You know what'd be neat?

[Xel]

A waterblock that had a fan on it just incase your pump fails (with a sensor to turn on when water temp gets too high in the block.

I think it could be neat if done right.

[Brad]

just make a block with a side inlet and side outlet and use AS epoxy to mount a normal hsf on there

[DigitalChaos]

damn.. i was thinkin about riggiing a few blocks like this up.. i got LOTS of block ideas... but they are slowly becommin prototypes

[Kevin]

This is actually not a good idea as it reduces waterblock efficiency. Before the heat ever goes into the water it goes to the copper. Then the copper heats up the water. By putting a hsf on the waterblock, it attracts heat away from the water and into the heatsink, thus taking efficiency away. OCH made a block based on the "pump failure" idea, and it was a horrible performer both due to adding the hsfs and just an overall shitty design. Xel, I like your idea though. I think a circuit attached to the power on/off button wires would be much better. Like to detect that if the pump no longer pumps, to shut down the PC. Surely in a quality watercooling system, if the pump fails, the PC has a few seconds before death...
-Kev
Here's maybe a better explanation. Think of how hot the top of your waterblock gets. Now stick a heatsink on it. Think of how hot that heatsink would get. Where would it send the heat? It wouldn't... It would just sit there on the heatsink and heat the block up more than need be.

[Brad]

kev, what you are describing is true, but is very minimal at best, I honestly think you've exaggerated there

[Kevin]

Originally posted by Brad
kev, what you are describing is true, but is very minimal at best, I honestly think you've exaggerated there

No. It is quite severe. Copper is an excellent material for transferring heat. All heat that is reasonably transferred to the copper base of the waterblock transfers all throughout the copper block. Putting a large copper/aluminum block on the top of the copper waterblock would be like putting another CPU on top of the waterblock for it to cool.
-Kevin

[Brad]

lets get Joe to test this then

[Leuf]

Originally posted by Kevin

No. It is quite severe. Copper is an excellent material for transferring heat. All heat that is reasonably transferred to the copper base of the waterblock transfers all throughout the copper block. Putting a large copper/aluminum block on the top of the copper waterblock would be like putting another CPU on top of the waterblock for it to cool.
-Kevin

No, it's not another heat source it's a heat mass. What it will do is make the system slower to react to changes in cpu load. Ie, it will take longer to go from the idle temp to the load temp and vice versa. Depending on the test, you may see a higher average temperature because of that, but you shouldn't see much change in the maximum temp.

The question is, how useful would it be? If you are using watercooling to overclock your cpu to the max, then presumably the hsf couldn't handle the load even if it were on the core, much less hanging off a waterblock. It would buy you a little more time before the inevitable crash, but that's it. If you are using watercooling to reduce noise with limited overclocking, then it might work, though it wouldn't be much fun to epoxy that hsf you just did all that work to get rid of back in your box

[Volenti]

The only time I can see this detracting from cooling performance is when you're using sub-ambient coolant, or mabie if you have a high internal case temp.

you could use one of the small high performance thin fin copper heatsinks with a delta screamer. You can get normally open thermal switches that close at a certain temps, imbed one into the block and you have automatic thermal protection if the pump dies, sure it may run hot, but if you have set the shut down temps in the bios it will have a pretty long time to shut down if necessary.

[Brad]

but if your pump just dies, your temps might climb so rapidly that your software can't react to turn your comp off before it crashes.

whereas with a hsf on there, it should have time to react, and shut off before it crashes

[Cyph3r]

Now i don't have any experience first hand. But i've heard that ppl have had pump failures and it didn't kill their sys... It was hot... very hot... like 55c but not dead.... I dunno.... I was planning on rigging a DigiDoc 5 into a relay that'd cut the power on top of having mobo monitor shut down. Some systems u can even set a thermal shutdown in the bios.

I asked a Q about this in the hard forums cause i was wanting to know if I needed dual pumps for redundancy.

Cyph3r

[pHaestus]

Originally posted by Brad
but if your pump just dies, your temps might climb so rapidly that your software can't react to turn your comp off before it crashes.

I would be more concerned that the system's OS would freeze and MBM wouldn't be able to shut things off. If you were conservative enough in how you set it up tho (60C?) then you can probably get away with this.

There are better alternatives. The Maxim ICs I have for my diode readers have some interesting functionality that might prove useful. They have Overtemp functions, sometimes 1 and sometimes 2 that can be associated with one of the remote temperature readings. So for example you could assume that with normal watercooling that for the internal diode of your CPU to read 60C the cause was definitely pump failure and have the first overtemp turn on a 2nd pump via a circuit. Then if the temp continued to rise then you could have the 2nd overtemp shut the system down.

I think this is all a bit of overengineering personally though. If one of my pumps fails then I want to replace it and not run on a backup

[gmat]

I had a pinched tube some time ago. Happened when i closed my case, a tube got stuck.. I didnt notice, and used my comp normally for about 30mn.
Windows crashed with a BSOD, then i could hear strange noises coming from the waterblock. Those noises water does when it's halfway to boil.
I immediately shutted down the comp and found out about the pinched tube immediately. The WB was *very* hot, though it wasnt burning and neither the CPU nor the thermal paste did suffer. I just replaced the tube, let the system cool down and all restarted finely.
My WB is a DangerDen Maze2 and i have 1/2" tubing, and my system is totally closed (no res, no air trap). My CPU is a Tbird 1000/AXIA @ 1487.
For one the thermal inertia of a filled system seems quite high, and for two the DDM2 is big enough to sustain such a heat alert.
Couldn't say the same for pelt systems though... those could butn out quite quickly.

[Kevin]

Leuf - you are very right. That's what I ment to say. Another heat MASS.

Anyhow bottom line is that I think pHaestus hit the nail on the head. Just set a conservative auto shutdown temp. If you are H2O cooling, it shouldn't get too high anywyas.

And the pump dieing in an H2O system is not quite as severe as the fan dieing on a heatsink. It takes longer for it to overheat. I remember when I first watercooled my Celeron 300A back in the day, I unplugged the pump while it was on and watched the BIOS temps escalate... took a bit... I had time to watch the screen for a while and plug the pump back in. Granted, that CPU doesn't exactly put out as much heat as a 2 ghz beast.

-Kev

[pHaestus]

I suppose there are caveats. If you are direct die cooling or using something like the nanoblock (if you don't already have one I guess you aren't getting one) then you don't have much buffer there like with larger copper blocks. In those cases, the pump failure might quickly kill a CPU. If you are doing something extreme like that then you might want to make a flow sensor like this:

http://www.overclockers.com/tips642/

Then connect it to a NO relay (normally open) that is suitably low in power (they make 5V relays I am not sure about 3.3). Have the relay turn off the power_good line of the CPU in the event flow stops. Might take a bit of hunting for the appropriate relay but other than that it should be cake.

[DigitalChaos]

ok if adding a HSF to the other side of the block adds heat MASS then why didnt everyone use those minature vtuned blocks that were only the size of the core????

also the only way for a HSF to add heat is if you are cooling below ambient. If you are not using active cooling, a HSF would actually increase performance because it would passively cool a block that is slightly warmer than ambient!

As for it slowing down the overheating: a heatsink PLUS the fan would be designed to stop the overheating... if you wanted to slow it down you could simply put a big chunk of metal on it.

[DigitalChaos]

also.. if you use the average block like a Maze2 the temps do take some time to rize... the only way your computer would lock up and not be able to stop it in time is if the water block came off.

Test it out. unplug your pump and watch the temps climb, it takes quite a bit for the temps to go up. also a good temp for shutdown in a water cooled system would be 50C being that this is a very safe temp and even if your system was heating up fast, you would be able to do something before it locked up.

[gmat]

Usually tha waterblock surface is 1°C or 2°C above *radiator* ambient. If not, water doesnt remove heat efficiently enough. (just checked my Maze2 surface - its ok )
And usually the *inside* of the case (when closed) is about 4-5°C above *radiator* abient. But putting a HSF on your waterblock you're just adding heat to it.
That, unless you've got a 'tornado' kind of case with zounds of fans... and what's the point of watercooling your comp if you're about to put extra fans just for it...?
The N-O relay is quite a smart solution IMHO.

[DigitalChaos]

good point.. case temp will probably be above water temp so that would be bad...
yea my personal preference for thermal protection is a DD5 with a relay on the PowerGood

[Kevin]

Yup yup... putting a heatsink on is not the best solution. The best is good protection via auto shutoff when the pump fails, however you want to wire that is up to you.

[DigitalChaos]

if you dont have a digital doc there is a circuit you could build.. ill see if i can find that thing. but in the end its about the same cost as an older digital doc.

[GigaFrog]

Pump failure is a problem is you are peltier cooled. With flat water cooling, I would think it will take time for the water block to heat up to dangerous levels. A hardware protection is good for flat water cooling. It is necessary for peltier cooling (this is my opinion).

The problem of software protection is that as pH said, the system freezes before it reaches the shutdown temp. Conservative settings solve that, but we may not want conservative settings.
If I want to be real conservative, I'll go buy a Dell, its guarenteed, and I don't have to worry (OK, this is ultra conservative!).

There are different ways to protect against water pump failure with simple hardware mods. There is even a very good one describe in this website, that directly reads the temp of the CPU core (it evens protects for uncorrectly sitted water blocks or heatsinks).

The water flow sensor looks very cool, but it is one more risk of leaks. Plus it looks a little complex.

A thermal protection, like any protection, must be as simple as possible, to prevent its own failure that would lead either to false alarms, or not be able to react to failure.

As far as over engineering, I personnaly think water cooling is at the base already over engineered (but it is a cool hobby!).

Just my 2 cents.

[Fixittt]

my machine ran for close to 30 minutes before my temp alarm went off, my ferret had unplugged the pump. I have my temp alarm set for a very low temp, it hits 45C and things start shutting down.


Isnt the reactor setup with 2 heatsinks and fans, and we found that wasnt a very good setup.

[DigitalChaos]

LOL your ferret is jealous of your computer from all the posts i have read

[Cool 16]

Why not just buy a reliable pump? Get a Eheim and wire it with a relay or something so there is no chance of having it unplugged. Has anyone heard of a Eheim pump dying???