H50 not cutting it? Don't hack it! Chill it!

[Rubycon]

As promised here is a quick run down of my results of cooling down the H50 radiator. Unfortunately I was VERY short on time so no playing around. There was just enough time to set this up, take pix and break it down.

The thought of leaving water and ice cubes on the bench under way next to racks of expensive equipment is frightening to say the least!

Ok enough about that off to the goods!

First here's idle temps with a single fan on the H50 Radiator:

Loading up with some Linpack action after a few min:

Idle iced down:

Under load under ice:

Actual pix of the hardware:

Some other (important) facts to consider:

Hardware used:

Dawn's "beater" system!
Asus P6T7WS "supercomputer" motherboard w/latest BIOS 604
12GB Corsair Dominator 1866 memory at 8-8-8-20-1T 1.635V
Intel Xeon W3520 CPU at 208x21, 1.3V 1.25VTT
Intel X25M G2 160GB SSD
EVGA GTX 280 1GB Graphics card
Enermax EVO Galaxy 1250W PSU
Corsair H50 (Duh!) w/Nidec thermistor fan (air) and Coleman playmate style cooler with 3kg ice + 950ml water at 1.040 sg "kosher" salt crystals. Fluid temp measured at +25.5°F
Radiator hoses were cold enough to sweat - the Miami air had a dewpoint of +73°F so no surprise there. IR thermometer showed these hoses at 67°F FROM cooler and 62°F FROM radiator.
The load speeds in Realtemp are SLOWER because of this stupid thing called turbo throttle. It kicks the multiplier back to 20X (from 21X) very quickly dropping CPU speed. As you can see the speed was higher when the radiator was in the ice so it is thermally triggered. A better test would have been with my Classified since it does not exhibit this silly and most undesired behavior! Temps would've been considerably HIGHER especially without ice because the CPU would have been running at 21X208 the entire time. Easy mid 90s possible.

In any case you can see that bringing the temperature of the radiator down can significantly reduce operating temperatures. (no surprises there)

 

[daw123]

How is that so different to running it under LN2 or dry ice?

I'm not trying to cast dispersions on your post Rubycon, but I honestly can't see a difference in terms of upkeep; there is very little difference between having a vessel planted on top of the CPU IHS, filled with LN2 or dry ice with the MB insulated, and the separate cooler you showed filled with ice.

Or am I missing an obvious point you are making?

 

[SonicIce]

Thats awesome. Next, a mini fridge.

 

[Rubycon]

The point is to show cooling the radiator down to a temperature that would not normally be possible or practical can yield an improvement in performance. Honestly I thought it would be better.

It's a LOT different than LN2/Dry ice simply because NO prep on the motherboard is required. Just removing the fan from the radiator and dropping it in a bucket of ice. Easy as falling off a log backward.

Not to mention LN2 and dry ice present a hazard if not handle properly particularly the former. I can get dry ice but not to use in a pot simply because adding alcohol or acetone would create a hazardous condition that violates all kinds of rules here.

Bottom line it was for fun, I mentioned that I would give it a whirl months ago (the copper plate on the bottom of my H50 is actually tarnished from sitting and needed to be polished!). I have not seen anyone else try it as well. So there you go!

 

[MJinZ]

You have a bad unit or improper mounting.

My RealTemp 3.4 Core idles are low 30s, sometimes even 29s. Overclocked to 4.2ghz, it idles in the high 30s.

I've had 4x H50s so I would know, some of them have bad pumps.

 

[daw123]

The point is to show cooling the radiator down to a temperature that would not normally be possible or practical can yield an improvement in performance. Honestly I thought it would be better.

It's a LOT different than LN2/Dry ice simply because NO prep on the motherboard is required. Just removing the fan from the radiator and dropping it in a bucket of ice. Easy as falling off a log backward.

Not to mention LN2 and dry ice present a hazard if not handle properly particularly the former. I can get dry ice but not to use in a pot simply because adding alcohol or acetone would create a hazardous condition that violates all kinds of rules here.

Bottom line it was for fun, I mentioned that I would give it a whirl months ago (the copper plate on the bottom of my H50 is actually tarnished from sitting and needed to be polished!). I have not seen anyone else try it as well. So there you go!

Sorry. I was looking at it from a maintenance point of view; I saw the bucket full of ice and thought, how the hell can that be easier than LN2 or dry ice with keeping a CPU continuously cooled. I know that condensation is a big problem with exotic cooling solutions; I wonder if Aerogel would help. When I get more $$$, I may I'll give it a try... What do you think because let's face it, you are a lot more technically minded than I am?

 

[Rubycon]

You have a bad unit or improper mounting.

My RealTemp 3.4 Core idles are low 30s, sometimes even 29s. Overclocked to 4.2ghz, it idles in the high 30s.

I've had 4x H50s so I would know, some of them have bad pumps.

What's your ambient?

When I tested this before I did have lower idle temps but loaded temps were still too high. I've had more than four of these units pass through my hands. The first one was defective and the rest have been like this.

Loaded temps were spot on as before so I know the unit is working fine. Mounting when removed yielded a perfect patch.

Additionally unless you have the CPU temps comparisons are invalid!
I've had crazy differences with the same model CPU!
Don't even get me started on 980X temps! The system I'm posting from says it's idling at 14C!

 

[aigomorla]

i really thought it would be better too.

i guess the fins in the ice bucket isnt working out as well as we thought it would.

Can u have another pump inside shooting water into the radiator

 

[Rubycon]

i really thought it would be better too.

i guess the fins in the ice bucket isnt working out as well as we thought it would.

Can u have another pump inside shooting water into the radiator

The tubing stayed pretty cold.
I suppose a small powerhead in the cooler pushing water around would enhance it. Probably the biggest difference would be a larger cooler so the entire radiator was completely submersed!

Here you go - for those that sleep at night have a chiller make ice while your computer sleeps then the ice keeps it cool. Silly I know. (if you had a chiller you would just use that on a real water block!) And if a frog had wings he would not have to bump his ass on a lilypad - just saying.

 

[Corporate Thug]

cool mod (pun intended). Too bad it's not a lot term solution.

 

[slayernine]

Very insightful on how to get lower temps. One big difference I realized with this cooler with if you have two fans on it doing intake you get lower temps than two fans doing exhaust. This is because you are pushing cooler air through the radiator.

 

[n7]

Ah, you did try this!

Interesting results.

 

[Zap]

Gotta keep "kosher" right? Is that like black computers being faster?

 

[Rubycon]

Gotta keep "kosher" right? Is that like black computers being faster?

Haha yes indeed!
I asked for salt and was asked if kosher salt would work.
I'd rather do that than get water from Biscayne Bay. :twisted:

BTW if you REALLY want to see temps jump just unplug the pump! It won't last long though.

 

[HybridSquirrel]

bumping because the pic links are dead

 

[Rubycon]

I can see them and I checked to make sure they are not cached via star peer.

 

[HybridSquirrel]

I can see them and I checked to make sure they are not cached via star peer.

hmm...red x for me. ill try on a different computer. was interested when you posted this in the other thread


edit: ahh...they are from bbzzdd, which is blocked at my school for some reason.


tag for when i get home then =D

 

[deimos3428]

It'd be interesting to see results with room temperature water for comparison. I'd suspect a significant portion of the delta between the results is due to the change in medium rather than the chilling itself.

 

[MagickMan]

That's what I was thinking too. Hell, if I could get near that performance with room temp water I'll just hang my H50 out of my case in a bucket.

 

[WoodButcher]

It'd be interesting to see results with room temperature water for comparison. I'd suspect a significant portion of the delta between the results is due to the change in medium rather than the chilling itself.

That's what I was thinking too. Hell, if I could get near that performance with room temp water I'll just hang my H50 out of my case in a bucket.

Over time the bucket water would heat up, needs to be chilled.

 

[deimos3428]

Over time the bucket water would heat up, needs to be chilled.

Over quite a long period of time, as that's a huge reservoir. Four kilograms of water is simply not going to get warm very quickly. You'd need slightly more than 16 kilojoules per degree C, not counting the significant constant heat loss to the ambient air. Natural convection should generate sufficient motion to distribute the heat fairly evenly.

As for maintenance, consider that for most people it will be much easier to find a source of cool water than to go to the trouble of creating ice or constantly chilling the same water.

 

[yh125d]

Now rig it up with longer lines and a stronger pump (to make up for the longer travel), and route the lines into a rube goldberg contraption that consists of an icemaker dumping ice into a cooler which opens at predetermined intervals

 

[HybridSquirrel]

It'd be interesting to see results with room temperature water for comparison. I'd suspect a significant portion of the delta between the results is due to the change in medium rather than the chilling itself.

do you play wow for seaghyn? offtopic.


anyway, why is condensation a problem? would the lines inside the pc condence? or were you concerned about water on the desk that could spill over into your over fancy stuff? i had always wondered if this would actually work.

 

[deimos3428]

do you play wow for seaghyn? offtopic.

Nope. (Maybe I'm old or something; I had to google "seaghyn".)

 

[Juddog]

Over in the xtreamsystems.org forums they do something similar, except on a more permanent basis, using a "beer chiller". The downside seems to be that condensation will form over parts that you cool, so if you get into the super low temps you may need to seal it. Seems like a nice idea though to add into a loop.