Cooling What to expect from Water Cooling

We get asked this question on a regular basis so thought i'd try a direct comparison between the two to find out some of the pro's and cons.

Starting off with two identical systems with the following spec...

Wavemaster Case
Lan Party UT NF3 250GB Motherboard
3000+ AMD 754 CPU (stock)
Gainward 6800Ultra (both set at 450 core - 1200 memory)
2x 512mb Ram
Tagan 480 PSU

Both PC's gave almost identical Sandra Benchmark results

Water Cooled Arithmetic Benchmark = 8383
Air Cooled Arithmetic Benchmark = 8354

The case on the right has the CPU and GPU water cooled using a Danger Den TDX and NV-68 blocks and the case on the right is totally stock air cooled and all tests where carried out with both sides of the case attached as they would normally be.





On powering both systems up the first thing that is strikingly obvious is the noise level. The water cooled system has 2x 120mm fans running very slow and quiet and is almost inaudible compared to the high speed CPU and GPU fans of the air cooled system.

Both systems were left running for an hour to come up to working idle temperature

Water Cooled CPU = 35 degrees (idle)
Water Cooled GPU = 43 degrees (idle)

Air Cooled CPU = 40 degrees (idle)
Air Cooled GPU = 53 degrees (idle)

Quite a difference considering the noise levels of both machines.

Anyway, time to start stressing both machines with a constant 15 minute run of NFS2 on each machine. Again the noise level increased on the air cooled system as the GPU fan kicked into high speed as soon as the 3d application was opened taking the noise level way over acceptable for every day use in my opinion where as the Water Cooled machine stayed practically silent.

As you can see from the results, even though the fans on the air cooled machine were running flat out it was struggling to keep temperatures down especially on the GPU.......the Water Cooled machine only increasing a couple of degrees over idle

Water Cooled CPU = 36 degrees (stressed)
Water Cooled GPU = 45 degrees (stressed)

Air Cooled CPU = 47 degrees (stressed)
Air Cooled GPU = 72 degrees (stressed)

To sum up i would have to say the water cooled system was far superior in the noise department and not having had an air cooled machine for some time now the noise level is way above what i expected and i don't think i could tolerate that day in day out especially if it were running 24/7

The temperatures speak for themselves really. The Air Cooled temperatures fluctuate quite a bit between different applications along with the noise from the fans which i found really annoying where as the temperatures of Water Cooled machine stayed much more stable which i would imagine would equate to more stable and better overclocking.

The only slight down side to Water Cooling is that you may need to periodically change the liquid maybe once or twice a year but most would argue that this is in fact a plus point and the fun of Water Cooling is the constant tinkering and the enjoyment that brings.

A good and well thought out Water Cooled system is easily as safe as an Air Cooled system and apart from the odd water change, can run trouble free, silent and stable for years........the investment is easily justified by the silence, stability and WOW factor in my opinion!

 

It's taken way too long to get a direct comparison like this. Thanks coolmeister! Even not touching on overclocking, it's a clear advantage.

 

whats with the brown tubing?

good test though, personally i run a the setup in my spec, and i would NEVER EVER EVER go back to air after my triple rad

 

This is interesting, but I have one question that will most likely decide whether I do air or water cooling...

How are these types of computers for portability? I go to LAN parties about once a week, and I haven't heard of how fun watercooled rigs are to carry around.

 

Generally heavier, but as long as you tighten everything properly, nothing should come loose. My comp is huge and pretty heavy so it's not overly transportable, but I'm not more paranoid about moving it because it's watercooled.

 

Interesting, but is that a fair comparison?.

"Stock" air cooling?
Put on a decent CPU HSF and an aftermarket video card cooler and the differences would not be nearly as dramatic.

For instance, in my case, switching from an AC Freezer64 (which is nearly inaudible) and a Zalman VF-700cu to a DD TDX/NV-68 waterloop resulted in iidentical CPU temps, GPU temps lower by 5C, but higher northbridge and PWM temps.
I assume the latter two components suffer from lack of secondary cooling previously provided by the heatsink fans.

This not-so-dramatic improvement has lead me to question whether H2O is really worth it.
Working on the PC has become considerably more involved...it's no longer possible to just pull the vid card and remove the motherboard without completely draining and dismantling the waterloop.

I am seriously considering moving Sprocket out of her watercooled TJ-06 and into a straight aircooled case (either a Stacker-again!- or a Lian-li V-series) and seeing if the performance is noticably degraded.

Watercooling is still perversely attractive to me, but my experience does not show a great benefit in either cooling or silence when using it.

 

How come this is the case, are your lenghts too short? Whenever I have to remove a part like that I just unclamp the waterblock and move it out of the way without ever disconnecting the water lines.

For the topic in general the biggest plus to water cooling for me is the noise reduction. With a well designed WC system it is very easy to only have one or two near silent fans in high performance system, zero fans in a system designed for it. Combine that with a near silent pump like an Eheim and you're back on the track to sanity.

 

The CPU block is not the problem...the GPU is.
Mounting the DD NV-68 involves four screws ( with the foam insulator and backplate to align also) and is no cakewalk.
It could be removed in situ I suppose, but not properly reinstalled.

 

I decided to take this one step further as it was pointed out that there are better air cooling solutions on the market than those that come as stock on the CPU and GFX card.

I got my hands on what i'm told is one of the best CPU coolers the Arctic Cooling Freezer 64 and for GFX the latest offering from Zalman the Zalman VF700-CU Quiet Copper VGA Cooler.

I also purchased the Zalman ZM-NB47J Northbridge Cooler but this wouldn't fit due to the placement of the chipset on the DFI board so it was left with the stock heat sink.





The Zalman GFX Cooler came with the option to run the fan at either 12volt or 5volt so eager to see how well this would cool the card over the stock Gainward cooling i decided to run it at 12volt so left the system to idle for one an hour as on the previous test.

What i will say is that this cooler is no quieter than the Gainward stock cooler and given that the fan runs full speed all the time whether in 3d application or not as its linked straight to a Molex connector and doesn't give you the choice to run slower when just idling so its flat out all the time.

I'm not able to tell if the CPU Freezer 64 is any quieter than the stock HSF as it can't be heard over the Zalman.........inaudible this certainly isn't!

Anyway, after an hour of running idle the following results were taken.

Air Cooled CPU with Arctic Cooling Freezer 64 = 40 degrees (idle)
Air Cooled GPU with Zalman VF700-CU @ 12volt = 52 degrees (idle)

Now then........i've heard good reports on these two coolers but from my own experience the Arctic Freezer 64 is giving the same results as the stock HSF and the Zalman VF700-CU is showing 1 degree better than the stock Gainward cooler and that is with the fan running flat out at 12volt which is anything but silent.

As before i ran NFS2 for 15 minutes to stress both CPU and GPU with the following results...

Air Cooled CPU with Arctic Cooling Freezer 64 = 46 degrees (stressed)
Air Cooled GPU with Zalman VF700-CU @ 12volt = 63 degrees (stressed)

CPU only marginally better over stock with a 1 degree decrease but this time the GPU was much better with a 9 degree decrease over stock.

And finally i tried the Zalman VF700-CU at 5volt which again although slightly quieter than running at 12volt, i still can't hear the Arctic Cooling Freezer 64, so even with two of apparently best air cooling solutions they fail in both departments against the water cooled Wavemaster.

i think the following figures were inevitable TBH

Air Cooled GPU with Zalman VF700-CU @ 5volt = 55 degrees (idle)
Air Cooled GPU with Zalman VF700-CU @ 5volt = 70 degrees (stressed)

I had high hopes with both these coolers so to say i'm disappointed with the results in both noise and cooling ability is an understatement.

Luckily i have been in a position to try all of these coolers first hand and share the results so personally it has to be said the water cooling is by far the superior option in my experience.

The air cooled Wavemaster has been running for some hours now and going by noise alone, i honestly could tell whether it was the stock cooling or the new cooling.........all i know is its getting switched off.

 

Thanks, that was a much better comparison, kudos for making the effort.

I was using the exact same combination of heatsinks prior to my latest switch to H2O.
Our results are nearly identical for the vid card but way different for the CPU.
I was averaging low twenties (ambient in the teens) and maxxing out @32C.

The only significant difference between our installations would be that my case is pseudo-BTX (SS TJ-06)...whether that would explain it I can't say.

I'm preparing to revert to aircooling (albeit into a totally different case) and, following your format here, will post a report of my findings.

 

nice comparison tho
<3 my watercooling

 

Be interested to know what CPU you are running as i've spent a lot of time going through various reviews on the Arctic 64 and i've yet to come across one that gives such good results as at best gave a few degrees drop over stock and it would seem the temperatures i got where just about spot on compared to independent reviews.

The CPU i used is a 3000+ AMD 754 CPU (stock) which already run pretty cool in comparison to other CPU's

Also what are you using to measure temps and also what were your watercooling temps??


On the CPU i am finding a 5 degree deficit between water and the Arctic 64 at idle and a 10 degree deficit between water and the Artic 64 while stressed.

Air Cooled CPU with Arctic Cooling Freezer 64 = 40 degrees (idle)
Water Cooled CPU = 35 degrees (idle)

Air Cooled CPU with Arctic Cooling Freezer 64 = 46 degrees (stressed)
Water Cooled CPU = 36 degrees (stressed)


And rather than getting a 5 degree drop over the NV-68 as you state previously using a Zalman VF-CU, i'm seeing a 9 degree drop idle over the Zalman and an 18 degree drop stressed.

Air Cooled GPU with Zalman VF700-CU @ 12volt = 52 degrees (idle)
Water Cooled GPU = 43 degrees (idle)

Air Cooled GPU with Zalman VF700-CU @ 12volt = 63 degrees (stressed)
Water Cooled GPU = 45 degrees (stressed)

I note you edited your previous post from Interesting but flawed to Interesting, but is that a fair comparison?.

Well i've tried both now but still fail to see how air can match water in both the noise and performance department!

 

Thanks for all the hard work (not to mention money) involved in conducting this important experiment, Coolmiester!

Thread stickified!

 

Winchester 3500+ 939

My 3200+ 754 tended to run somewhat hotter than the 90nm Winnie, but worlds better than the old Athlons.

I simply use the monitor apps provided by the motherboard maker.
I can't claim any accuracy for the temps, but they are useful for relative comparison on the same equipment.
Here is a fairly typical watercooling temp...

My chip is never run at stock speed, daily use sees it set to 2.5GHZ and benchmark runs go up 2.8GHz.

By comparison ( and admittedly, not a fair one) here is the same chip on my previous Gigabyte board while aircooled....


Since moving on to the DFI mobo, Sprocket has always been watercooled so I have no direct air-water comparisons yet.
I hope to change that soon.

Indeed, I did edit my original post...it seemed snotty and judgemental as first written and that was not my intent.
I do appreciate your effort, especially the effort to level the playing field with better aircooling parts.

Unfortunately, any comparison I perform will be inherently flawed as I don't have a duplicate Silverstone case to build into and it would take a major effort to convert my present case back to straight air.
Seriously considering moving back into a Stacker, which should allow for both air and water installs without hacking the case to pieces, but that must wait till the end of the month.

Keep up the good work.

 

Cheers Nexxo, i guess it is a good start for anyone considering water cooling over air however originally i just thought it would be quite an interesting experiment even though it wasn't carried out with any specialist equipment that some might own but as clocker mentioned, using programs such as Sandra on both machines does give a relevant and relatively fair comparison and are available to everyone.

Just out of curiosity clocker it would be great if you get a chance to d/l Sisoft Sandra SR2.a and see what your CPU temp is with that which presumably would be relevant to my findings.

It would be even better to get bigz (or someone equally impartial) to put both machines through their paces in a more in depth and professional manner than my admittedly pretty basic test.

If anyone is up to the task then they are more than welcome to come and do as much testing as they want as i'm starting to think this could be quite a significant and topical subject.

 

Right then.

I've been preparing to convert back to air cooling, but hit a snag.
The new PSU's (Seasonic S12 600w) 24-pin cable is too short to reach the motherboard header when installed in my TJ-06.
I don't particularly want to hack either the cables or the case to make it fit, so I'll be unable to use the same case for the conversion.

Which started me thinking.

Can this comparison be considered valid if I use two different cases*, but use the same basic components in each?
After all, if I knew beforehand I was going to water cool, the criteria I would apply choosing the case would be different than for air cooling.

For instance, the Wavemaster is not particularly well known for it's airflow, but would be fine for watercooling (as you've so amply demonstrated, Coolmiester), so I would consider the CM case only if watercooling was planned from the outset.

The CM Stacker is the obvious choice to handle both types of installation, but having just spent $150 on the new PSU I'm not sure the budget allows for a new case this month.

If you guys will accept results from two different cases as pertinent to this thread I'd love to participate, otherwise I'll bow out.

Furthermore, the question of "noise" must be addressed somehow.
Any judgements about how pervasive/objectionable the end result is are totally subjective...my idea of "silent" or "acceptable" may not correlate at all with someone else's.
Short of submitting to a full SPCR-style lab test, how shall we quantify this fairly?

*Edit:
Looking again at coolermiester's two cases, it occurs to me that the aircooled version might benefit from the same twin fan roof blowhole that is present on the watercooled version, so to be rigorously fair those two systems are not in "identical" cases anyway.

Edit, edit:
OK.
Stacker is in hand.
All aircooling parts are in place save the GPU unit.
Temporarily using a Zalman similar to the one Coolmiester mounted...mainly cause I had it, although I was not impressed the first time it was used.

As I expected, the CPU temp is nearly identical to the watercooled result.
The GPU is much worse, but that should change when the StasisThermal cooler arrives.

For the initial build I used all the fans supplied with the case and CPU cooler to establish a baseline performance level.
This too will be changing as I have a good supply of tri-blade 120x38mm Nidecs that I plan on using.

Apparently both CoolerMaster and Thermaltake value silence over performance as all of their fans are whisper quiet, but not moving much air.
Sprocket is dramatically quieter than when she was watercooled, but that will certainly change as I swap out fans and ramp up performance.
Gone ( thank God!) is the terribly annoying whine from the Polarflo TT pump.
I didn't realize how pervasive that was till now...if/when I go back to water that pump is history, for sure.

Happily, the Stacker will easily accomodate my Weapon heatercore/shroud (well, the shroud is a few mm too wide, but I'll deal), so converting back to water to complete the comparison will be relatively painless.

Not so painless will be dealing with the wiring on the Seasonic PSU.
I was going to post some pics of the new build, but the wiring is such a rat's nest that I was embarrased.
All of the cable bundles are twisted ( supposedly for EMI shielding or something), but they did a half-assed job of it and it looks terrible.
Plus, the twisting puts the connectors out of sync...the SATA connectors are aligned differently from each other on the string and I had to improvise to get everything hooked up.
Yuck.

Anyway, we'll see how this turns out.

 

Well, to get even records you two should use identical boards (same revision, same BIOS) as manufacturers tend to calibrate their temperature readings differently.

It has been reported, that at the best, single BIOS update would lower CPU temperatures over 10 degrees. Now, that simply is not possible, only way that can be achieved is by recalibrating the temperature readout.
It seems that in the war of who's-got-coolest-(as-in-temp)-board, everything is allowed.

Other way would be using completely different measurement, i.e. by specific thermometer, placing the sensor to same exact position.

By the way, watercooling is out for me, as air cooled solution is much more care free. I don't want to be changing anything periodic, unless it's broken.

 

We are not comparing the two systems to each other at all.
Coolmiester's twin Wavemasters are identical in hardware spec, he is comparing the cooling solutions 'tween the two.

I am going to be using one system and just swapping out the cooling on it.

The results might be simplistic, but both methods address the topic title ("What to expect from watercooling") in a generally consistent way.

 

Fair enough. erm.. Carry on.

 

A couple of graphs to make things more easy on the eye and gray matter...

First CPU


...and then GPU