]

But the reason Dangerden's must have high flow is to get around the ineffeciencies of their water block design.

In this regard Innovatech rules

(they don't need no stinkin high flow to work)

BTW I'm currently building a new waterjacket for my system... atm I think an all copper "6 shooter" revolver cylinder design would be best seeing that I can't get my hands on bulk copper here locally (without dropping a fortune)

[RichL]

I've not done any watercooling, but if you ask me the best way would be to split the flow over both CPUs simultaneously then recombine before going into the radiator.

If you run the water from Pump > CPU1 >CPU2, then the water is already going to be heated up when it reaches CPU2.

Edit : Should have read page 3 before posting

[KvHagedorn]

Mike, why don't you get some 4 gauge primary wire (used in car stereo) and melt it down? Or do the same with all the extra junk speaker (and other) wire I know you must have lying around there

From someone who has a whole subwoofer box stuffed with old speaker wire..

[SCompRacer]

Originally posted by Greebe
But the reason Dangerden's must have high flow is to get around the ineffeciencies of their water block design.

In this regard Innovatech rules

(they don't need no stinkin high flow to work)

BTW I'm currently building a new waterjacket for my system... atm I think an all copper "6 shooter" revolver cylinder design would be best seeing that I can't get my hands on bulk copper here locally (without dropping a fortune)

No argument about Innovatech, good design and high quality components. The margin between it and other blocks widens at low flow rates, but many blocks perform similarly at higher flow rates. Most of the overclocking crowd choose the high flow approach since increasing the line size and using higher GPH pumps on the Innovatech results in diminishing returns and equal performance to the high flow blocks. Some of this is due in part to the higher flow pump adding heat to the coolant. I would like to get one for myself and see how it performs.

The new Spir@l is gaining in popularity and performs well with lower flow. There is an Australian block called the Cyclone5 which performs well but it likes high flow. The DD block is designed for high flow due to it's simple, mostly open design with bead blasted interior. In my single CPU system it outperforms his Maze2 using the same pump by 1C. Performance wise, the Maze3 on an XP1800 with no overclock in 21C ambient runs 28C loaded. The XP1800 @ 1783 in same ambient runs 33C loaded. All temps via DigDoc5 sensor taped to the CPU touching core.

Your block idea sounds interesting, never seen one like that yet. It's very satisfying to use something that you made. Not many have the access or ability to make their own blocks either. There are a few guys over at ProCooling that make blocks. Always something interesting going on.

[Greebe]

hehe Chuck.. I was doing just that this evening... meltin copper just to see how well it would work... amazing what you can do with a high BTU propane burner

[Technoid]

Our Case supplier has imported a cooling kit.
And as soon as it is available I’m going to test one…
Sadly it’s the cheapest “Complete” kit that I have been able to track down in Sweden.
Our usual “cooling” supplier has all the bits but hasn’t bothered to make a “Kit” for us newbies

And after talking with the seller I definitely have to check it before I sell any….

Me: How well do it perform?
Seller: We have tested it with AtlonXP’s and it keeps the CPU at its optimal thermal curve!
Me: And that is?
S: 50C, if they are cooler they wont perform optimally! (Sounding as if he had revealed the secret of eternal life )
Me: To which sockets do it fit? (Still a bit shocked from his previous answer )
S: It has clips for Socket 370, A and 480, 7, and 8 (Hm… Really? )


Athlon’s don’t work if they’re colder than 50 Degrees Celsius??

I think I’ll take that with a metric ton of salt

What’s your opinion?

[Greebe]

He's a flippin clueless norom

[RichardI]

Do it! A quality pair o' fans are only about £7 a piece.

I'd buy some small rubber washers to put on the screws, to reduce vibration and noise.

[Greebe]

By adding two fans in parallel (quasy isobarik) you will not improve airflow by squat. It's effective pressure will nearly double but the overall airflow will remain nearly the same. Spend the same money on a good larger single fan sucking the air through the radiator will significantly improve cooling.