Loop order?....

[jmdixon85]

So, I'm planning on getting back into gaming

With this in mind and a serious upgrade needed in the GPU department I have a question.

If I was to water cool both the CPU and GPU's does loop order really matter?

From what I have read it doesn't, I have seen loops go res>pump>cpu>gpu>rad, and even GPU>CPU.

With this in mind I am wondering how to go about it on my rig if the time (and funds) come available.

I have a HAF932 with a 120.3 rad mounted up top that currenlty cools only the CPU.

If I was to water cool a couple of GPU's within the same loop I would choose a loop order of res>pump>cpu>rad1>gpu(s)>rad2>res

Surely this is better than say: res>pump>cpu>gpu>rad1>rad2>res???

I just like the thought of the coolant getting cooled before going into the next component if you get what I mean??..

 

[GTXJackBauer]

Loop order doesn't matter as long as the reservoir is before the pump.

Going to the CPU first is probably a better idea because the GPU's would dump the heat to the sensitive CPU. Should save some degrees but the loop eventually reaches a equilibrium so it shouldn't matter too much unless you want every little degree off the CPU. Most usually choose the best looking route with short use of tubing.

Depending on the heat load you'll most likely have to hang some radiators outside of your case, preferably in the rear.

Also for example, if you went with a single newer GPU say like the Kepler (GTX 670/680/770), the 120.3 should be fine for that and the CPU but if you went multi-GPU than you sure would need more than a 120.3 imho.

 

[jmdixon85]

Thanks Jack I won't be going with any current GPU, I did think about getting a 770 but im paying for just more VRAM and not much more GPU power. Hoping the 8xx series brings more to the table

I would love to keep the cooling all internal but I dont think it's possible using my case. I could fit one 140mm extra rad but that's it as far as internal space goes.

 

[GTXJackBauer]

Yes, you could wait for Maxwell (8xx) and see what they'll have in store.

Most case design nowadays are very water cooling friendly. If you're able to sell that case and use those funds towards a newer case would be a better idea but if you're stuck with it you still can pull it off. You'll just be hanging one externally.

Most 140mm spaces in cases won't fit a 140mm radiator but will fit a fan. Most forget the radiators aren't exactly the same size as a 140mm fan. They are longer and wider. I almost made the mistake of getting a 140mm a while back in my first few water cooling days for my 800D at the time and realized that it wouldn't fit after doing some measurements.

A simple bracket in the back like this would simply do the job well.

 

[jmdixon85]

Well it may be old but I don't think I be changing cases anytime soon. It will fit a 140mm rad in the back, its a rather large case.

But would a 120.3 and a 140mm be enough to cool a CPU plus two GPU's?? Highly overclocked too, I'm not shy of over volting things


If not then it will be another 120.3 hanging off the back of my case :/ it does have holes in the case to allow this

 

[GTXJackBauer]

If you're looking to do some high overclocks on both CPU and GPUs than yes I would advise a 120.3/120.4 off the back but we won't know how much heat will be generated. We're just guessing here. You could go with a 120.4 in the rear and just call it a day. I am sure it should be more than enough say 2x 250w+ GPUs and say a 88w+ CPU at a decently quiet sets of fans if you're also looking to keep it nice and quiet while on full load.

 

[BobbyBubblehead]

Check your pm

 

[jmdixon85]

Got your PM bubble head. Thanks I have a few spares already tho.

Its just the GPU blocks that cost a fortune

I have a extra pump and rad already, looks like if I water cool the GPU's I need to go external Even in the mighty (but old) HAF932..

 

[RnRollie]

You know what has been bothering me for a while?
The "loop order is not important as long as Res is before Pump" statement

I think it is a overgeneralisation and we've been become too nonchalant in throwing it around.

I think the statement should rather be more like:

"loop order is not THAT important as long as Res is before Pump AND ... "

"loop order is not THAT important as long as Res is before Pump, HOWEVER... "

"loop order is not THAT important as long as Res is before Pump, DEPENDING... "



Someone with more money as sense can take the "loop order is not important as long as Res is before Pump" as a given and create following loop


BayRes (monsoon, whatever) > 90° elbow > 10" of 1/4" tube > 90° elbow > EK DCP2.2 > 90° elbow > 5" of 1/4" tube > 90° elbow > HTSF2 > 90° elbow > 10" of 1/4" tube > QDC Bulkhead fitting > 90° elbow > 20" of 1/4" tube > 90° elbow > MO-RA > 90° elbow > 20" of 1/4" tube > QDC Bulkhead fitting > 5" of 1/4" tube > 90° elbow > Tri-SLi or Tri-Xfire KOMODO blocks (series or parallel) > 3" of 1/4" tube > EK Supreme LTX > 90° elbow > 90° elbow > VRM waterblock > 5" of 1/4" tube > 90° elbow > HTSF2 > 90° elbow > 10" of 1/4" tube > 90° elbow > BayRes


I've got my res before the pump, so it should work right?


Well yeah, for about 5 minutes, if you manage to get it filled up to start with.

Then you'ld be facing following problems:
The puny jingway DCP2.2 lacks the head to overcome all the elbows, the restriction of the tubing and the most restrictive Radiators & GPU blocks on market
The Supreme LTX isn't a flow king either but at this point it doesn't matter as there just isn't enough pressure to make the impingement/jet plate design work.
Since BayRes (and most reservoirs) are very rarely filled up to the brim there is no pressure (except some gravity) on the suction end of the pump to help it force the water around. Even if the pump aint cavitating at this point, it will soon die, if you're lucky only after the CPU has done a few thermal shutdowns.


See, the thing is, reality it isn't always as clear cut as the statement suggests... and we shouldn't forget that

 

[GTXJackBauer]

You know what has been bothering me for a while?
The "loop order is not important as long as Res is before Pump" statement

I think it is a overgeneralisation and we've been become too nonchalant in throwing it around.

I think the statement should rather be more like:

"loop order is not THAT important as long as Res is before Pump AND ... "

"loop order is not THAT important as long as Res is before Pump, HOWEVER... "

"loop order is not THAT important as long as Res is before Pump, DEPENDING... "



Someone with more money as sense can take the "loop order is not important as long as Res is before Pump" as a given and create following loop


BayRes (monsoon, whatever) > 90° elbow > 10" of 1/4" tube > 90° elbow > EK DCP2.2 > 90° elbow > 5" of 1/4" tube > 90° elbow > HTSF2 > 90° elbow > 10" of 1/4" tube > QDC Bulkhead fitting > 90° elbow > 20" of 1/4" tube > 90° elbow > MO-RA > 90° elbow > 20" of 1/4" tube > QDC Bulkhead fitting > 5" of 1/4" tube > 90° elbow > Tri-SLi or Tri-Xfire KOMODO blocks (series or parallel) > 3" of 1/4" tube > EK Supreme LTX > 90° elbow > 90° elbow > VRM waterblock > 5" of 1/4" tube > 90° elbow > HTSF2 > 90° elbow > 10" of 1/4" tube > 90° elbow > BayRes


I've got my res before the pump, so it should work right?


Well yeah, for about 5 minutes, if you manage to get it filled up to start with.

Then you'ld be facing following problems:
The puny jingway DCP2.2 lacks the head to overcome all the elbows, the restriction of the tubing and the most restrictive Radiators & GPU blocks on market
The Supreme LTX isn't a flow king either but at this point it doesn't matter as there just isn't enough pressure to make the impingement/jet plate design work.
Since BayRes (and most reservoirs) are very rarely filled up to the brim there is no pressure (except some gravity) on the suction end of the pump to help it force the water around. Even if the pump aint cavitating at this point, it will soon die, if you're lucky only after the CPU has done a few thermal shutdowns.


See, the thing is, reality it isn't always as clear cut as the statement suggests... and we shouldn't forget that

Talk about going overboard. MAN OVERBOARD!!!!!!!!!

 

[BobbyBubblehead]

All depends if people ask or buy first really. Not like anyone would recommend half a dozen obstructions and say use a low end pump so we can laugh about it.

Sufficient flow rate and we proved here years back it was negligible difference if detectable at all.

Still appreciate thoughtful replies though

 

[GTXJackBauer]

All depends if people ask or buy first really. Not like anyone would recommend half a dozen obstructions and say use a low end pump so we can laugh about it.

 

[outhouse]

You know what has been bothering me for a while?
The "loop order is not important as long as Res is before Pump" statement

Ya me too


It is a loop, the res will always be before the pump no matter where it is in the loop.


I fill my loop before powering on, so it really does not matter.

 

[Daddyjaxx]

So the res should always be before the pump.....

 

[RnRollie]

weeeeeelllll, ... yes, although it technically always is in a loop....

maybe we should start rephrasing it in the sense that "the Res should ALWAYS be DIRECTLY before the pump from a loop/flow direction point of view , although a (short) piece of hose between res & pump is allowed. One could refer to it as "a linked Res-Pump combo, where the res must feed the pump, and thus the pump must be lower in the loop as the res" "

I hope someone can come up with less words as above for that Res-Pump-Combo

But while we're at it.. there is of course one exception of a thing which is allowed between Res & Pump ..... a T-Line

(yeah i know it would defy purpose)

 

[EarthDog]

Ya me too


It is a loop, the res will always be before the pump no matter where it is in the loop.


I fill my loop before powering on, so it really does not matter.

LOL, if people always did that, there would be no need for that statement. People will think of placing it anywhere...

Part of the reason for doing that is to be able to feed the pump easily upon start/filling. A dry pump is a dead pump and all.

 

[Supertrucker]

weeeeeelllll, ... yes, although it technically always is in a loop....

maybe we should start rephrasing it in the sense that "the Res should ALWAYS be DIRECTLY before the pump from a loop/flow direction point of view , although a (short) piece of hose between res & pump is allowed. One could refer to it as "a linked Res-Pump combo, where the res must feed the pump, and thus the pump must be lower in the loop as the res" "

I hope someone can come up with less words as above for that Res-Pump-Combo

But while we're at it.. there is of course one exception of a thing which is allowed between Res & Pump ..... a T-Line

(yeah i know it would defy purpose)

Make sure the res can gravity feed the pump because as ED and so many others have said, a dry pump is a dead pump.

 

[outhouse]

LOL, if people always did that, there would be no need for that statement. People will think of placing it anywhere...

Part of the reason for doing that is to be able to feed the pump easily upon start/filling. A dry pump is a dead pump and all.

It sure would have mad efilling the loop easy.

I always do things the hard way


I just didnt have room amywhere else, and did not want a fill port above my graphics cards.

Guess I could have bought more fittings and ran a fill line

 

[gkline]

jmdixon85: I will try to answer your questions. I have the exact same HAF 932 case as you. I have the Advanced case (actually have a second one for my 3770k rig also).

My rig has a 3930k cpu and an Asus Sabertooth X79 mb. When I started in custom watercooling I had a XSPC RX360 rad and a XSPC Acrylic dual bay D5 res/pump. I was running 2 GTX670 FTWs in SLI and had 2 680 waterblocks (different makes). I decided the 360 rad wasn't enough and first added a 120 rad internally. It was TOO much work for not much gain. I then replaced the 120 internal rad with an externally mounted (vertically in the rear) XSPC EX360 which worked well. XSPC makes external mounts that are not real expensive (bought at Frozencpu.com)

http://www.frozencpu.com/products/1...Bracket_Set_-_632_Thread.html?tl=g30c637s1731

Since then I replaced the 2 670s with a single watercooled EVGA GTX780 Classified and quite frankly went overboard by replacing the EX360 with an external Watercool MO RA3-420 Pro (a beast). I got an incredible deal on a brand new one which included the pedestal feet and fan enclosure. Since I have so much rad capacity I switched out the single D5 bay res for a dual D5 bay res (XSPC Twin D5 bay res) with the pumps running in series. Coupling this with my internal Rx360, I could handle both the cpu and 3 or 4 gpus with excellent temps.

Getting back to your setup, your 2500k is ramped up at 4.7Ghz and throws off heat but not as much as my 3930k @4.6Ghz. Your big heat dump into the loop will be your dual GTX470s. I don't know where you will buy waterblocks since these gpus are becoming dated but it will probably cost at least $150 for both blocks.

My dual 360 setup was very good with both of my 670s even overclocked so I would hang a 360 rad or even a 480(if it will fit) to the rear. Going much smaller is not worth the effort.

 

[gkline]

Here's a photo of my HAF 932 with the external MO RA3 - sorry for the poor photo quality.