Dual Loop versus Single, the facts

[NaeKuh]

thats not a dual loop barf... lol..

if 1 pumps fails... you still got water though all your loops.

dual loop means if the pump fails, the loop fails, that loop is knocked out.


Your sharing a res, which means equalibirum will be atttained on both sides sharing that res..
its a single loop unless you had separate res's.

Example...
I am running 3 independant loops on 4 radiators and 6 pumps.
That means each loop has a dual pump option, its own res, and radiator. Completely separate from each other, doesnt mix at all.

While i understand gabes' theory in this approach, single loop is ideal for starters.
This methodology also assumes your NEVER going to get both cpu and gpu under constant load for long term which is true 85% of the time.
(15% is if you do distributing computing... or Folding / WCG ) Then good luck with a single loop.

While u get advance in our hobby a dual loop is the route, because of the more fine control you get, and when setup properly, you get better efficiency in your loops for lower deltas.

Do you think i should do this? or should i stay dual loop with shared res?

whatever will make you the shortest loop.
This will always apply... the shortest loop always = WIN.

[gabe]

thats not a dual loop barf... lol..

While u get advance in our hobby a dual loop is the route, because of the more fine control you get, and when setup properly, you get better efficiency in your loops for lower deltas.

1/ Pls qualify/elaborate (with data if possible) "setup properly"

2/ By "While u get advance in our hobby" do you mean something like your own setup (3 independant loops on 4 radiators and 6 pumps)?

3/ Other members comments/input request: is Naekuh's setup representative of the average XS member?

4/ Have you tried to connect your 3 loops in series? and if so, would you share the data?

[barfastic]

Its not exactly a single loop though either. If 1 set of pumps fails, i wont have flow to that part of the loop.

While u get advance in our hobby

I think of myself as advanced, just because i dont have your setup (Then again i dont ahve a MM Acension + Pedestal to fit 6 rads into anyways, and i dont want one either.... i have enough bar stools/couch tables), doesnt make me less advanced.

This is my system.

not sure if you can make it out, but the left side of the bench, is the CPU only loop, and the right side is for everything else.

It has dual d5's on each "loop". if both pumps on one side die, that loop (either CPU, or GPU/ram/mobo) wont get any flow... I have the ability to turn on/off the pumps, and ive tested to see (with a flowmeter) if one side shuts down if it will get any flow.... it doesnt. Can you please explain to me where you assumed that it would get flow?

[NaeKuh]

1/ Pls qualify/elaborate (with data if possible) "setup properly"

2/ By "While u get advance in our hobby" do you mean something like your own setup (3 independant loops on 4 radiators and 6 pumps)?

3/ Other members comments/input request: is Naekuh's setup representative of the average XS member?

4/ Have you tried to connect your 3 loops in series? and if so, would you share the data?

1. because your still dealing with carrying capacity of water, you will get a higher delta from inlet and outlet in your system.
Gabe, if you have 3 monster class GPU's + an assortment of blocks where it crimples your flow, you cant tell me by not separating them out your not going to get better performance.
The lowest you can go is ambient... the highest you can go will be determined by how much flow you have in relationship to how many blocks are connected in a series with the overall radiator capacity.

2. By advances i mean as you get more skilled and you want customization. Are you telling us ONLY to use single loops and conform to the mac way? im sorry, but watercooling is about customization, how you build it and how you use it is totally up to the op. Telling a person the single loop is always the best and only way is blatantly wrong.

3. My system is no where near the average, its what happens after you stuck with this hobby for a very long time with a lot of resources outside of being a vendor. I can count more then a handful of systems in the worklog which would rival mine and possibly even beat it.

4. Yes i have... with everything loaded it was terrible. Were talking about 3 x 580GTX + full board block (which is given to kill flow) + cpu block. I have TOO MUCH stuff on 1 loop, if i add too many DDC's in a series my flow only goes up 1/3 for each DDC i added, and they get WORKED harder and heat up faster. Gabe have u ever attached 3 ddc's + in a series and seen how hot they run? ignoring the 35x...


Now i said single loops are ideal for the starter, who doesnt need to worry about the cache of blocks.
But not everyone is solely on just your blocks.
We use board blocks, mosfet blocks, some guys even use ram blocks.. EACH of these require flow.
To a certain point you can only get so much because water can only carry so much.

Now dont get me wrong, if your happy with your 10+ Deltas dont let me stop you... but i didnt water for that. And i sure many others will follow my shoes and strive for lower temps over bling.
If your a starter, a 10C delta might seem wow, as you get advance its dismal... you get greedy like all human nature and you strive for the less then 5C delta, even the 2C delta.
Which isnt possible in your model with EVERYTHING LOADED on the cache of blocks i have, without damaging my pumps in the future from overworkage by having too many in a series.

Its not exactly a single loop though either. If 1 set of pumps fails, i wont have flow to that part of the loop.

I think of myself as advanced, just because i dont have your setup (Then again i dont ahve a MM Acension + Pedestal to fit 6 rads into anyways, and i dont want one either.... i have enough bar stools/couch tables), doesnt make me less advanced.

This is my system.

not sure if you can make it out, but the left side of the bench, is the CPU only loop, and the right side is for everything else.

It has dual d5's on each "loop". if both pumps on one side die, that loop (either CPU, or GPU/ram/mobo) wont get any flow... I have the ability to turn on/off the pumps, and ive tested to see (with a flowmeter) if one side shuts down if it will get any flow.... it doesnt. Can you please explain to me where you assumed that it would get flow?

Im sorry i should of clarified the terminology advance...
It means as you know what each thing does in relationship to how it works, as your customizing, you plan things out accordingly so it works within the set parameters you want it to work at.
This means sometimes you will have to go outside the normal way things are done, and improvise, or innovate new ways to get the job done.

When i say advance.. it typically means, when you planned your loop, you didnt ask for help, you didnt use any resources because your brain is already a mountain of resources.

And Barf, a single loop is always a single loop if 1 component on each loop is shared regardless of how its split up.
A closed loop works on equalibirum, meaning it tries to hold constant until you exceed the maximum holding potential of water where the water temp slowly rises in a gradient form inside your loop.
Your sharing the res, that means as your pulling water from the res, your pumping it back, you will get some movement inside that second portion of the connection inside your res.. it maybe small, but its still a off parallel loop model because your sharing the res. Flow is just very dismal inside that second portion of your parallel because of the imbalance in restriction, which your second pump helped negotiate.

We never had this problem b4, GPU's didnt put out much heat (No one really SLI'd and when you sli'd it ended at 2 and Xfire was even rarer) and cpu's didnt, unless u were on a kentsfield under constant load all the time... there was no board blocks til Bitspower, Bei and I started innovating, so there was no true flow killers out there (Except the EK supreme).
Gpu blocks were just solid slabs of copper with channels (not micro channels which are more efficient) and now have they have water channels on the vrm's so they pick up even more heat.

We didnt have these options, but we do now... things add up, nothing in the world is for free, and everything comes at a cost unless you pay for the cost. <-- this is the rule of thermo.

[vivithemage]

I do like the redundancy of the single loop, dual pump.

[MrBean]

3/ Other members comments/input request: is Naekuh's setup representative of the average XS member?

Nope, and no disrespect here to Naekuh, and others, that is serious overkill, and a waste of money.

I run a single-loop, dual-quad Xeons, as well as an ATi 5990, and on a single Feser 240, with fans running 1200-1300rpm, and yes, even in Melbourne summer temps, I can keep it running full-load for days on end, and yet have to have a BSOD or any issues.

I guess, Gabe, the biggest problem is perceptions on what modern electronics can handle, what are acceptable temps vs what we all think it should run at.

95% of Watercoolers nowadays go wayyyy overkill, simply because of the availability of components.

Oh yeah, and in some cases, epeen

But, in all honesty, a $2000 watercooled setup won't give much overclocking headroom over say, a well-planned and implemented $400 setup. Been there, done that, and as I have stated time and time again, if you want to unlock the overclocking potential of your system, you simply HAVE TO go phase-change.

I have been in the Futuremark Hall of Fame Top 10 enough times to know that, and to know that for a very reliable, 24/7 watercooling system, simpler is mostly always more reliable.

I do run dual DDC's on my Blackbird system though, but simply for redundancy, as this PC is a doing services as a process simulator, and I would like to keep availability as high as possible.

My simple point is that we're spending way to much money on watercooling due to the 'expert' advise often dished out, with no scientific substance to back it up.

In our Industrial control platforms, and Scanning platforms, consisting of very delicate electronics, traversing a web of paper, running at 1000+ m/m, with ambients of 48 deg C +, we watercool these very sensitive beta-radiation weight, IR moisture, Optical color, and laser-caliper measurement sensors, and the best part is, we actually use heaters to stabilise the water-temp at 35 deg C, on the other end, we have a heat-exchanger interfaced to tap-water, at around 20 deg C.

What makes water better is the fact that it can absorb so much heat, and act as a damper for those sharp transient temps you find with aircooling, at an acceptable equilibrium temp - and this is the point of contention in threads like these - what is acceptable = well, much higher than what most think.

Anyway, just my opinion, and no offense to anyone - if you like the way you do it, no worries, then do it - with watercooling, and the latent heat capacity of water, you'd have to be really silly to end up with a system doing worse than standard aircooling

[rge]

Nope, and no disrespect here to Naekuh, and others, that is serious overkill, and a waste of money.

hmm....hobby....waste of money....nagging......wait not wife....

I guess, Gabe, the biggest problem is perceptions on what modern electronics can handle, what are acceptable temps vs what we all think it should run at.
95% of Watercoolers nowadays go wayyyy overkill, simply because of the availability of components.

thank god you took the time and explained to us 95% WHY we do it, I always wanted to know why I bought those 2 rads that I clearly did not need, to cool the components that I did not need to upgrade, in the new build I did not need to buy, in the case I did not need to have painted.

My simple point is that we're spending way to much money on watercooling due to the 'expert' advise often dished out, with no scientific substance to back it up.

Wait, your confusing me. I thought it was because I did not understand what modern electronics can handle and because there were components everywhere I looked...now your telling me I have been brainwashed as well....oh god no....no....nnneeed tooo find pssychiatrsstst nowww.

[NKrader]

3/ Other members comments/input request: is Naekuh's setup representative of the average XS member?

oh yeah

[shazza]

An interesting thread - but please don't let this get into any personal flame wars. It's fine to disagree, and state your opinions. Let's just not make personal attacks.

[MrBean]

Thanx Shazza, appreciated, as pretty often, keyboard-heroes need to revert to the facetious.

Or, it is probably the case that certain members look at one's postcount and decide if they're dealing with noob, or expert

Wait, your confusing me. I thought it was because I did not understand what modern electronics can handle and because there were components everywhere I looked...now your telling me I have been brainwashed as well....oh god no....no....nnneeed tooo find pssychiatrsstst nowww.

Kindly point where I have at any point in time, quoted you, or referred to you in my previous response. Clearly you were confused long before I posted my comment. Chill out, this is a discussion, not a personal attack.

[NaeKuh]

u dont need to watercool.

have you seen a 2600k on air not clock as well as a 2600k on water?

its all personal preference... so whats the real point in water?

Its about art... and how you want your system to be.


If you want to build your loop one way, go ahead, i wont stop you. Infact i applaud you for attempting water.
But you should never be forced to build only 1 way ever. Not in something which id driven by customization and fun.

That kills innovation...


Anyhow im not arguing with the DATA, i said single loops is ideal for most users.
But its not how i or many others would build a system of my nature or the many other XS systems which lurk the worklog as monsters.
As i said, there are more then a handful of systems in the worklog sections which would easily rival mine / ascend past mine...

[MrBean]

its all personal preference... so whats the real point in water?

For me it's the stability associated with water vs air. As mentioned in my earlier response, water has a high heat-absorption capacity, and this tends to assist dampening down massive temp-transients, leading to improved system stability. But then again, you know that already.

I guess the only point I was trying to make earlier was your need/requirements vs real requirements.

As an example, my Wife's setup is running a Feser 480, dual D5's, single-loop. My own Blackbird, similar specced hardware, make do with a 240mm rad. In the Wife's PC case, I had the space for a 480 rad, I had the rad, and it was a practical solution - was such a big rad needed for her setup - simply, no - I can use the Innovatek to stop 3 of the 4 funs, run the last at at 1400rpm, and game for hours, without any concern.

But, the rad goes well with the big Lian-Li case, simplified the mod, and looks good.

Luv it - but, as I also mentioned earlier, if you don't have the space for a 480mm rad, a decent 240mm would cool most systems out there, considering practical application. Of course, my input revolves around real-life application, empirical data, vs what a mod may require to look good, or what you want your system to be because you simply have the money to blow, and want to do something outrageous - no problem with that, we all do it.

Is it needed though, and should it be prophesized as such, to allow your system to operate reliable at full, and most importantly, stable potential - a resounding no

@ Naekuh: As always, I appreciate debating with you, and respect your opinion - I don't always agree with it though, but, that's why we are individuals. Would luv to see some pics of yer setup too, must have been tough to get that all routed proper.

But you should never be forced to build only 1 way ever. Not in something which id driven by customization and fun.

I cannot agree more with this statement

[Vicey]

Personally I always run a single loop but I try to separate my radiators between components. For example right now I have a 140mm radiator after my CPU but before my graphics as an exhaust. It fits nice and neat uses no more tubing than going straight to the GPU from the CPU would. I also have a 1080 Radiator straight after my graphics cards and finally a 280mm radiator just before my CPU but after my pump. I didn't plan it this way originally it just sort of happened to be the neatest way to get it. The 1080 radiator is external connected with Koolance quick disconnects but the system runs perfectly fine on just the internal 280mm + 140mm radiators that are inside the case (just a little hotter and louder). I love those Koolance disconnects because I can connect them together without the 1080 rad and run the system all internal. Perfect if I wanna take it to a friends house for a lanparty, one extra large piece I don't need to carry.

[Church]

MrBean: are very small rads often sufficient for stable usage of particular hardware set on generic usage, not folding/crunching? - indeed, often yes.
BUT! 1) - is it more effective then air? I see no sense in spending lot on cooling solution with efficiency less then current top air coolers, and in such case for smart investment good air cooling imho would be way to go. If one turned to LC because of wish for better then air, one shouldn't limit to less heat dissipating area then even top air coolers would give, or as result user has spent hell of a lot more and got something worse cooling efficiency/noise wise.
2) reduced maximum temps (86°C? imho not that hard to reach with high load and high deltas of insufficient rad area) for new LGA2011 cpu-s alarms me a bit, if that's not some trend to worsen even more for future top cpu-s, and thus something one has to take into account when building cooling system for future hardware upgrades aswell :/. Of course, these cpu-s may run at insane clocks without crashing, but what such high clocks can give in reality, if cpu-s start throttling half the time with resulted performance like that of cpu @clocks much lower?

[basserdan]

I'm concerned it won't sound nearly as cool.

Example; ...me, I'm on water.
Huh?
Oh, just that I water cool my pc!!!
Really?
Yeah, see here, it's a dual loop configuration. I've taken taken the two major components that need cooling and broken them down into two primary, separate subsets of cooling systems - processor on one and graphics on the other. Procs are far more dependent on or, sensitive to, temps for oc'ing so they get more radiator than needed to over cool them in order to maintain a very narrow temp delta. Graphics tend to peak their oc for reasons that occur before cooling becomes a limitation so...

VS.

See, the water goes from here to here, to there, over here, through there and right back to here again. Neat, huh?
Um, yes.

Now, as you can clearly see, the second guy has a decent setup but, that first guy, he had a real serious system! It was designed to provide a specific application to a particular area, to squeeze every drop of performance... or something. He takes it to a higher level, and so on and so forth. He's probably at least, a tad more expert than and could even be slightly cooler than that other guy, probably.

Once they find out the 2nd guy is running his single loop to a drip-irrigation hose in a planter box in the window sill well, it's all over. The extreme pc honeys, the computer groupies, will not be stickin' around for the all-night WoW-a-thon!

It's really a matter of who you want to be: sophisticated, cool Vs. practical, lonely. I ain't gonna risk it!


I am a two looper though, mainly because of the desire to have a dedicated solution to keep cpu temps in check for oc. I have considered the likely pointlessness in this just because of what Gabe revealed. For me, the two hottest subsystems; cpu and gpu are really, never being taxed simultaineously and would probably amount to just more cooling surface on either system when it alone, was getting a workout. The redundancy of a single loop with dual pumps provides such a logical solution to a personal concern of mine, pump failure, it must be considered again. Of course, I can go a long time w/o opening my box, one pump could dead be and I still might not know that my system again, hangs on by just one pump.

Plus, I just like the way it looks having a dedicated system to and from an individual source with neatly routed tubing, all tying in to their respective ends.

[MrBean]

are very small rads often sufficient for stable usage of particular hardware set on generic usage, not folding/crunching? - indeed, often yes.

Good points. For the record though, I (very stupidly) had my system 10 hours running OCCT, with dual instances of Furmark.

Ambient was around 32 deg C, water temp stabilized at 37 or 38, GPUs maxed out at 42, and CPUs at 50-52 deg C. The problem with ever shrinking dies are not really raddage et al, rather getting the heat out of the CPU.

Last comment from my side, a Feser 240mm rad has a lot of surface area, and tied with the heat-capacity of water, well.....you would not get aircooling on both my quad Xeons, as well as the 5990, to get temps this low and stable, not with 30 deg C plus ambients. I invite anyone on here to prove me wrong

Have a good one guys, talk later again.

Procs are far more dependent on or, sensitive to, temps for oc'ing so they get more radiator than needed to over cool them in order to maintain a very narrow temp delta.

I will respectfully disagree in parts with this statement. Even 10-20 deg C real die-temp decrease will not yield significant overclocking gains - to really push your system, unfortunately, you will need to drop die temp to way low negative temps, like -20 and lower, much lower. So, the difference to overclocking potential between a Feser 480 and 3x Feser 480's will be, well, I wonder if you will see 1-2%.

I am serious overclocker - if I cannot gain 20-30% +, why waste time I simply will not spent 4x the money on extreme WC to gain 1-2% over what I have now - in that case, I'll rather buy a more powerful gfx card...

Again, simply my opinion, based on my experience stemming from watercooling/extreme cooling since 1997, but as stated earlier, whatever works for you....

[Church]

My thoughts:
to get smartly relatively cheap budget with better noise profile cooling: Air, and only air. Top cooler like thermalright archon or silver arrow/phanteks ph-tc14pe/zalman cnps12x and likes + something of 3rd party air cooling for gpu-s with big rads nad more & bigger fans on them (i'm not competent which are best here). +Maybe few extra case fans to enhance overall airflow. Nothing will beat air on relative simplicity/reliability and by price/performance.
next stage: still top air on cpu, but contrary to what many do (LC only for cpu), small-rad LC for gpu only. As TDP of top gpu-s are more then that for cpu-s, but their cooling solutions have to fit in much more constrained limits in 1-3 slots, thus gpu-s should be prime candidates for LC instead of CPU-s to get noise and temps down. +Higher overclocks of gpu will net more FPS then higher overclocks of cpu, which probably matters more for majority of us consisting from gamers instead of crunchers/folders. Even single 120mm rad section per top gpu with even high deltas will bring temps and noise probably on par or better then top 3rd party air cooling solutions, and much much more so then stock cooling.
next stage: LC all major heat dissipating components in single loop. Here i take into account original findings of Gabe of pros and cons of single vs dual loop. Probably one will have to add relatively more rads though, as acceptable high deltas for gpu-s may hurt CPU temps/overclocks, but still overall single loop wins. I also did mention "major heat dissipating" not "major" components, as cooling 7W chipset/RAM/or HDD with LC is everything but smart gains/cost wise. Yes, many do (mostly for bling so), but for smart spending those should be left out.
instead of next stage (with big pile of rads of whatever configuration (two MoRa3-s, 3 quads or whatever ), LC everything, uber controlling solutions) i'd probably would start to consider single phase cooling instead. Costs get similar, but cooling not. Getting down further below 5C delta will cost same, but will net worse temps then extreme cooling, so why go for it? So while extremities like NaeKuh's build is interesting experiment, it simply is unwise as to what you get relative to investments, as there are better alternatives available.
next stage .. even more extreme coolings? Heck, no. NOT for everyone. NOT wise in any reasonable way to actually using PC to do some job. Clustering/more PCs will use less electricity/compute better then these, made for select uber-minority for whom PC is only for momentary max overclock, not actual use. So imho it should be left out of scope for 99.9% of majority of users.

[rge]

Good points. For the record though, I (very stupidly) had my system 10 hours running OCCT, with dual instances of Furmark.

Ambient was around 32 deg C, water temp stabilized at 37 or 38, GPUs maxed out at 42, and CPUs at 50-52 deg C.

A delta air to water of 5-6C on a 240 rad with 1300rpms fans with furmark on gpu and 2 quad xeons loaded with OCCT is impressive. Your 240 rad with 1300 rpm fans is cooling your dissipated power (whatever that is) better than my 360rad + 280rad with 1800 rpm fans in push/pull with just my 1 cpu loaded with OCCT and 1 GPU loaded with furmark. My delta air to water is 9.2C. It is 5.0 C with just cpu loaded with OCCT and gpu idle. (measured with calibrated dallas 1 wire/crystalfonz). My dissipated power is cpu 185W (voltage calc by DES on GB), loaded GPU + idle gpu ~250W + 2 pumps 35W). Cpu only delta includes both gpus at idle heat.

Can you list what quad xeons you have and any overclock settings, ie ghz/speed. And list same for gpu. And whether your rads are external or internal, (internal rads lose about 20+% efficiency from air resistance, compared to dropping side panels of a case eliminating resistance of pushing/pulling air into case.) That way I can estimate dissipated power to compare to mine.

If everything is stock with very low heat output, then your unusually low dissipated power/stock settings are as different from norm on this forum as your cooling.

[MrBean]

See my project-log in my signature, the Blackbird one.

More specifically, refer to post 60 at the bottom of this page, made nearly 2.5 years ago - ambient was a little lower then, but you get the idea -> http://forums.overclockers.com.au/sh...=717445&page=4

Take note, back then, I had the same basic system, but with a 4870X2 - nowadays it's a 5990, with the same result, no real increase in watertemps considering the same ambient. Should you wish to dispute these facts, I have no issue whatsoever doing a few screendumps over the weekend with said system, just to satisfy you.

Then, please do me a favor and reread my post just prior to your comment, about modern processors, ever shrinking dies, and the problems getting the heat out of the die. Your problem is not getting the heat out of the water, rather getting it into the water to begin with - yes, Houston, we have a surface-area problem, but, it ain't with the rad.

Current waterblock-technology is about as advanced as you can make it with current materials, and as efficient, let's not split ball-hairs over 2-3 deg C, because that means nothing other than bragging rights.

We need something way more advanced to get the heat out of the CPU-die, maybe in a few decades, there will be nano-fans embedded inside the processor-substrate to assist with heat transport.

But, that's a topic for another time. Someone on this board may have the answer to the next waterblock revolution, he mentioned we should think about submarine propulsion - all I can think of is some form of creating a low-pressure area inside your waterblock, allowing water to boil at much lower temps, of course this phase-change will require a lot of energy, taking it from the surrounding environment - ditto, pull the heat a bit quicker, and more efficiently, from the CPU die.

Mmm...probably the rantings of a madman

Anyway, signing off, need to get some sleep - pretty late Down Under, so see you guys later.

[basserdan]

I will respectfully disagree in parts with this statement. Even 10-20 deg C real die-temp decrease will not yield significant overclocking gains - to really push your system, unfortunately, you will need to drop die temp to way low negative temps, like -20 and lower, much lower. So, the difference to overclocking potential between a Feser 480 and 3x Feser 480's will be, well, I wonder if you will see 1-2%.

I am serious overclocker - if I cannot gain 20-30% +, why waste time I simply will not spent 4x the money on extreme WC to gain 1-2% over what I have now - in that case, I'll rather buy a more powerful gfx card...

Again, simply my opinion, based on my experience stemming from watercooling/extreme cooling since 1997, but as stated earlier, whatever works for you....

I will definitely defer to your experience and I agree with what you said.

I was imagining, as I wrote that part, that it might create some disagreements. I thought I might have suggested that water cooling had more power over overclocking than I meant to. It won't provide the awesome temp. drop needed for cooling to affect signifigant oc's. I have found a number of times there's an extra 200mhz hanging around that will not pass a temp point stabily. Drop that 200mhz and you run right through that temp zone w/ stabilty. Keep it below that temp and you can keep the extra mhz. You might find you can do this easily and very quietly, with a 360 rad where even the best air coolers will not. I probably couldn't have made my best oc with my current proc if it shared water with my 4870x2 stinkin' up the place.

The difference I meant to point out is when I've clocked my graphics I've found a larger tolerance for heat. That is to say, on a number of cards, I've found that you can only clock it so far regardless of heat. You will a max oc and it will operate at a fairly high temp, often the temp it's rated for. But, tweak just a little more and it won't run at all, even while it's at cooler temps. In this case the graphics card isn't maxing out from a decreasing acceptable max temps but rather it's getting tripped up on too much voltage or clock speed etc. For this fairly consistent, reason I have found most of my graphics to not be as temperamental over heat for the max oc.

[Church]

basserdan: imho LC can be easily thought as air cooling with limits removed. You can easily increase heat dissipating area as in rads with no limits to their size/mass and distance from cooled area unlike limited by those accounts classic air cooling. Hence in rough picture better but similar not xtreme performance, and gets more used not for uber xtreme overclocks at subambient temps cooling, but rather to get quieter cooling solution at still reasonable price and for 24x7 usage.

[rge]

MrBean, you got a 6C air to water delta with 1300 rpms fans on 240 rad with OCCT on 2 Xeon 5430 (TDP 80W each) so ~50W each OCCT load at stock settings or 100W both + dual GPU 5990 furmark which is an OCed 5970, from reviews similar power consumption to my gxt295 dual gpu.

I got 12.4 C air to water delta with 1300 rpm fan speeds push/pull on 360 rad (plus passive cooling 280 rad) with OCCT on my 1 CPU stock MHZ, vcore set to 1.15v + GPU furmark load (both gpus). My cpu temps were only 61C max since at stock.

I got 17.2C air to water delta same except my cpu/gpu at 24/7 settings OC settings. My cpu temps maxed at 87C.

I dont really care about the 6C discrepancy from error/different power/full cover or not.

But, by using my rad preference (360+240/280) instead of yours (240), just running prime on cpu I can shave 5C off my temps from say near 80 to mid/low 70's by ~halving my delta air to water to current 5C. You are making no greater gain on your liquid cooled cpu at 50W power with your liquid cooling vs high end air. Not to mention, if I am already forking out $$$$ for liquid cooling, gaining an extra 5C running prime for cost of another rad is IMO good bang for the buck given $$$$ paid for best cpu block/tim/fans, etc.

So I dont understand the soap box against 5C from extra rad (or even 3-4C for external rads) going from say 80C to 75C (given one already paid $$$ for water), when you are watercooling your cpu going from ~60C air to 50's C with your liquid.

As for your statement, biggest issue is gradient from core to water, I agree and have made the exact same argument, hence didnt argue with that. But until I can go after the biggest fish, will have to settle for smaller ones.

[stephenswiftech]

1. because your still dealing with carrying capacity of water, you will get a higher delta from inlet and outlet in your system.
Gabe, if you have 3 monster class GPU's + an assortment of blocks where it crimples your flow, you cant tell me by not separating them out your not going to get better performance.
The lowest you can go is ambient... the highest you can go will be determined by how much flow you have in relationship to how many blocks are connected in a series with the overall radiator capacity.

2. By advances i mean as you get more skilled and you want customization. Are you telling us ONLY to use single loops and conform to the mac way? im sorry, but watercooling is about customization, how you build it and how you use it is totally up to the op. Telling a person the single loop is always the best and only way is blatantly wrong.

3. My system is no where near the average, its what happens after you stuck with this hobby for a very long time with a lot of resources outside of being a vendor. I can count more then a handful of systems in the worklog which would rival mine and possibly even beat it.

4. Yes i have... with everything loaded it was terrible. Were talking about 3 x 580GTX + full board block (which is given to kill flow) + cpu block. I have TOO MUCH stuff on 1 loop, if i add too many DDC's in a series my flow only goes up 1/3 for each DDC i added, and they get WORKED harder and heat up faster. Gabe have u ever attached 3 ddc's + in a series and seen how hot they run? ignoring the 35x...


Now i said single loops are ideal for the starter, who doesnt need to worry about the cache of blocks.
But not everyone is solely on just your blocks.
We use board blocks, mosfet blocks, some guys even use ram blocks.. EACH of these require flow.
To a certain point you can only get so much because water can only carry so much.

Now dont get me wrong, if your happy with your 10+ Deltas dont let me stop you... but i didnt water for that. And i sure many others will follow my shoes and strive for lower temps over bling.
If your a starter, a 10C delta might seem wow, as you get advance its dismal... you get greedy like all human nature and you strive for the less then 5C delta, even the 2C delta.
Which isnt possible in your model with EVERYTHING LOADED on the cache of blocks i have, without damaging my pumps in the future from overworkage by having too many in a series.



Im sorry i should of clarified the terminology advance...
It means as you know what each thing does in relationship to how it works, as your customizing, you plan things out accordingly so it works within the set parameters you want it to work at.
This means sometimes you will have to go outside the normal way things are done, and improvise, or innovate new ways to get the job done.

When i say advance.. it typically means, when you planned your loop, you didnt ask for help, you didnt use any resources because your brain is already a mountain of resources.

And Barf, a single loop is always a single loop if 1 component on each loop is shared regardless of how its split up.
A closed loop works on equalibirum, meaning it tries to hold constant until you exceed the maximum holding potential of water where the water temp slowly rises in a gradient form inside your loop.
Your sharing the res, that means as your pulling water from the res, your pumping it back, you will get some movement inside that second portion of the connection inside your res.. it maybe small, but its still a off parallel loop model because your sharing the res. Flow is just very dismal inside that second portion of your parallel because of the imbalance in restriction, which your second pump helped negotiate.

We never had this problem b4, GPU's didnt put out much heat (No one really SLI'd and when you sli'd it ended at 2 and Xfire was even rarer) and cpu's didnt, unless u were on a kentsfield under constant load all the time... there was no board blocks til Bitspower, Bei and I started innovating, so there was no true flow killers out there (Except the EK supreme).
Gpu blocks were just solid slabs of copper with channels (not micro channels which are more efficient) and now have they have water channels on the vrm's so they pick up even more heat.

We didnt have these options, but we do now... things add up, nothing in the world is for free, and everything comes at a cost unless you pay for the cost. <-- this is the rule of thermo.

Water doesn't loose its physical properties with heat load. Heat load tends to increase water temperature but if you have a system that can maintain your water temperature at a decent level, it doesn't matter how much heat load is going through, water still has the same properties.

I've asked you several times to explain precisely what you mean with "holding capacity of water" because it does look like you are mixing things up.

[MrBean]

You can easily increase heat dissipating area as in rads with no limits to their size/mass and distance from cooled area unlike limited by those accounts classic air cooling.

I guess the problem I have with people comparing the real surface area of aircooling with that of watercooling heat-exchangers is a simple one - there can be no direct comparison, as water has that much higher efficiency in absorbing/carrying heat than air does. So, in practice, you will need a significantly smaller area, if airflow remains the same, to keep the CPU/GPU at the same temperature level.

Keep it below that temp and you can keep the extra mhz.

Yep, seen that too, even goes the same for phase-cooling, where you will eventually hit the ceiling anyway. I was only pointing out that with average WC, compared to extremely good, high-end and very expensive WC, there are only very small gains to be had -> to realise the CPU's max potential, we have to revert to drastic measures, which is economically non-viable, as well as not to practical, as 24/7 under sub-dewpoint temps brings its own issues.

Nice debate though, appreciate your input and feedback.

Of course, sometimes big rads are just better to use, becuase it suits a particular mod, or you want to run your system in near-passive, absolute silent mode, I can see the point - but, like my Blackbird case, you're forced to a maximum of 240mm, maybe, a 280mm rad, and it's good that it still does the job, at moderate sound-levels.

I dont really care about the 6C discrepancy (plus im using 360 rad) from error/different power/full cover or not.

One thing I should point though, this Blackbird case is a real heavy aluminium monster, with massively thick panels, acting as heatsinks. I have specifically focussed my roject to have the rad transfer heat into the case itself, which I suspect contributes to my low deltas. To put it in perspective, this system, must weigh in close to the 30-35kg range, so there is a huge metal structure assisting with cooling, as there is direct metal-metal contact with the rad.

I refer to this statement:

Water doesn't loose its physical properties with heat load. Heat load tends to increase water temperature but if you have a system that can maintain your water temperature at a decent level, it doesn't matter how much heat load is going through, water still has the same properties.

The point I tried to raise earlier was the heatload your system will be comfortable and stable with at 24/7 operation is at much higher levels than what we think it is I have also stated whatever works for your application, is what you must use - if you need absolutely the last drop of OC capability from your system, in WC format, so be it, then go for it, no prob with that. Then you fall outside the 95% group I mentioned earlier - most people will not want to OC to the brink of instability though, hence my comment of earlier.

I've been there, done that, it's captivating, fun, and awakens that competitive tiger slumbering in most of us.

As a final word, I just got to the point 5 years ago, after spending a huge number of hours overclocking, tweaking, voltmodding, running bench after bench, that I realised no form of WC is good enough to bring you even close to the top 10 in FM Top10, but, it is abolsutely the best for quiet, 24/7 operation, at least for my application and requirements.

Talk later.

[MrBean]

But, by using my rad preference (360+240/280) instead of yours (240), just running prime on cpu I can shave 5C off my temps from say near 80 to mid/low 70's by ~halving my delta air to water to current 5C. You are making no greater gain on your liquid cooled cpu at 50W power with your liquid cooling vs high end air.

=

I have also stated whatever works for your application, is what you must use - if you need absolutely the last drop of OC capability from your system, in WC format, so be it, then go for it, no prob with that. Then you fall outside the 95% group I mentioned earlier - most people will not want to OC to the brink of instability though, hence my comment of earlier.

Your arguments are good, no problem with that - if you buy a Porsche, why have 13" wheels on it. Get the 20" ones......but we probably did not need a Porsche to begin with

All jokes aside, I am down that same path with my Wife's rig, having a Feser quad 480, ATI 5990, I7 930 at 4Ghz, etc.....and it sports some pretty serious cooling with single-loop, daul D5's, and using an Innovatek fan-controller -> I have all that quality hardware in there, as it suits the build pretty well, is near-silent, and I realise to well, overkill.

Why did I do it then - well, for aesthetics, practicality, symmetry. So, I am not against this, merely saying it's not really required - but I do it anyway, because I can.