Watercooling Guide @ Toms

[asdfqwertyuiop]

This article is pretty funny.

http://www.tomshardware.co.uk/2007/03/2...or-watercooling-your-pc_uk/page10.html

Our methodology is simple: push the e4300 as far as it will go with its stock air cooler, then benchmark it with the water cooling system and compare results. As it turned out, the e4300 was willing to go much higher than its stock 1.8 GHz. With the CPU frequency increased from a stock 1800 MHz to an overclocked 2250 MHz, the e4300 easily handled the 450 MHz over clock without a voltage increase or other problems. However, the stock cooler proved it wasn't up to the task, with the CPU temperature rising to an uncomfortable 62 degrees Celsius under load. While the core would have gone higher, we weren't happy increasing the temperature at this point so we recorded the results and installed the water-cooling system.

So they decided how far the PC would OC on its stock cooler, and then compared temperatures with water cooling. What for? If you're spending extra $$$ on a watercooler, then it's to push it further. So if the CPU does 2250 on stock cooling, then try push it to 2500 on water.

And why compare a stock cooler ($0) with a water cooler ($500)? You compare it with the most bad-ass air cooler.

They forgot to put the graphs in the article, but here they are

http://www.pollardbanknote.com/_demos/benchs.gif

WTF does high fan speed mean? RPMs please guys. And noise measurements.

Nobody cares if their CPU is at 45 or 35 degrees.

What we want to know is that

(a) the max OC achievable with the best HSF available at full speed (noisy)
(b) the max OC achievable with the watercooling setup running full speed (also noisy)
(c) the max OC achievable with a silent fan - i.e. a Thermalright Ultra 120 with a fan running at about 1200rpm
(d) the max OC with the watercooling with a silent fan/off

The only things that matter are: max OC, and noise.

Why would anybody compare a watercooling solution with the stock HSF one and conclude that

"The results we recorded from liquid cooling our test rig are pretty clear: liquid cooling is vastly superior and more efficient than air-cooling."

What can their watercooling setup do that the free one cannot? Nothing, according to the article - they didn't OC it. Is it quieter? Who knows, no measurements from these guys.

Their cooling of their X1900 XTX is a joke too.

"We had one of the best air coolers money can buy on it: the Thermalright HR-03. After a 10 minute stress test with Atitool's artefact tester, what advantage did water-cooling provide?"

"We can see that the stock temperatures are atrocious: 89 degrees on the GPU and over 100 degrees on the voltage regulator! The Thermalright HR-03 does an incredible job cooling the GPU down to 65 degrees, but the voltage regulator is still incredibly high at 97 degrees!

The water block cools the GPU down to 59 degrees. This is a massive 30-degree improvement over the stock setup, but only a six-degree improvement over the HR-03, which looks even more impressive in this light.

However, the separate water block for the voltage regulator works wonders for that component. While the HR-03 has no means to cool the voltage regulator, the water block brings it down to 77 degrees, which is 25 degrees below the stock heat sink. This is a very positive result."

In the graphs (above), they show the Koolance with 77 degrees on the voltage regulator and the HR-03 doing 97. THat's b ecause THEY HAVEN'T COOLED THAT PART!

So um, cooling a part is better than not cooling it. Wow. Must buy a watercooling kit now.

So Toms, water cooling is now mainstream is it? Everybody should get it, because, er, according to their article it won't make their PC any faster at all.

Ok....

 

[aigomorla]

Originally posted by: asdfqwertyuiop
This article is pretty funny.

http://www.tomshardware.co.uk/2007/03/2...or-watercooling-your-pc_uk/page10.html

Our methodology is simple: push the e4300 as far as it will go with its stock air cooler, then benchmark it with the water cooling system and compare results. As it turned out, the e4300 was willing to go much higher than its stock 1.8 GHz. With the CPU frequency increased from a stock 1800 MHz to an overclocked 2250 MHz, the e4300 easily handled the 450 MHz over clock without a voltage increase or other problems. However, the stock cooler proved it wasn't up to the task, with the CPU temperature rising to an uncomfortable 62 degrees Celsius under load. While the core would have gone higher, we weren't happy increasing the temperature at this point so we recorded the results and installed the water-cooling system.

So they decided how far the PC would OC on its stock cooler, and then compared temperatures with water cooling. What for? If you're spending extra $$$ on a watercooler, then it's to push it further. So if the CPU does 2250 on stock cooling, then try push it to 2500 on water.

And why compare a stock cooler ($0) with a water cooler ($500)? You compare it with the most bad-ass air cooler.

They forgot to put the graphs in the article, but here they are

http://www.pollardbanknote.com/_demos/benchs.gif

WTF does high fan speed mean? RPMs please guys. And noise measurements.

Nobody cares if their CPU is at 45 or 35 degrees.

What we want to know is that

(a) the max OC achievable with the best HSF available at full speed (noisy)
(b) the max OC achievable with the watercooling setup running full speed (also noisy)
(c) the max OC achievable with a silent fan - i.e. a Thermalright Ultra 120 with a fan running at about 1200rpm
(d) the max OC with the watercooling with a silent fan/off

The only things that matter are: max OC, and noise.

Why would anybody compare a watercooling solution with the stock HSF one and conclude that

"The results we recorded from liquid cooling our test rig are pretty clear: liquid cooling is vastly superior and more efficient than air-cooling."

What can their watercooling setup do that the free one cannot? Nothing, according to the article - they didn't OC it. Is it quieter? Who knows, no measurements from these guys.

Their cooling of their X1900 XTX is a joke too.

"We had one of the best air coolers money can buy on it: the Thermalright HR-03. After a 10 minute stress test with Atitool's artefact tester, what advantage did water-cooling provide?"

"We can see that the stock temperatures are atrocious: 89 degrees on the GPU and over 100 degrees on the voltage regulator! The Thermalright HR-03 does an incredible job cooling the GPU down to 65 degrees, but the voltage regulator is still incredibly high at 97 degrees!

The water block cools the GPU down to 59 degrees. This is a massive 30-degree improvement over the stock setup, but only a six-degree improvement over the HR-03, which looks even more impressive in this light.

However, the separate water block for the voltage regulator works wonders for that component. While the HR-03 has no means to cool the voltage regulator, the water block brings it down to 77 degrees, which is 25 degrees below the stock heat sink. This is a very positive result."

In the graphs (above), they show the Koolance with 77 degrees on the voltage regulator and the HR-03 doing 97. THat's b ecause THEY HAVEN'T COOLED THAT PART!

So um, cooling a part is better than not cooling it. Wow. Must buy a watercooling kit now.

So Toms, water cooling is now mainstream is it? Everybody should get it, because, er, according to their article it won't make their PC any faster at all.

Ok....

Umm its a well written guide. However his product choice wasnt the greatest. Koolance should be avoided! i am telling every member here right now. THERE ARE CHEAPER AND BETTER UNITS.

The editors at Toms should check out my Opty.
Load Temps on Opty

Also my E6600 was doing 26C idle and 38C Load at 1.45V at 3.6ghz. So Toms can BITE me. I havent seen ANY AIR kit that gets close to my Load temps on my E6600.

If your going to do a comparision with top water and AIR. USE MY SETUP


I'll EAT any airsetup in both noise and performance. Only setup quieter then my NAS is the Zalman full Passive.


Super Quiet Setup

And yes that top quiet setup will EAT any koolance setup you compose.


I dont like toms hardware. For the longest Time they stuck to intels HT when AMD64's came out. They kept insistant that intel was the better machine. ROFL... X2's will Stomp the prestler and Smithfields with blind folds.

I just hope they dont intend to lead a bunch of noobies to migrate to koolance products. This would be a very very sad day in watercooling.

Also Not one person who has done water correctly will ever regret the move nor go back. Its a black hole thats very very deep and very expensive

To also be honest, i think more people regret migrating back to air on a successful setup, then people wanting to migrate back to air. But keyword is sucessful setup!


EDIT: heh i just made a membership on there fourms, to show them wat my watercooling setup looks like. I think there in for a rude awakening.

 

[BonzaiDuck]

Haven't posted much for a while, took a look at this thread, and offered some observations about cooling methods.

For a while, I had casually planned to build an evaporative "bong" water chiller setup. The additional parts for it, depending on the peculiar opportunity to use seemingly unrelated household junk (like a discarded Igloo beverage cooler, or automotive "heater-core" junk-yard radiators), could cost maybe $60 to $150, a little more or less, depending on a perference for more than one water pump, a spare power-supply, some seemingly-incidental insulation grommets for CPU or GPU to eliminate condensation, choice of chilled-water reservoir filtering.

Power-consumption in any chilled setup varies by choice of TEC or the more simple and primitive evaporative "tower." The latter probably uses enough electricity for one extra fan and maybe an additional pump depending on whether you want the steam and mist outside the house, or inside the house. Having it inside the house has benefits to consider; locating it outside might increase the rate of evaporation and increase effectiveness. Some attention to keeping the chiller supplied with water, or some "drip-regulation" device hooked up to the home plumbing, addresses some risk implications and inconvenience. Having it outside during a rainy spell would decrease effectiveness, since the evaporative design works best with low air humidity.

And all total, the problem with the "bong" that isn't necessarily the problem with "completely internal" water-setups, is that moving the computer from room to room also means moving the cooler from room to room. The "outside" cooler pretty much limits your choice of where to put the computer indoors -- once you've determined how to run the hoses into the house, or where to put them. That's the same problem with external water-cooled extensions in general.

With water, there may be some limited -- almost imaginary -- risk of leakage damaging internal components -- probably more a result of negligence than anything. Putting components like the radiator and fan on the back of the computer may lower the risk by putting a segment of the water-cooling apparatus outside the case. Case cooling might be improved as well, given the opportunity to reduce the number of hose connections that are inside the case. Adding complexity in general will increase risk, although a lack of redundancy also increases risk for key components, to include examples of fan-deployment in air-cooled setups.

For many of the reasons I mentioned as disadvantages, I decided to pursue air-cooling just before heatpipe coolers become broadly available. I actually made my last build based initially on the XP-120 -- later exchanged for an SI-120, and I put a ThermalRight heatpipe cooler on the VGA card. And about the time I was installing the XP-120, I read some articles on motherboard ducting at OverClockers by John Cinnamon and another cotributor whose name (I think) was Legget.

Effective motherboard ducting will reduce CPU temperatures by 25 to 30C degrees, in comparison to any given stock cooler's load-level thermal power. In Cinnamon's tests, ducting provided CPU temperatures just 9C degrees higher than a water-cooling rig, although the choice of the cooler he deployed may not have been the best -- readers should judge for themselves. Also, in these articles, the test results showed an even greater improvement to VGA cooling with some special yet simple duct extensions to the GPU side of the VGA card.

But keep in mind that heatpipe coolers have surpassed some very fine water-cooling kits in effective thermal resistance, a high but acceptable level of fan noise from an amply powerful but PWM-limited 120mm fan -- running at its highest around 2,500 rpm, which probably isn't even necessary. The appearance of these heatpipe coolers occurred after the dates of these "ducting" projects, which only used copper heatsink-fan assemblies. For processors generating within 120W +/- 10W of thermal power, fan speed can be kept to around 1,800 rpm, and additional minor improvements to case interior and fan-deployment may reduce again the effective dBA contribution of fans in any sort of air-cooled rig.

Several comprehensive comparison reviews show, for example that the ThermalRight Ultra 120 has just about the lowest thermal resistance and greatest reduction in load temperatures under controlled, chosen room-ambients among all available heatpipe coolers for LGA775 or socket-939/AM2-class processors. The C/W values were reliably measured in two different ways by Joe Citarella -- OverClocker's heatsink-review guru -- and shown to be 0.11 C/W. I've been able to use other comprehensive test reviews to estimate the thermal resistance of the yet-to-be-released Ultra 120 Extreme, and my calculations put it at 0.097 C/W.

By comparison, the effective thermal resistance of a water-cooling kit like Swiftech's high-end H20-220-APEX ULTRA is about the same as that of the Ultra 120 heatpipe cooler, or 0.11 C/W.

More than likely, water-cooling will deploy fans. Radiators need fans. An additional PSU for chilled-water-cooling will increase the number of fans, or it will increase other requirements if the PSU is to be cooled by water or some other passive-air method.

An increase in the number of intake or exhaust fans with lower user-chosen speeds will reduce noise and further change the interior case air-pressure. A decrease in number of fans may reduce certain aspects of amplified noise, but increase noise due to turbulence or noisy motors. So an increase in the number of fans may reduce noise from the latter factors. Carefully choosing an airflow strategy, the number of fans can be kept at minimum for the desired effect, and they can be deployed to keep them out of sight, with mounting locations that further reduce noise. For instance, a fan behind a drive cage and in case-center can cool the drives with less apparent noise at the front of the case.

Ducting benefits from pressurized-case configurations, because it requires air-flow patterns that force air through narrower apertures near the components to be cooled, and in such a way as to exhaust the warmer air immediately from the case after it passes by the component.

We also see that Dell and Sony have developed "hybrid" products which use both TEC and water. And with chilled-water models, some sort of temperature-control for TEC will hold temperatures closer to a constant level. How easily this can be accomplished at any given room-ambient without TEC might also mean that average component temperature is raised slightly to keep idle and load temperature range at a minimum, for idle-to-load range also would provide an enhancement to stability.

Still, this makes DIY projects more complex, and adds to complexity and number of components -- components that can eventually fail.

It may be that water-cooling can make for lower temperatures with chilling, or even slightly lower temperature-levels for the same idle-load range. It also may be chilled water-cooling may be more controllable, further minimizing the idle-load range and the overall temperature level. But heatpipe coolers have matched or surpassed water-cooling without TEC, and there are some heatpipe coolers in hybrid products which are supplemented by TEC.

It would seem possible that for essential components, the temperature reduction from ducted air-cooled designs could overtake the minimum temperature levels of a mainstream water-cooling kit without TEC. It might also be possible to improve unassisted water-cooling with air-cooled ducting in a hybrid of a different combination of technologies.

The example in the UK THG article refers to the temperature of a VGA voltage-regulation component. Air-cooling with ducting can decrease the temperatures on these and other components of lesser concern, without adding equipment -- provided the essential CPU, chipset and VGA cooling devices can be chosen for best performance.

So even if some individual component may have a higher minimum temperature, there is an overall effect on a multitude of components that would not be addressed in water-cooling because of a limitation on the number, availability and feasibility of water-blocks. Reducing temperatures on any single component electrically and physically connected to other components would have some effect toward reducing temperatures on all other components. It makes more sense to cool the motherboard and a plethora of components with air, and we now have motherboards which employ single heatpipe cooler designs to reduce temperatures on Northbridge, Southbridge and a collection of Mosfets around the CPU.

For the trouble involved, and some minor concern for the longevity of my parts-investment as a result of complexity, I've indefinitely deferred building my "bong" cooler, or experimenting with TEC. So for the time being, I'll stick to heatpipe coolers and foam-board prototypes that eventually get replaced by constructions of Lexan and Poly-Zap.
Fans are simple, replaceable, easily redundant; you need fans with water-cooling anyway; and I might want to move my computer between rooms without double trouble. Now, there are fans with 140 to 250mm diameters, which reduce noise, increase CFMs and filter room-air. CFMs are important in air-cooling with or without ducting, but the fan-generated noise level can be reduced since the bigger fans run at lower rpm and generate less turbulence and "white-noise." The improved indoor air-quality will also relieve your sinuses.

I continue to fantasize about converting a cheap office water-cooler for use in some Rube Goldberg contraption. Maybe. Who knows? It depends on how much I want to build a "War Games" WOPR or something to rival Ben Kingsley's liquid-nitrogen-cooled super-computer in the movie "Sneakers."

"I tell you, JOSHUA called me back, and asked me to continue playing 'Thermonuclear War!'"

"Marty -- you know this -- we started this -- now it's all about the little ones and zeros. Think of the places we could go, Marty!"

"I know, Cos'. There's nobody there."

It's the information, Marty! The in-for-mation!! . . . "

What is it "really about?" That's almost a personal matter. For me, though, it's not about the highest CPU over-clock setting or "THE" highest benchmark scores. It's about a higher over-clock setting in conjunction with widening other bottlenecks -- more L2 cache, faster memory running at lower latencies, faster hard disk arrays with blinding-fast improvements in bandwidth, and more bandwidth within financial reason for VGA components. And it's about getting that balance within warranty specs, and longevity expectations.

That's it. That's it. Holistic Bandwidth Improvement, Stability, and Reliability, plus Lebensraum for expansion later according to a modest budget, since money may be part of the object. Air has some advantages; chilled water may offer some others. When you can over-clock a C2D from 2.6 to 3.6 Ghz, run memory rated at DDR2-1000 at DDR2-800 with latencies like 3,4,4,9, and get sustained transfer rates from a drive array that are double or triple those of a single SATA-II drive without grabbing clock cycles from the CPU, your need and desire to run the system at a constant 25C may not be worth the trouble when your peak temperatures never exceed 40C.

Even with an extreme commitment to refrigeration and benchmark scores, you may find your need for bragging rights begging to be fed with more money the following year.

I'd rather just focus on "building a good computer," and watch again for technological waves to surf -- capturing a trade-off of reasonable price trends and performance gains.

 

[aigomorla]

its too long to quote so i'll be brief.

show me a computer thats on AIR with these load temps on BOTH GPU + CPU

Nah, i couldnt pull this off with AIR either

My board would melt on AIR with this

gg?

But i wish you did your bong chiller. I want to see one up close with pics. :T

:]

 

[BonzaiDuck]

Well -- I want to be both gentle and modest here. I've looked over all three of your screen-shots -- I assume this is the same Opteron -- the 175 used in the 2005 Max PC Dream Machine, and that one screen-shot shows an idle CPU of between 28 and 31C, while your load temperature at a higher clock is 41 and 33C for the respective cores.

This is all about how much peak thermal wattage your processor generates, and I didn't pay close attention to your percentage over-clock achievements.

My P4 Prescott 3.2E is rated at 105W. I could be wrong, but your Opteron should probably fit the profile of several Athlon 64 and Athlon 64 x2 processors, which all have a rated TDP of around 95W -- at stock speeds. All I see on your screenshots is a TDP of 95W, and I don't know if that's a changing value in your BIOS or just part of coded information read by the CPU-Z program.

In my over-clock settings (which top out at 25%), my TDP rises to something between 115 and 130W over a range of settings to that limit.

At a room ambient of 75F, my idle core temperature averages around 35 and load of 41C, and at 70F, those temps would be just above 32 and 38C respectively.

This is all with ducting, two 120mm case-front intake fans running at 1,800 rpm and two 92mm exhaust fans running at around 2,300 each. The CPU fan changes with temperature in increments between 1,600 and 2,400 rpm. Modest opportunities were taken to deaden sound inside the case, and the ducting is a combination of black foam-board and Lexan, so that there is an additional muffling of sound for some of it.

You haven't said what water-cooling apparatus you use, or whether you're using a thermally controlled TEC chiller. But for the range of your idle and load values, I'd say the thermal resistance of your cooling apparatus might even be greater than mine -- and I would guess that you do not have a TEC chiller. And my stress-test (I was using Prime-95 among others) seems to run the same sort of iterations as your Orthos.

But it's good for lower temperatures overall. By a few C degrees . . . but that at least replicates Cinnamon's test results using older processor, and no heatpipes with the CPU cooler. My difference over yours is maybe 3C degrees -- Cinnamon's is 9C degrees hotter than his water-cooling reference test.

 

[BonzaiDuck]

My GPU -- on the Prescott system -- runs between 47C and 58C with room ambient of 70F. Since the "red-zone" temperature is around 80C something, I've yet to complete the ducting extension to the VGA card. I suspect that changing the fan deployment on the VGA card and adding the duct-extension to cover some of the low side of the card (this is ATX configuration) -- might be worth another 5C degrees. The VGA heatpipe cooler would still show the same idle-load temperature spread.

But since I can't prove it yet to you, I agree that your water-cooled VGA at near-CPU temperatures is "out of my league" here. And my expected improvement -- unless it proves greater than my guess -- still rides some 8C or so above my CPU temperature.

 

[aigomorla]

Originally posted by: BonzaiDuck
My GPU -- on the Prescott system -- runs between 47C and 58C with room ambient of 70F. Since the "red-zone" temperature is around 80C something, I've yet to complete the ducting extension to the VGA card. I suspect that changing the fan deployment on the VGA card and adding the duct-extension to cover some of the low side of the card (this is ATX configuration) -- might be worth another 5C degrees. The VGA heatpipe cooler would still show the same idle-load temperature spread.

But since I can't prove it yet to you, I agree that your water-cooled VGA at near-CPU temperatures is "out of my league" here. And my expected improvement -- unless it proves greater than my guess -- still rides some 8C or so above my CPU temperature.

actually the last post my NVmonitor got fubard.

The voltage is actually 1.51V. So its TDP is a lot higher then it should be. Hence the high idle.

CoreTemp blows with an AMD -DFI Expert combo. Dont ask me why, but tons of people have had errors on the first core reading. The second core reading is more accurate as it correlates with RMClock, and Speedfan together.

I just threw in coretemp because everyone on this forum lives by it., and my coretemp values are still concdierably lower then most people on AIR.


And no i couldnt push her to 3.2ghz on AIR no matter how hard i tried. Im an avid overclocker. I am in a transition of jumping to TEC's, but i havent quite merrited the extra 100 dollars or so on the electricty bill. Your right in aspect that air coolign technology has increased over the times. But saying its caught up with the highest end of water kits, is very far fetched.

Oh you wanted to see the parts and stats on my loop, look at my sig below. Im the author, and all my parts pictures are nicely displayed on the sticky.


I dont think water is for everyone. You'll see me on the forums always answering watercoolign questions, and half the time i'll be questioning the reason why the op is jumping to water. And 1/4th the time i'll be telling the op to get a tuniq tower

If you have the time and resources. I say more on resources, then water is definitely an upgrade worth investing. If your a grampa or old foggie who does web surfing and looks at porn all day, water is something you wouldnt want to do.

But for me, i play games, i play supreme commander. I need my machine pushed at its extreme limit, or i got 7 computers dicking around sitting with there army causing nothing but lag.

And to mention. C2D @ 3.6ghz with SLI 7900GT @ 700mhz with 2 gigs of ram <-- my sig

Will not lag in supreme comander at 1600x1200 under high settings. But the moment i lose my 3.6ghz POOF~ i get shuttle sometimes.


AT my settings right now, ambients close to yours ~ 70F My idle temps on my cpu at 1.45 was around 27C load temps in the 38-39C.

Ummmm.. if a pump dies on my dual pump layout, the temp will increase by almost 5-6C :X Thats how imporant parts ands loop is in water. 1 wrong thing can devistate your temps. Yet 1 correct thing will push your machine to a new lvl.

 

[BonzaiDuck]

I can't make up my mind at this point. If I do it, it will be "chilled" -- using one or another of the methods I mentioned. I'll want to find some way of auto-controlling the temperatures -- especially with the otherwise "low-tech" bong.

Some others have said here -- over the last couple years -- that water-cooling is worth an additional 300 Mhz in OC settings. So I wonder how much chilled water is worth in Mhz.

 

[aigomorla]

Originally posted by: BonzaiDuck
I can't make up my mind at this point. If I do it, it will be "chilled" -- using one or another of the methods I mentioned. I'll want to find some way of auto-controlling the temperatures -- especially with the otherwise "low-tech" bong.

Some others have said here -- over the last couple years -- that water-cooling is worth an additional 300 Mhz in OC settings. So I wonder how much chilled water is worth in Mhz.

chilled water IMO gives nothing but headache. Half the time its done incorrectly, and the other half it costs more energy to maintain then a simple phase unit.

True it is a lot cheaper, an AC will run you around 119 dollars, and the other stuff is dirt cheap, however, you need to insulate everything down from your board, gfx cards, tubes and block! YES your blocks will have a nice frost forming around it without proper insulation.

Personally id put my cards on water, and jump phase if i was in your bucket. If your going to bong it, i would ditch sub ambient cooling all together.

 

[BonzaiDuck]

I won't commit to the bong project, but I may eventually do it.

It's a TEC chiller that would boost the power requirements as in your comments.

The bong chiller -- it can't use more than the cost of a 120mm fan and another waterpump.

The author of an article (circa 2002) demonstrating the bong concept at Overclockers used a single water-pump, accepting the eventual contamination of the water even with biocide added to it, and the water circulated in one loop with the shower-head and chilled-reservoir in between. That's the drawback on the more effective design of the bong.

My idea -- not based on any "thermo" calculations or data on thermal transfer between the materials -- was to grab a bunch of junkyard heater-cores, link them together in series (even with generic automotive water-hose and "bracelet-clamps"), immerse them in a beverage cooler filled with tap-water (reservoir for the other cooling loop), run in- and output hoses from the radiators to the computer water blocks, and allow the reservoir to chill the "clean-water" loop.

With the single-loop "dirty-water" project, he was able to sustain temperatures that or 10 or 15C below ambient.

So the trick here would be to find a way to regulate temperatures to bounce closely around an ambient-average. Choosing how that's to be done, given the indirect dual-cooling loop, its lower effectiveness compared to the single-loop model -- I'm not sure.

You could have a separate reservoir in the computer, and stop the water circulating from the evaporative apparatus through that reservoir when things start to chill too much. Maybe the controller would simply shut off a pump circulating from the chiller into that reservoir. This would imply using three pumps -- one between the water-blocks, radiator and internal reservoir; one circulating the water from the heater-cores; and the third circulating water in the evaporative loop. Maybe a two-pump solution would just increase or lower the flow rate between the radiator-cores and the water-blocks, if such a control mechanism could be found or devised. They may even be available at water-cooling specialty houses.

I know nothing about air-conditioning, and it's been 40 years since even so much as a Thermodynamics course, so the ideas could be unrealistic -- but maybe worth a try unless someone else could point out their folly. But the possibility -- despite filtering -- of dead insects passing through my water-blocks -- pushes me toward the two-loop idea.

In any such setup, you'll either be servicing the evaporative reservoir every couple days with a quart or two of water. Hooking it up to a depth-regulated flow from the household plumbing is also impractical.

I don't see any web-sites sporting projects with two loops. The alternative about burying pipes in the back-yard -- a closed loop -- and going that route seems too restrictive to equipment "mobility."

I'll think about it some more . . .