Newbie question: Water cooling vs. Air cooling

The alternative to direct air cooling

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

peterson
Posts: 256
Joined: Sat Jan 15, 2005 2:09 am
Location: Gothenburg, Sweden

Newbie question: Water cooling vs. Air cooling

Post by peterson » Sun Jan 30, 2005 5:01 am

I've read chylld's guide but still have some questions. :)

I've recently started to seriously getting my PC noice to minimum. But i'm not satisified yet. I know what to do to get it quieter with air cooling, i just need the cash and acceptable reasons ;)

But of course i'm considering water cooling. But i need some answers first.

With water cooling, is it possible to get it quieter than a very quiet actively air cooled PC?
I now have 2 120mm case fans running at 400rpm. Also a XP-120 CPU cooler w fan running at 800rpm. And a 6800GT w VF700 running at 1000rpm.

If i installed a water cooling kit, cooling the CPU and GPU, with only 1 120mm fan cooling the radiator and the pump inside the reservoir. Would it be quieter than my present system at the same temp?
And can i have everything inside the case and get enough cool air through the radiator?
How long time can a water cooled system run w/o maintainance?

A friend is going to New York in a few week so i'm considering asking him to buy DangerDen waterblocks there. I'd save a lot of money. Also, will an Eheim 1024 be sufficient? And a radiator w only 1 120mm fan?
What speed does your radiator fans normally run at?

Thanks a lot for your help :)

/Lars Peterson

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sun Jan 30, 2005 6:40 am

I've built lots of quiet air cooled systems and own (and constantly fiddle with) a mad-scientist-experiment water cooled machine with twin loops, cooled HDD, automatic fan control and all kind of weird shit like an alu case acting as a heatsink itself. These are my recommendations, might be true or just stupid.

Only one 120 fan on a single radiator cooling a CPU/GPU like that won't even be close to quiet. You need at least 2x120 fan area on a good active rad, or rads, or the fan(s) speed will ramp up. Don't forget the rad itself adds to the air woosh noise.

Your computer already seems to be rather quiet. It takes a fair amount of expertise or trial-and-error to achieve a really quiet w/c setup. And really quiet is what you must aim for to beat what you've got. Unless your HDD is a problem (and needs w/c in an isolation box) I'd strongly recommend you to stay away from w/c, instead, duct your CPU fan and get yourself a Zalman heatpipe thingie for your GPU.

Discussion thread: Watercooling obsolete?

However, should you feel inclined to spend a considerable amount of time and money (and generally more headache) on a good w/c system, it surely will beat your current one.

You can have the rad(s) inside the case if you mod it to fit the minimum 2x120 rad(s) and use them as blowout.

Maintenance should not exceed perhaps changing water once every year if you build everything very thoroughly and get a quality pump. After a couple of years the tubing may start to crack, the blocks start to corrode and everything generally turns into a shit-load of shit but by then it's all going to be obsolete anyway.

Never heard of Eheim 1024, perhps I'm not up to date. Is it the 12V version of the 1046? Or did you mean the 1048? Anyways, check up the stats for the pump and then compare it to this excellent chart, keeping in mind that one or two degrees is nothing but a piss in Mississippi. That means almost any pump is enough. But a too weak pump in a too restrictive loop will damage the pump and or make it start clanging or sort of rattling. I can from personal experience recommend the Laing DDC pump. People always say Eheims are reliable. My 1048 is very quiet but vibrates a lot. Be prepared to mount it in elastic cords and let it touch nothing solid or you're ass is grass.

Your last question is about my radiator fan speeds. They cool an AMD 2600+ and a Radeon 9800. I have two rads, each with one Papst 4412 F/2GL at 3.5V, almost 400 rpm. Mounted with some space between fan and rad it's pretty darn quiet but the best thing is that the sound is very soft, smooth and pleasant. It's the annoying sounds, like a whining HDD, I want to kill. Anyway, you have more watts to handle and if you like gaming it might not be good enough. I have an automatic fan controller ramping it up for me if needed. When playing games the noise problem is not much of an issue compared to when sleeping with DC++ on.


Hope it helped. Din dator lär rocka!

EDIT: ugly spelling and unnecessary poor english found

peterson
Posts: 256
Joined: Sat Jan 15, 2005 2:09 am
Location: Gothenburg, Sweden

Post by peterson » Sun Jan 30, 2005 6:56 am

Thanks for such a rapid and thourough answer! :D

I considered maybe to install 2 radiators w 1 120mm fan each.
The HDDs i intend to silence w the Zalman heatpipe thingy. Not that they disturb me now but they're the loudest component. :)
Zalman heatpipe thingy for GPU is incompatible.

Of course i meant the Eheim 1048. :oops:

It does use too much power. 3.0 prescott. :( I intend to buy a new PSU soon, so that'll be taken care of then. :) And i do game from time to time, but i also have an automatic fan controller, which also could take care of the pump if necessary. It's the T-Balancer.

Thanks for the tip on Laing pump. I'll check it out.

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Mon Jan 31, 2005 1:28 am

Just a note on the DDC pump: It must be isolated because it makes a rather irritating sound. The Eheim is much quieter, but bigger and vibrating instead. Get the DDC wrapped up in something heavy (dense material stops sounds) and you'll be fine.

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Re: Newbie question: Water cooling vs. Air cooling

Post by bobkoure » Thu Feb 10, 2005 10:48 am

peterson wrote:With water cooling, is it possible to get it quieter than a very quiet actively air cooled PC?
IMHO it'll be about the same at idle, but water gives you a lot more "headroom" - the temps don't go up much when running for long periods at 100%, even if you've overclocked the CPU.
I now have 2 120mm case fans running at 400rpm. Also a XP-120 CPU cooler w fan running at 800rpm. And a 6800GT w VF700 running at 1000rpm.

If i installed a water cooling kit, cooling the CPU and GPU, with only 1 120mm fan cooling the radiator and the pump inside the reservoir. Would it be quieter than my present system at the same temp?
The answer is yes - but you need to be using a radiator bigger than 120x120, but "shrouded" so that a single 120mm fan can pull air through it. In theory, a thinner radiator with wide-spaced fins (like the Black Ice Pro series) should give you the best air flow resistance (low noise) to heat transfer trade-off. However the 120mm BIP isn't big enough for you and the BIP2 uses two 120mm fans. If you have room for the BIP2, this still may be your best option. Personally, I use a 6"x6"x2" heater core with a stacked pair of undervolted nexos fans pulling. This system is very quiet, but not silent. (maybe 17dB, including a pair of 200G SATA drives, another pair of nenxos fans on the exhaust side and a Seasonic "tornado" power supply).
And can i have everything inside the case and get enough cool air through the radiator?
Depends on the case and the radiator, of course :)
If you're putting this in a mid or full tower you should have no problem. If you can, try to install the rad in the lower front of your case (so it gets cooler incoming air) and, again, if possible, set it up so that the exhaust from the fan(s) pulling blow at your graphics card. You may get best luck with watercooling for your CPU and a Zalman heat-pipe heatsink for your 9800 - which should work very well with some directed airflow. BTW I'm set up this way, but with a Matrox card that doesn't require the heat pipe.
How long time can a water cooled system run w/o maintainance?
You'll need to check water levels every couple of months (more often if no reservoir) - and vacuum dust out of your radiator about that often, too. It wouldn't hurt to vacuum out the rest of your case, with attention to heat sinks, at the same time.
Other than this, assuming an Eheim 1046/1048/HPPS, I'd give it three years. I built a quiet system for a local recording studio (really a digital mixer with a PC motherboard in it), Eheim 1046, Swiftech (now very obsolete) block and BIP that's been running 24x7 since hmmm.... 1998 or 99. Water gets added every so often, and I've drained and re-filled once.
A friend is going to New York in a few week so I'm considering asking him to buy DangerDen waterblocks there.
IMHO, for a low noise pump (like the Eheim 1046 or HPPS) you'd be better off with the Swiftech 6000. As I mentioned above, I'm not sure I'd watercool the GPU at all.
The 1046 would be sufficient for CPU cooling alone, but you'd need a 1048 or HPPS for GPU as well.
The HPPS is the quietest pump I've ever heard run, then the 1046, then the 1048

burcakb
Posts: 1443
Joined: Tue Mar 09, 2004 9:05 am
Location: Turkey

Post by burcakb » Thu Feb 10, 2005 12:13 pm

Totally unrelated to w/c but, this comment caught my attention.
17dB, including a pair of 200G SATA drives
I was under the impression that hdds alone were around 22-24 dBA??

Rusty075
SPCR Reviewer
Posts: 4000
Joined: Sun Aug 11, 2002 3:26 pm
Location: Phoenix, AZ
Contact:

Post by Rusty075 » Thu Feb 10, 2005 1:43 pm

That's why non-verified dB comments should always be taken with a grain of salt the size of a Volfswagon minibus. :lol:

I'm not calling Bobkoure or anyone else a liar, don't get me wrong. And it is possible that from a listening position it really is that quiet.

But given those HDD's and fans, I doubt it. If it was that quiet, it would likely be described as "silent" since 17dBa is probably below the ambient volume in the room to begin with. I'd also be surprised if any "normal" user would have an SLM capable of measuring down that low. That would require "MikeC" caliber equipment.


Rather than given absolute measurements, which are essentially meaningless, its a better idea to describe sound comparitively: "Its as quiet as an L1A at 12 volts" (which is about 21dBa) :lol:

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Thu Feb 10, 2005 3:23 pm

burcakb wrote:I was under the impression that hdds alone were around 22-24 dBA??
In the open air, sure. However, I'm using Seagates (pretty quiet) and I've lined my case with Blackhole 5, which does a particularly good job soaking up hard drive noise. This foam has an absorbtive depth greater than a half wavelength of the noise I wanted to get rid of - and it has a "limp wall" embedded in it to interfere with mechanical transmission.
I can still hear 'em on seek, of course - lower frequency and transmission through the case metal.
I didn't bring this up as regards radiator fan noise as most of that seems to escape on-axis through the heater core. I'd had real problems finding an under 20dB fan that could pull sufficient air through the radiator. A Japanese 120mm Panaflo at 7V pulled enough air, but was way seriously too loud. At 5V it was almost quiet enough for me, but was in stall when presented with 2" of heater core. Stacked panaflos at 5V could pull sufficient air but were, well, louder. (but quieter than a single one at 7V).

Anyway, I could not hear the drives over a panaflo at 5V, which I estimate to be slightly under 20dB.
The PWMed Nexus fans I'm using now are noticeably quieter than the panaflo at 5V (so 18? 19?) ... and I still can't hear my drives over these, hence my guess of 17dB
Rusty075 wrote:Rather than given absolute measurements, which are essentially meaningless, its a better idea to describe sound comparitively: "Its as quiet as an L1A at 12 volts" (which is about 21dBa)
OK - so it's quieter than a japanese L1A at 5V, which in turn is some unknown amount quieter (and I'm pretty sure it's at least 1dB quieter) than an L1A at 12V which rusty says is about 21dB.

peterson
Posts: 256
Joined: Sat Jan 15, 2005 2:09 am
Location: Gothenburg, Sweden

Post by peterson » Fri Feb 11, 2005 2:15 am

Thanks bobkoure! :)

I've read through so many reviews and the Swiftech is probably one of the best alternatives for me. I like the good low flow performance. And i'll probably start with just cooling the CPU. :)

Also, i've understood that at least 2*120mm rad is needed. I hope i can mount it at the bottom of my future case and suck in air that way. The pump is still something i'm deciding on. Not easy w/o experience. Might go for the Eheim 1048 and try to damp the vibrations. Or i may not. :?

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Fri Feb 11, 2005 6:31 am

The vibrations are pretty easy to damp. Put the pump on a foam rubber mat and be sure to use the most flexible tubing you can find leading to/from the pump to reduce mechanical transmission through that.
the downside of flexible tubing is that, it seems like, the more flexible the tubing, the faster water evaporates (maybe the right word is permeates?) through it. You'll probably want a reservoir. Innovatek makes a push-on reservoir for the Eheim 1046/1048/HPPS called the agb-o-matic, that turns the pump into a very slick hybrid inline/submerged setup. They make some round push-on reservoirs as well, that I've never seen in-person but that appear to restrict flow in the base (looks like a pair of milled holes leading to the pump inlet) the agb-o-matic actually has the "nose piece" of the pump inlet protrude right into the middle of the reservoir area.

If you put the rad in the floor of your case, think about orienting it so airflow from one of the fans is directed at your GPU, which may let you go to a passive heatsink. One less fan in the system.

Using the bottom of the case has the advantage that you can use something non-sound-reflective under the case and reduce a bit of the on-axis fan noise coming out through the radiator.
I tried an experiment with a BIP (1 not 2, but should apply to the 2) in that I put a spacer (made from the frame of a dead fan with "guts" removed) between the pull fan and the radiator. In theory my temps should have gone down a bit. I saw no change - but I think it was slightly quieter. May have been psycho-acoustics playing games with me (very hard to judge relative noise levels with a half hour between) but it's cheap and easy to do and might help your noise level - and it won't make your temps go up.
The downside of using the bottom of the case is that you'll get plugged up with dust faster and you'll have a more difficult time getting a vacuum cleaner brush onto the outside of the radiator (where the dust will mostly accumulate). Maybe your house is cleaner than mine?

Finally - non-watercooling, if you're about to get a new case, consider getting one that's big enough that you can line it with sound-absorbtive foam. The stuff that I've found works best is 1.5" (38mm) thick, which is too thick for most cases. I'll leave the "best sound absorbtive foam" discussion for the "cases" forum, but, when picking out a new case for watercooling and quiet, remember that you are probably going to want to line the case walls with something and that, in this regard, more room is better than less.

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Fri Feb 11, 2005 9:30 am

I second most of bobkoures tips.
Athough I have to disagree with bobkoure on this one:
bobkoure wrote:The vibrations are pretty easy to damp. Put the pump on a foam rubber mat and be sure to use the most flexible tubing you can find leading to/from the pump to reduce mechanical transmission through that.
Perhaps his case is of an exceptionally sturdy type so he gets away with it. But generally you must suspend the pump in elastic cords, letting it touch nothing. Otherwise you can probably forget about finally beating that notorious Panaflo @ 5V.

Watercooling doesn't equal silent bliss. Not trying to be a smartass, but bobcure's claim of total system noise under 17 dB is probably a bit exaggerated. Just the Seasonic Tornado makes more noise even at idle, according to the SPCR review.
The Nexus fans bobkoure uses are excellent. I use them too, but have 4412s on the rads. Even now @ 3.5V I can hear the soft woosh from the rads - and they are soft mounted with shrouds and everything the doctor ordered.
Adding up noise from HDDs (unless watercooled and completely isolated), perhaps a HDD fan, and pump noise, even if in an isolated case, makes it hard to go perceptively silent. Don't hope for w/c to bring you wonders. It's not that easy.


Also, I'd recommend you get yourself a water-block for your graphics. It offers uncomparably better cooling than a Zalman heatpipe and you can still let your water temps go high. Split the water line to ease on the head pressure from the pump. Recommend one line for CPU and rad1, another line with GPU and rad2.
Water-cooling northbridge and stuff is pointless unless you want overclock-dragracing.

If you find the reservoirs expensive and have good fiddling skills you can use a shampoo bottle or whatever. PL-400 construction glue kicks ass. Just be sure everything's bolted and play it safe. A reservoir makes life easier down the road. I've tried both with and without.


Good luck on your w/c endeavour!

peterson
Posts: 256
Joined: Sat Jan 15, 2005 2:09 am
Location: Gothenburg, Sweden

Post by peterson » Fri Feb 11, 2005 12:57 pm

Thankyou both :)

I may not get watercooling components very soon, as i have yet expenses for case and PSU, but it's good to be prepared and knowing hat to get :). I intend to get a Coolermaster CM Stacker case. I just need to detach the grill at the bottom. I also intend to swap the big round intake in the door with piece if plexi glass., or something.

I have 3*4412 on 5V in my case and i too hear them. I'm aware that silence is impossible, so i don't aim for that. I have to dampen HDDs more too.

@snutten: what do you mean by splitting the water lines? I f i have 1 rad that takes 2 120mm fans, how can i split it? Or do you suggest 2 120mm rads?

About graphic card water block. Should i aim for a block that cools memory as well or is it enough with only GPU? I have a 6800GT so there are some blocks that also cools memory.

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Fri Feb 11, 2005 8:10 pm

snutten wrote:. Otherwise you can probably forget about finally beating that notorious Panaflo @ 5V.
erm... that was a reference to damping Eheim pump vibrations. They are relatively low frequency, large amplitude (A.K.A. "shaking"). Just put the pump on a thick foam rubber pad and let it shake a bit.
bobcure's claim of total system noise under 17 dB is probably a bit exaggerated.
It might be - I've no way to be absolutely sure. I know what 20dB sounds like. AFAICT the Nexus fans at 12V are a bit under 20. PWMed to about 3/4 speed, they are considerably quieter - but are still the noisiest thing I can hear from my chair.
... hard to go perceptively silent
This system is not silent. I can hear it, particularly late at night, and particularly now, in the winter with windows closed and a blanket of snow on the ground (I'm out in an exurb so no "city noises").
Split the water line to ease on the head pressure from the pump. Recommend one line for CPU and rad1, another line with GPU and rad2.
Parallel paths? General thinking is that the loss of C/W due to flow loss not worth the small rise in coolant temp through each block in series. Have a visit to the ProCooling liquid cooling forum and either read up or start a new thread explaining why you think parallel is best (depending on how confrontational you care to be).
Water-cooling northbridge and stuff is pointless unless you want overclock-dragracing.
I'd agree that it isn't worth water cooling a northbridge - but then we differ on whether it's worth it to water cool a GPU. If you're overclocking the GPU, well, then, sure. From calling it "overclocking-dragracing" I'd suspect that you are not - the idea being to keep the CPU and GPU cool enough to run safely at the lowest noise level. Think systems-level. You've got an air stream that may be all you need. No, it won't be sufficient for O/Cing the GPU, but it will let you run a smaller, quieter pump, like a 1046.
A reservoir makes life easier down the road.
Yep, in two ways. When filling your system you're going to have a bunch of entrained air. If the water can slow down, the air is more likely to be released. Then later, it's less of a big deal if you lose some coolant as it's just a matter of the level going down in the reservoir (even 'T' line systems have a bit of reservoir - the hose leading to the 'T', but it's not really much volume)

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sat Feb 12, 2005 5:13 am

bobkoure wrote:
snutten wrote:. Otherwise you can probably forget about finally beating that notorious Panaflo @ 5V.
erm... that was a reference to damping Eheim pump vibrations. They are relatively low frequency, large amplitude (A.K.A. "shaking"). Just put the pump on a thick foam rubber pad and let it shake a bit.
The vibrations produce a low sound. Can't stand it.
bobkoure wrote:
Split the water line to ease on the head pressure from the pump. Recommend one line for CPU and rad1, another line with GPU and rad2.
Parallel paths? General thinking is that the loss of C/W due to flow loss not worth the small rise in coolant temp through each block in series. Have a visit to the ProCooling liquid cooling forum and either read up or start a new thread explaining why you think parallel is best (depending on how confrontational you care to be).
The water temp rise passing a block is of course negligible.
If you want Procooling numbers on why parallell loops are the better choice, read Cathar's article on pumps and flow.
In short: Pumps can manage good flow but are weak against head pressure. (I had my 1048 line split and now both loops have better flow than the original single ever had. Passing only cpu-rad and gpu-rad.)

Also, big head pressure damages the pump over time.
Here's another rattling pump, certainly not the first but a perfect example.
Here's me experimenting with pumps and head pressure again.
bobkoure wrote:... No, it won't be sufficient for O/Cing the GPU, but it will let you run a smaller, quieter pump, like a 1046.
Again, the very same procooling article presents the numbers. The 1046 is enough if the line is split. Unless of course you think a temp raise of some 3-6 degrees over best flow is important? All this talk about flow is folly for the o/c crowd. They use big air-flow too, through their tightly finned rads.

Btw, the temp raise in the GPU, if relying on a Zalman heatpipe, can be substantially higher. The airflow needed on a Zalman on a powerful GPU is tricky to achieve in a quiet w/c system. Not to mention petersons 6800GT won't even fit, as he pointed out to me earlier. Zalman warns us about both temps and fit too.


peterson wrote:@snutten: what do you mean by splitting the water lines? I f i have 1 rad that takes 2 120mm fans, how can i split it? Or do you suggest 2 120mm rads?
In fact, you can split it any way you like as long as water resistance in the loops are fairly equal. Especially make sure the CPU block gets fair flow. Since you are going to use low airflow through your rads then the water needs not go that fast either. Use clamps to direct flow if you must but try to engineer it from the outset so the water flows where it's best needed. My recommendad CPU-rad1 + GPU-rad2 is just an easy way to get equal resistance in both loops. No need to think about where the water passes first and so on. It's not like the water is way hot after a block and ice cool after a rad.
peterson wrote:About graphic card water block. Should i aim for a block that cools memory as well or is it enough with only GPU? I have a 6800GT so there are some blocks that also cools memory.
GPU only unless o/c is the standard answer. You can add up with some regular passive heatsinks on memory if you wish. But I'm not up to date on the 6800 so please check it up somewhere else. Just don't check it up on any overclocking forum because they will tell you to get an extra overcharged vapochill per GPU memory circuit :wink:


EDIT: added smiley to prevent unintentional o/c insult
Last edited by snutten on Sat Feb 12, 2005 11:28 am, edited 2 times in total.

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Sat Feb 12, 2005 10:52 am

snutten wrote: Just don't check it up on any overclocking forum because they will tell you to get an extra overcharged vapochill per GPU memory circuit.
Do you know what a "straw man" argument is?
1) Go to procooling
2) post a thread asking about GPU cooling - be sure to mention your emphasis is on quiet/silence
3) See if anyone even mentions vapochill, other kinds of compressor-based cooling, or even TEC (peltiers) which is silent, other than the additional heat shedding required by the add'l heat put into the loop (so potentially more pump, more fan - but also possibly not as you might be able to make it up with increased radiator frontal area)

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sat Feb 12, 2005 11:26 am

I'm not sure what you meant by that last post, bobkoure. But I'll see to it that in the future my last sentence, quoted by you, isn't regarded as completely serious by adding a smiley!

Banten
Posts: 5
Joined: Sat Aug 14, 2004 11:30 pm

Post by Banten » Sun Feb 13, 2005 12:34 am

snutten wrote:Split the water line to ease on the head pressure from the pump. Recommend one line for CPU and rad1, another line with GPU and rad2.
Don't split the flow. You would have better flow in the circuit as a whole, but it would effectively halve the flow through each block (if they offer the same resistance). Sounds reasonable, no?

I also recommend Pro/Forums if you really want to know what's what

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sun Feb 13, 2005 7:33 am

Banten wrote:Don't split the flow. You would have better flow in the circuit as a whole, but it would effectively halve the flow through each block (if they offer the same resistance). Sounds reasonable, no?
In fact it doesn't. Pumps manage big flows but poor head. Did you even follow up the link I presented above? It links to the much appraised Procooling. The numbers are quite clear.

I've tried both ways. I empirically know the flow from a 1048 (peterson's leaning towards a 1048 right?) will in fact increase in both loops compared to the single original loop if it passes 2 blocks and 2 rads. And my blocks are old Innovateks. There are new blocks out there with even bigger water restriction I think, but admittedly I haven't bothered to check this up thoroughly enough to know.

With a DDC things might change. It has smaller flow and much better head. Haven't tried it out.

There are so many myths circulating around about water-cooling. Please, if you chose to follow up on this, give me some facts to prove it and I'll admit I was wrong about the whole thing. Or try it yourself if you have the pump. I don't care to turn this into a pie throwing contest.

Peterson, I'm sorry this seems no longer to be helpful for you. You're going to have to use your own judgement!

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Sun Feb 13, 2005 10:47 am

snutten wrote:I've tried both ways. I empirically know the flow from a 1048 [...]will in fact increase in both loops compared to the single original loop if it passes 2 blocks and 2 rads..
Let me try so re-say that so you can tell me that I understand what you are saying.
Say you have a set length of tubing, two raditors, two blocks and one 1048.
Setup entirely in series you get some flow rate, say X (no idea what it really is
If you set this up in parallel, not only is total flow higher than X, but the flow in each branch is greater than X, meaning your total system flow has gone to >2*X.
Do I have this right?
If I do, I'm wondering how applicable your results are to the general situation (mostly wondering if this is due to the combination of a low-head pump with what is (in comparison, say,with 3/8" ID tubing and Swiftech 6000 blocks) quite restrictive loops made of Innovatek blocks, radiators (series flow?) and possibly-small-ID tubing.

I'd strongly encourage you to post this empirical data over on procooling. They (we) are really not just a bunch of overclocking crazies. Not that some/most filks there don't overclock, but silence is regarded as a legitimate systems engineering goal. You read some forums there, or you wouldn't be pointing at Cathar's posts, so it's not like you're a stranger there.

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sun Feb 13, 2005 2:52 pm

You said it better than me, bobkoure. Flow in each of the two loops superseed the flow in the single original loop. Total flow is more than double.

I use 12/16 mm (12 mm inner diameter) aquarium tubing (= 0.4624/0.6299 inches, quite bigger than 3/8 ID) and there are no sharp corners or anything.
I was under the impression that modern blocks like the whitewater are more restrictive than my Innovateks (latest revision)? If so, then we can expect even better results from a modern system.


I'm just a silent freak and I dunno if I really want to engage myself in yet another forum. But thanks for asking!
Anyway, my tests aren't groundbreaking or anything. Anyone who knows how to applice the numbers can read read it black on white (or actually white on blue @ procooling) from Cathars tables. Feel free to point to this thread if you wish to tell them about it.

There's so much baloney going around focusing on maximum cooling effect. People getting a bigger pump to gain a degree or two. Flow this much, use hippest block and rad with four fans. Cathars article simply excellent. Clearly showing us that flow is not at all crucial and still people continue the hunt for decimals. I'm just too old I guess :(

What matters is how you dump the heat. Where the noise comes from to begin with. Big rads are key. That's my take on the whole w/c business. Like I said, the water's just there to move the heat a couple of centimeters. People with heatpipe solutions rarely talk about how turbo-charged the pipes are, they concentrate on the actual heatsink/fan situation. I think they're right.
And by the way, bobkoure, thanks for taking me seriously instead of going on arguing about beliefs (nobody mentioned, nobody forgotten)! :D

pdf27
*Lifetime Patron*
Posts: 264
Joined: Wed Nov 12, 2003 12:59 pm
Contact:

Post by pdf27 » Mon Feb 14, 2005 8:41 am

I'd say it probably is possible, but only for some pumps/block combinations. Presumably you'll have to be operating in a very shallow slope section of the P-Q curve, and have in practice to be operating identical blocks. It isn't like you've got any straight lines on the graph, so things get a little bit complicated.
Problem is, most good CPU blocks are more equivalent than the equivalent GPU blocks, so unless you fudge it a bit you'll end up cooling the CPU rather badly.
Were I ever to watercool a Northbridge though, I'd stick a restrictive NB block in paralell with a free-flowing GPU block - total restriction shouldn't be any more, and the NB wouldn't need much flow. Helps keep all the dP in the CPU block, and hence better performance (probably more flow too).

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Tue Feb 15, 2005 8:10 am

pdf27 wrote:Problem is, most good CPU blocks are more equivalent than the equivalent GPU blocks, so unless you fudge it a bit you'll end up cooling the CPU rather badly.
Did you mean "more restrictive than"?
snutten wrote:Use clamps to direct flow if you must but try to engineer it from the outset so the water flows where it's best needed.
Letting some 1/3 of the water pass the GPU on a parallell loop is bound to hurt flow through the CPU less than having them run in serial. In my case, it improved CPU flow. Almost all pumps have that flat curve. The 1048 certainly does.
This seems controversy and judging from the responses even more so than I imagined. All the modders cry serial so I started out with a serial setup too. The myth feeds of itself.
Reports of broken our clanging pumps increase and nobody stops to think why.
The rads only need high water-flow if you use high air-flow but even on SPCR flow is emphasized.
Few w/c setup pics presented on internet look anywhere close to quiet. Big pump isolated with bit of foam, single rad blowing in - doesn't matter then how ultra extra spiffy the blocks are.

I'm not going to push my case anymore. Only way you're going to know is to try it yourself. Feel free to dismiss it as a freak accident :)

pdf27
*Lifetime Patron*
Posts: 264
Joined: Wed Nov 12, 2003 12:59 pm
Contact:

Post by pdf27 » Tue Feb 15, 2005 9:16 am

snutten wrote:
pdf27 wrote:Problem is, most good CPU blocks are more equivalent than the equivalent GPU blocks, so unless you fudge it a bit you'll end up cooling the CPU rather badly.
Did you mean "more restrictive than"?
Yes :oops:
snutten wrote: Letting some 1/3 of the water pass the GPU on a parallell loop is bound to hurt flow through the CPU less than having them run in serial. In my case, it improved CPU flow. Almost all pumps have that flat curve. The 1048 certainly does.
This seems controversy and judging from the responses even more so than I imagined. All the modders cry serial so I started out with a serial setup too. The myth feeds of itself.
Reports of broken our clanging pumps increase and nobody stops to think why.
The rads only need high water-flow if you use high air-flow but even on SPCR flow is emphasized.
Few w/c setup pics presented on internet look anywhere close to quiet. Big pump isolated with bit of foam, single rad blowing in - doesn't matter then how ultra extra spiffy the blocks are.

I'm not going to push my case anymore. Only way you're going to know is to try it yourself. Feel free to dismiss it as a freak accident :)

I don't think it's necessarily a freak accident, but nor do I think you can really apply it as a general rule to all systems.
You've actually got your own thread on ProCooling right now, and you certainly aren't being dismissed out of hand. Hopefully at least one person (Nickd) will be running some numbers to get an idea of when this is and isn't a good idea. For instance it may turn out that for low restriction blocks with a 1048 it is good, but with a DDC it isn't. Until we have some numbers to back it up (either lots of theoretical ones or quite a lot more practical ones) it really isn't much beyond a nice theory or coincidence.
http://forums.procooling.com/vbb/showth ... post135254

HammerSandwich
*Lifetime Patron*
Posts: 1288
Joined: Sat Oct 25, 2003 3:21 pm
Location: 15143, USA
Contact:

Post by HammerSandwich » Tue Feb 15, 2005 11:57 am

I posted in the thread at PC, with an incorrect interpretation of your claim, Snutten. I apologize for any confusion.

Now, I did conclude that each block can get more flow in parallel. I strongly suspect that getting this to occur depends on a few items (some discussed above).

First, the blocks must be the most significant restriction. Other restrictions - think of a system with only one radiator - will become far more important when flow doubles.

Second, the pump must be relatively close to deadhead/stall with series operation. The new operating point on the P-Q curve will have >2x the flow and cannot afford to lose a lot of pressure. With the shape of a P-Q curve, this can happen only in the high-pressure/low-flow region.

Third, the pump must have relatively limited pressure capability. This provision helps ensure that the previous one applies. But while it's theoretically possible to choke a DDC to 1LPM, that's probably not going to happen in a real-world system. Low-head pumps are far more likely to benefit from parallel blocks.

As always, it becomes a matter of overall system design. For Snutten's 2 rads and weak pump, parallel is better. For a more typical WC system with a single rad and higher pressure pump, series is probably superior.

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Tue Feb 15, 2005 11:57 am

Of course there's controversy. You're suggesting something that's exactly opposite to the way thing have been observed to work - at least with high-flow, high-head pumps.
I got the same thing when experimenting with axial fans undervolted to the point of stall (searching for quiet).

I'm not questioning your results (although I've asked a friend I've loaned a 1046 to if I could "borrow it back" to try to replicate your results) but I would strongly suggest you not try to generalize from this one data point to all pumps/loops.

I'm curious as to how you were measuring flow. A couple of buckets and a stopwatch? That's how I've done it in the past - but just to get a guess at flow numbers. I was not trying to control for zero head at the inlet.

I'd also suggest avoiding any kind of claims about restriction being bad for pumps - other than the exact circumstances you have observed (which are interesting but may not generalize to other situations). But please do share the data points you do have.

IMHO, there's nothing wrong with a "single rad blowing in" (so long as the rad is appropriately sized). I've had best noise/cooling results with a radiator with the fan(s) on the "pull" side.

I wouldn't so much dismiss what you've found as a "freak accident" as a small data set - which may or may not generalize to more situations. The stuff I ran into re axial fans most definitely did not generalize (but I wasn't expecting it to

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Tue Feb 15, 2005 4:16 pm

This is becoming rather intriguing. Hope I can clarify some of my standpoints here. Happy to answer any questions now that you well informed guys have taken an interest and I no longer feel I'm just pissing against the wind.

HammerSandwich wrote:I posted in the thread at PC, with an incorrect interpretation of your claim, Snutten. I apologize for any confusion.
No problems you can quote my on anything @ procooling. I've just read the thread. Coming out as an obnixious asshole just goes to show them I've been around on the internet for a while :D
Thanks for the link!
HammerSandwich wrote:Now, I did conclude that each block can get more flow in parallel. I strongly suspect that getting this to occur depends on a few items (some discussed above).

First, the blocks must be the most significant restriction.
Why?
HammerSandwich wrote:Other restrictions - think of a system with only one radiator - will become far more important when flow doubles.
I totally agree! Anything in the line where flow is double (not split) naturally increases on the resistance even more. More than double in fact!
snutten wrote:Recommend one line for CPU and rad1, another line with GPU and rad2.
snutten wrote:I empirically know the flow from a 1048 (peterson's leaning towards a 1048 right?) will in fact increase in both loops compared to the single original loop if it passes 2 blocks and 2 rads. And my blocks are old Innovateks. There are new blocks out there with even bigger water restriction I think, but admittedly I haven't bothered to check this up thoroughly enough to know.
...blocks and rads on a 1048
HammerSandwich wrote:Second, the pump must be relatively close to deadhead/stall with series operation. The new operating point on the P-Q curve will have >2x the flow and cannot afford to lose a lot of pressure. With the shape of a P-Q curve, this can happen only in the high-pressure/low-flow region.

Third, the pump must have relatively limited pressure capability. This provision helps ensure that the previous one applies. But while it's theoretically possible to choke a DDC to 1LPM, that's probably not going to happen in a real-world system. Low-head pumps are far more likely to benefit from parallel blocks.
Of course!
snutten wrote:With a DDC things might change. It has smaller flow and much better head. Haven't tried it out.
Bigger bad-ass (or noisy as I prefer call them) pumps not considered.
HammerSandwich wrote:As always, it becomes a matter of overall system design. For Snutten's 2 rads and weak pump, parallel is better. For a more typical WC system with a single rad and higher pressure pump, series is probably superior.
I'm with you all the way.


bobkoure wrote:I'm not questioning your results (although I've asked a friend I've loaned a 1046 to if I could "borrow it back" to try to replicate your results) but I would strongly suggest you not try to generalize from this one data point to all pumps/loops.
...blocks and rads on a 1048
bobkoure wrote:I'm curious as to how you were measuring flow.
Never measured the flow. It was such an obvious improvement, easily seen in my reservoir.
bobkoure wrote:IMHO, there's nothing wrong with a "single rad blowing in" (so long as the rad is appropriately sized).
Yep, size is key. But keeping rads on intake, blowing into the case, raises temps for PSU, HDDs and MB. Cooling them makes noise. Or did you interpret "in" as "push" config with fan pushing through rad, never mind if it's a blow-out?
bobkoure wrote:I'd also suggest avoiding any kind of claims about restriction being bad for pumps
Perhaps you're right. I'll take heed and add a "may" henceforth. Or a "probably" when feeling a bit naughty :wink:
I also kept aquariums as a hobby. Same pumps, but free flowing. They lasted for years and years, never any clanging noises. Ruined two pumps in my computer. Numerous problems reported here on SPCR. Exempli gratia (my bold):
Lapinou wrote:(....)Anyway, thanks to your suggestions my (external) 1048 is now completely silent (barring a very very slight normal hum). The problem was the impeller blade, and the fact that it is loose on the the support (it can move through half a turn). Supposedly this is to allow it to pump stones (prevalent in fish tanks, the pumps main market) with snapping blades, but uneven back pressure can cause it rattle back and forth making a terrific noise.

Not having too many stones rattling around in my resertator I glued the impeller to the support and voila, the noise disappeared.

Fantastic! Silence truely is golden.

Edward Ng
SPCR Reviewer
Posts: 2696
Joined: Thu Dec 11, 2003 9:53 pm
Location: Scarsdale, NY
Contact:

Post by Edward Ng » Tue Feb 15, 2005 6:54 pm

snutten wrote:Bigger bad-ass (or noisy as I prefer call them) pumps not considered.
Who reported DDC being louder than 1048? All the things I've read so far indicate them to be comparable, with neither pump winning out in acoustics.

HammerSandwich
*Lifetime Patron*
Posts: 1288
Joined: Sat Oct 25, 2003 3:21 pm
Location: 15143, USA
Contact:

Post by HammerSandwich » Tue Feb 15, 2005 7:50 pm

snutten wrote:Coming out as an obnixious asshole just goes to show them I've been around on the internet for a while :D
Or at least you're familiar with PC's modus operandi. See my sig (from Phaestus) over there!
snutten wrote:
HammerSandwich wrote:First, the blocks must be the most significant restriction.
Why?
Poorly written on my part, considering your 2 rads. How about "any components to be changed to a parallel configuration must be the most significant restriction when in series"?
snutten wrote:...blocks and rads on a 1048
That's my second "OOPS!" today. Guess I let thread title at PC fool me. Sigh...
snutten wrote:I'm with you all the way.
Yeah, it's good to be reaching some consensus. I'm curious to see where PC ends up on this issue.
snutten wrote:Never measured the flow. It was such an obvious improvement, easily seen in my reservoir.
BOOM! This is a real problem, and doesn't look good after your earlier "I empirically know the flow from a 1048 (peterson's leaning towards a 1048 right?) will in fact increase in both loops compared to the single original loop if it passes 2 blocks and 2 rads." No doubt there was a useful increase in total flow, but double or better now seems a pretty dubious claim. The response at PC should be fun.
snutten wrote:Yep, size is key. But keeping rads on intake, blowing into the case, raises temps for PSU, HDDs and MB. Cooling them makes noise.
Some of us have separate thermal zones. To be fair, I found that a necessary change when my PSU ramped up from post-rad warm air.

Rusty075
SPCR Reviewer
Posts: 4000
Joined: Sun Aug 11, 2002 3:26 pm
Location: Phoenix, AZ
Contact:

Post by Rusty075 » Tue Feb 15, 2005 8:39 pm

HammerSandwich wrote:
snutten wrote:Never measured the flow. It was such an obvious improvement, easily seen in my reservoir.
BOOM! This is a real problem, and doesn't look good after your earlier "I empirically know the flow from a 1048 (peterson's leaning towards a 1048 right?) will in fact increase in both loops compared to the single original loop if it passes 2 blocks and 2 rads." No doubt there was a useful increase in total flow, but double or better now seems a pretty dubious claim. The response at PC should be fun.
Worrying about what the flow rate is or isn't is a red herring. Lets not forget what the purpose of the cooling loop is. If the temps went down while using the same components, rads, fan voltages, etc, that's all the empirical evidence needed to show that the parallel config. is working better in snutten's rig than the serial config was. (You could say that if the temps went down the flowrate must have gone up, but that's not necessarily true either)

I could see the situation where the right combination of PQ curve, pressure drops, dT, phases of moon, etc, all come together to make it work. A limited number of situations to be sure, but possible through either luck (in snutten's case) or careful design. There are plenty of larger scale cooling systems that use parallel flow designs exclusively, but for our little systems, serial is almost always a better answer; not just because of the temperatures produced but for issues such as all that extra piping parallel runs require you to squeeze into your case.

bobkoure
Posts: 191
Joined: Sun Aug 11, 2002 3:26 pm

Post by bobkoure » Tue Feb 15, 2005 9:12 pm

Rusty075 wrote:Worrying about what the flow rate is or isn't is a red herring. ... If the temps went down...
Well, sure - if you have a reliable way to measure temps. I don't have one - I certainly wouldn't call MBM readings "reliable". Then, if the wbs have been dismounted/remounted there's a definite possibility for variation there (better/worse mounting job, slightly more/less hold-down pressure, etc.). Probably some other issues, too.
Sigh...

Post Reply