1 Big Loop vs 2 Loops?

[Vinas]

Shunail, for no OC you will almost always be just fine with one loop. That PA 120.3 and 120.2 is pretty massive overkill for a stock system, though! With that kind of cooling I bet you could do two CPU and two GPU's at stock no problem.

[Hannibal Lecter]

I guess I see your point. However, I like to keep it simple. Making two loops is not cost effective when compared to making one properly optimized loop. Its very easy to thermally isolate your components with the proper radiators while maintaining a single loop. In your case, you are describing a loop that is not properly optimized as far a heat load and radiator capacity is concerned, which IMHO is nonsence. Just my .

Could you please explain how you "thermally isolate your components with the proper radiators" in a single loop? In any closed loop, the water temperature will quickly equalise whatever the order of your blocks and radiators. In other words there is no such thing as isolating your components: the heat they generate will be carried by the same water.

Contrary to what you just said, two loops are about optimizing the cooling ability of your system, not the opposite. Regarding the added cost, this is really for anyone to decide whether it is worth it or not.

@shunail, I agree with Vinas, if you won't O/C, one loop will be perfectly capable of efficiently cooling your system.

[Humminn55]

Could you please explain how you "thermally isolate your components with the proper radiators" in a single loop?

Well, you may never completely thermally isolate the components from one another in a single large loop, but you may be able to mitigate the effects of multiple blocks on each other.

Take the "typical" loop that seems to be the most prevalent around....res -> pump(s) -> rad(s) -> block(s) -> res.

Now, try out this theoretical large single loop: res -> dual DDC 3.2 pumps -> Thermochill 120.3 rad -> Thermochill 120.2 rad -> cpu (i7 920 OC'd to 4GHz on 1.3V) and X58 nb blocks -> Thermochill 120.3 rad -> dual 5870 gpu blocks -> res.

In that case, instead of a loop consisting of a bunch of rads followed by a bunch of blocks, you have rads after each group of blocks. Admittedly, you still have a temp delta coming off each group of block/rad combo, but I'd bet it'd lower than you'd experience from just having block after block after block without any cooling placed between.

And since we know gpu blocks are fairly temp agnostic, just giving some decent cooling after the cpu/nb set of blocks, a TC 120.3 rad, would probably be enough to give decent enough overall temps throughout the entire loop.

Just a thought about what StAndrew was talking about....

[Hannibal Lecter]

Well, you may never completely thermally isolate the components from one another in a single large loop, but you may be able to mitigate the effects of multiple blocks on each other.

Take the "typical" loop that seems to be the most prevalent around....res -> pump(s) -> rad(s) -> block(s) -> res.

Now, try out this theoretical large single loop: res -> dual DDC 3.2 pumps -> Thermochill 120.3 rad -> Thermochill 120.2 rad -> cpu (i7 920 OC'd to 4GHz on 1.3V) and X58 nb blocks -> Thermochill 120.3 rad -> dual 5870 gpu blocks -> res.

In that case, instead of a loop consisting of a bunch of rads followed by a bunch of blocks, you have rads after each group of blocks. Admittedly, you still have a temp delta coming off each group of block/rad combo, but I'd bet it'd lower than you'd experience from just having block after block after block without any cooling placed between.

And since we know gpu blocks are fairly temp agnostic, just giving some decent cooling after the cpu/nb set of blocks, a TC 120.3 rad, would probably be enough to give decent enough overall temps throughout the entire loop.

Just a thought about what StAndrew was talking about....

I am sorry but loop order doesn't matter; after a little while the water temperature will be the same everywhere in your loop give or take 0.5C. IIRC Vapor even proved it a while ago. So you cannot optimise your loop by changing the order of your blocks and rads, or to be more correct the difference you may see is close to irrelevant.

[jedihobbit]

Shunail, for no OC you will almost always be just fine with one loop.......

@shunail: I agree with Vinus, if no oc'ing is involved and going for quiet then one loop should work. Also if this is your first attempt, I've found KISS is the best key until you're totally comfortable with wc'ing.

@StAndrew: Hello from Chesterfield.

From a noob wc'ing person who has managed to fry two mobos. Just my

[bmg]

I am sorry but loop order doesn't matter; after a little while the water temperature will be the same everywhere in your loop give or take 0.5C. IIRC Vapor even proved it a while ago. So you cannot optimise your loop by changing the order of your blocks and rads, or to be more correct the difference you may see is close to irrelevant.

You're exactly right. Unless the water is moving extremely slowly through a rad, the water temp doesn't drop very much on a single pass through it.

[shunail]

Ok, Since I've Dual loop res, I'll go with dual loop.
Since I'm not O/C and single or dual loop doesn't matter (so far). Does it matter if I rather have Rad place last in each loop?? Meaning

RES => CPU => MOFSets => RAD =>RES
RES => N/B => S/B => GPU1 => GPU2 => RAD => RES

I think its better managed this way, gain I THINK, unless I'm wrong!

[StAndrew]

Well, you may never completely thermally isolate the components from one another in a single large loop, but you may be able to mitigate the effects of multiple blocks on each other.

Take the "typical" loop that seems to be the most prevalent around....res -> pump(s) -> rad(s) -> block(s) -> res.

Now, try out this theoretical large single loop: res -> dual DDC 3.2 pumps -> Thermochill 120.3 rad -> Thermochill 120.2 rad -> cpu (i7 920 OC'd to 4GHz on 1.3V) and X58 nb blocks -> Thermochill 120.3 rad -> dual 5870 gpu blocks -> res.

In that case, instead of a loop consisting of a bunch of rads followed by a bunch of blocks, you have rads after each group of blocks. Admittedly, you still have a temp delta coming off each group of block/rad combo, but I'd bet it'd lower than you'd experience from just having block after block after block without any cooling placed between.

And since we know gpu blocks are fairly temp agnostic, just giving some decent cooling after the cpu/nb set of blocks, a TC 120.3 rad, would probably be enough to give decent enough overall temps throughout the entire loop.

Just a thought about what StAndrew was talking about....

Correct. If you have a rad that can dissipate the same or greater amount of heat than the amount of heat introduced by a component, then this will work. However, it is commonly thought that it is more beneficial to optimize the loop order for a shorter more direct tube route. I try to do both

I am sorry but loop order doesn't matter; after a little while the water temperature will be the same everywhere in your loop give or take 0.5C. IIRC Vapor even proved it a while ago. So you cannot optimise your loop by changing the order of your blocks and rads, or to be more correct the difference you may see is close to irrelevant.

Incorrect. It isnt the same, its just that the temperature differences are minimal. They are so close that most ppl dont worry as it makes a minimal difference (ie, temps from a CPU block output are only about .5C hotter than temps from the CPU input). The argument here is that newer components are much hotter, thus making a bigger impact on water temps. Im saying that staggering rads will fix this.

For example: My loop has a 120 between the CPU and GPU's. I can run Intel burn all day with my CPU at stock and not change the temps of the GPU's. However, with it overclocked, the temps on my GPUs raise ~.5C with fans on low and minimal change with the fans on high (BI GTS with a generic 120mm on pull and Ultra Kaze 3000 on push).

@ Jedihobbit a Chesty eehh ?

[Humminn55]

I am sorry but loop order doesn't matter; after a little while the water temperature will be the same everywhere in your loop give or take 0.5C. IIRC Vapor even proved it a while ago. So you cannot optimise your loop by changing the order of your blocks and rads, or to be more correct the difference you may see is close to irrelevant.

I did read Vapor's test when he produced it last year and fell into the same belief as you have right now. But really look closely at what he did in that test.

He made a circular loop....pump, rad, block. The radiator was always after the cpu block....no matter how he labeled it "before" the block or "after" the block. The radiator always received coolant heated by the cpu. So, he didn't change the cpu/radiator relationship at all.....in that loop, he couldn't.

What he did do, though, was change where the 3 pumps were positioned in relation to the radiators, that's all. His three DDC 3.2's, using 54W of electricity to work, either dumped their heat into the coolant and then into either the cpu block without going through the radiator, or had their heat dump cooled by the radiator prior to going into the cpu block.

If you look at his chart, the average temps. he recorded were slightly lower in every test scenario (GTZ, KL-350, FV2+3.5) when pumps were separated from the block by the radiators.

Now, it's true the temp. differences were very slight, but that may be indicative of how little heat the 3 pumps put into the coolant. While the 3 pumps consume 54W of elect. while working, at least, they don't dump all the heat they create into the coolant. They do radiate some of their heat through the pump bodies to be carried off by air convection.

But the trend is definitely there.

Now, to truly see if there is a difference would require a full blown test putting several blocks, such as a cpu, nb, and gpu blocks, in series followed by a few radiators, as is typical of most setups here at XS.

Then, put a good rad between each block, like cpu -> rad -> nb -> rad -> gpu -> rad.

I think you may see a little bit of a difference in ultimate chip temps.

Why? Well, water does have a heat carrying capacity and preheating the water before it gets to the next chip, be it cpu or nb or gpu, reduces its capacity at least a little and may, I repeat---MAY---reduce the amount it cools subsequent chips it flows across.

We tend to run the loop with cpu first in an attempt to give the cpu the "coolest" water and better chance to cool the cpu to the fullest. Subsequent blocks the coolant passes across how has preheated coolant and may reduce how much it gets cooled.

Well, that's my theory anyway. But Vapor's test almost does support this theory. After all, he changed a heat producing device's location in relation to the radiators and got a small drop in cpu temps when the rads were between the pumps and cpu block.

Could the same thing happen if a decent radiator was placed between say, a cpu block and gpu block? Would the gpu be cooled more effectively?

As I said, the only way to really test is to have a loop set up like:

pump(s) -> cpu -> nb -> gpu -> rads -> pump(s)

And then retest the loop set up as:

pump(s) -> rad -> cpu -> rad -> nb -> rad -> gpu -> pump(s)

In the second setup, you may not notice lower cpu temps, but you may see lower nb and gpu temps.

In fact, to really see a difference, run the loop like gpu -> nb -> cpu then rads. Bet a dime to a dollar you'd get worse cpu temps in that sequence. Then run it as gpu -> rad -> nb -> rad -> cpu -> rad. Bet your cpu temps would improve.

Now, this would take someone with the proper equipment and lots of time and effort. Unfortunately, I lack the equipment. I have the time but lack the temp. probes, flow meters, etc. But I wish someone had the gumption to do such a test. The results would be interesting.

Like I said, just a theory. But it only makes sense that if you remove the heat from the coolant before it hits any particular block, that block could be cooled better.

[PiLsY]

Spot on - you've described perfectly what I've been trying to for a long time on here. Gave up posting on these threads where people shout "loop order is unimportant". Think i'll just link back to your post in future .

[shunail]

So, here's my new system:












And this is the temp (w/o overclocking)



XSPC RES1 => Rad 120.3 => NB/SB Block => CPU => MofSet => RES 1
XSPC Res 2 => 8800 -1 Block => 8800 -2 Block => Rad 120.2 = Res 2




What do you think?

[Waterlogged]

What do you think?

I think you would have been better served getting some GT's instead of those SilensuX's.

[facboy]

why is all your tubing so long? does it need to be?

[StAndrew]

Yeah, you can probably shorten those tubes a bit, but no big deal. Some wire management would help as well.

Also, most ppl read temps in C not F. Im not crying over it, but just wanted to point it out.

[rge]

I did read Vapor's test when he produced it last year and fell into the same belief as you have right now. But really look closely at what he did in that test.

vapor was just testing, in his own words "does placement of the pump in relation to the CPU block matter? Specifically, does the increased inlet pressure from placing the pumps directly before the block increase performance as suggested by many. "

I dont think most people use that specific test as a reason to say loop order does not matter regarding gpu, cpu, etc...or no one should anyways.

And first you need to decide is 0.3 to 1C important. If it is, then loop order may matter, if it isnt, then it doesnt.

But I have placed sensors before and after cpu block and before and after gpu block, and get similar results to others.

And if you look at single rad testing, with varying heat loads, you can gather the same info: http://www.skinneelabs.com/swiftech-mcr320.html?page=4
280W cooled by triple rad with 1500 rpm fans drops temps about 0.5C.
530W cooled by triple rad with 1500rpm fans drops temps about 1C.

So loop order can manipulate temps 0.3 to 1C (multiple gpus) in terms of you can sacrifice one component temp ie gpu for another ie cpu for that 0.3 to 1C. If you went rad rad cpu gpu gpu gpu, you would expect cpu temps could be ~1C lower (maybe 1.5-2C with hottest gpus then chained through multiple rads) and gpu temps ~1C higher versus rad rad gpu gpu gpu cpu. Or if alternated as you described, have that .5C added to one or subtracted to another...so again, you need to decide if less than 1C is important or not to answer does it matter.

But regardless, I do agree, I would be interested in an in depth test of what you described, but you would need 3 high end hot gpus, OC cpu, and mobo blocks to get enough temp difference to be interesting, and then your talking about 1.5C probably max. With my gtx 295 OCed and i7 oced as only components + my 2 large rads, I can only play with about 0.3 to 0.4C, and need 9 sensors and very careful measurement conditions with avg 3 tests to be able to measure that very small difference.