Odd idea, looking for pros/cons

Cova · 07-23-2002, 03:43 PM

I'm not sure why, but I've had this idea in my head now for a few weeks, and I'm trying to decide if it's as good as my brain wants me to think it is (I just can't forget it), or if there's something fundamentally wrong. So I'm looking for a list of pros and cons over such a system compared to a regular pelt setup (cold side on CPU with cold-plate, hot side on CPU WB).

Pros:
- If pelt fails, CPU will not fry even with extended use
- Condensation will either always be a problem regardless of CPU load (big pelt) or never be a problem (smaller pelt)

Cons:
- CPU won't be as cold as direct pelt-cooled

The pelt I'm considering using is an old one I have from ... I don't remember, it's only about a 70W one though - not enough to direct-cool a hot chip. I'm also thinking about only running it at 5V (it won't move too much heat, but it will be more efficient with the heat it does move, putting less waste energy into the system to do it).

The theory is to move heat/energy around in the system without moving around the coolant. By doing this the water in the block should be cooler (block more efficient), and the water in the rad should be hotter (rad more efficient), and add a minimal amount of complexity and excess energy into the system in the process.

But I can't get over the feeling that I'm missing something.

bigben2k · 07-23-2002, 03:51 PM

Instinct says yes, but theory says no.

The pelt creates its own heat, from usage, which would increase the load on the total rig.

But regardless, some guy tried it (see OC), and basically, he found that the cold part wouldn't get very cold at all: it's hard to pull heat from a liquid passing by. He tried it with three pelt, in some large sandwich type assembly, using seperate loops to heat/cool.

airspirit · 07-23-2002, 04:00 PM

The only way to make that work is to have two separate self contained water cooling units. Unit one is chilled by the cold side of the pelt and directly cools the CPU. Unit two cools the hot side o' the pelt. Caveats: Unit one should not have a radiator or the radiator would actually HEAT the water, if the water gets chilled below ambient. Further, unit two should, save the water blocks, be isolated completely from unit one to prevent bleeding.

This is why it isn't a good failsafe system. If the pelts fail, you will now have no cooling except for the water on the CPU. Once that water reaches a certain temp, either the hoses will burst from increased temp or the CPU will fry. Either way you're screwed.

In order to make it truly failsafe, you need a radiator in Unit 1 in case of pelt failure. Unfortunately, you're now cooling your room, but it'll give you slightly lower temps than straight water. Your rad will become a condensation magnet, though, so isolate it from your comp. For that matter, all hoses and EVERYTHING in Unit 1 will need insulated for that same reason. Talk about a PITA.

Best case is with the rad inline, though it is much less efficient than direct pelt cooling. The benefit is that your system should never fry unless you lose both the pelts and the pump.

Without the rad in Unit 1, you will essentially have a larger time window before your CPU fries if you lose your pelts.

Does that make sense? I'm on dental drugs right now, so I'm not totally with it.

Cova · 07-23-2002, 04:07 PM

I've got a whole 'nother set of ideas I want to experiment with at home airspirit that much more closely fits your 2-loop setup. Thats essentially a water-chiller setup. I've seen a lot of the dual-loop pelt-chiller setups - this is different and new. If you want a water-chiller setup you're better off running phase-change to cool the water (as in #rotors dehumidifiers).

I know pelts create their own heat, as well as moving heat from side A to side B. I also know that pelts run more efficiently at lower voltages (create less of their own heat) while moving less heat from A to B. The theory in my above picture is to increase the efficiency of both the rad and the block (and I'm sure that it would), and if I'm right, that efficiency increase will be enough to offset the heat that the pelt creates in making them more efficient.

bigben2k · 07-23-2002, 04:29 PM

Ok, so then your idea is good, but you'll have to figure out what Kev did wrong here

Maybe a cross drilled block on the cold side would have been more efficient.

ambient · 07-23-2002, 04:32 PM

I think chip from overclock-watercool used to sell a sandwitch pelt water chiller. I don't remember if anyone ever used it. Maybe Joe, Una or Brad would know?

Cova · 07-23-2002, 04:46 PM

I'm not sure what kev did wrong in that article. There is a site in the ProCooling links section - www.frostyfox.com (don't go there now, site is dead and you'll get popups from hell I just found out) of a guy that built a pelt-powered water-chiller, but using 6 pelts at 12V. With an additional 2 pelts right on the CPU he was running in the -60C range with no load if I remember correctly. In kev's article it looks like he did 2 experiements, one with 2 buckets of water, and one with his sink's - in the buckets of water experiment it appeared to work, and in the sinks it didn't. I think he likely made a mistake in one of them and didn't realize it at the time (eg. the chiller block had some air caught in it).

Either way - I'm not building a 2-loop water-chiller here. As I said above if you want to chill water phase-change is a much more efficient way to do it. I want to use a single water-cooling loop, and potentially make that single loop more efficient through the use of a pelt in a different position than I've seen before.

bigben2k · 07-23-2002, 04:49 PM

I wish I could take a look at that chiller.

Either way, a maximum surface area (like that provided with a cross drilled design) should allow a fair amount of heat to be pulled from the water.

It's really a good, innovative idea!

JimS · 07-23-2002, 06:31 PM

The chiller block kev used was a simple "U" shape inside the block which yielded very little surface area. My system has a similar setup but is more like what airspirit's design describes with two separate loops, the cold side has no radiator. Using the same pelts kevin used ,5V pelts pumping 50W (x2)I can keep my CPU right around ambient with this setup. The water is below ambient, but I rarely get any condensation and I have critical points insulated.

I am redesigning my blocks now because it now looks like my flow is not even. I will try to remember to take some snapshots as I go and will publish the finished product in a thread here for all to see.

My goal has never been subzero cooling, just wanted to keep the CPU and NB at ambient or thereabouts. I could get some bigass power supplies and huge pelts to get cooler, but not worth it. These 5V pelts run fine off of a standard PC power supply and work well for this purpose. My electric bill went up $4/month when I put it in use.

Cova, the flaw in your design above is that you are using the one loop. No way water can change temp. that fast.

For these chiller blocks to work, they need to have a large amount of surface area exposed to the water in the cold loop. If setup properly, they can be effective.

Cova · 07-23-2002, 07:20 PM

Hmm..., that likely would be a problem - without really long blocks on either side of that pelt the water likely wouldn't change temp by enough to make much difference. Hmm..., perhaps a solution to this will pop into my head while I'm trying to solve some other problem (thats how I got the original idea). The goal at the end is still to have a single cooling loop, but make it more efficient.

airspirit · 07-26-2002, 06:12 PM

What was done wrong in the article is in no experiment was the hot side of the pelt isolated from the system. In all of them there was an avenue for the heat to bleed back in some way. If the hot side of the pelt is isolated from the main CPU loop and sufficiently cooled, you will have superior results to watercooling. I guarantee you will have better results with standard pelt cooling ... plus, you won't have to worry about your water lines freezing or collecting huge amounts of condensation (zappo).

07-23-2002, 03:51 PM	#2
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Instinct says yes, but theory says no. The pelt creates its own heat, from usage, which would increase the load on the total rig. But regardless, some guy tried it (see OC), and basically, he found that the cold part wouldn't get very cold at all: it's hard to pull heat from a liquid passing by. He tried it with three pelt, in some large sandwich type assembly, using seperate loops to heat/cool. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

07-23-2002, 04:29 PM	#5
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Ok, so then your idea is good, but you'll have to figure out what Kev did wrong here Maybe a cross drilled block on the cold side would have been more efficient. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

07-23-2002, 04:32 PM	#6
ambient Cooling Savant Join Date: Jan 2002 Location: Planet Earth Posts: 312	I think chip from overclock-watercool used to sell a sandwitch pelt water chiller. I don't remember if anyone ever used it. Maybe Joe, Una or Brad would know? __________________ water cooled 1.4 tbird @ 1580 143x11 vcore 1.88 vio 3.55

07-23-2002, 04:49 PM	#8
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	I wish I could take a look at that chiller. Either way, a maximum surface area (like that provided with a cross drilled design) should allow a fair amount of heat to be pulled from the water. It's really a good, innovative idea! __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

07-23-2002, 06:31 PM	#9
JimS Cooling Savant Join Date: Jun 2001 Location: Connecticut Posts: 140	The chiller block kev used was a simple "U" shape inside the block which yielded very little surface area. My system has a similar setup but is more like what airspirit's design describes with two separate loops, the cold side has no radiator. Using the same pelts kevin used ,5V pelts pumping 50W (x2)I can keep my CPU right around ambient with this setup. The water is below ambient, but I rarely get any condensation and I have critical points insulated. I am redesigning my blocks now because it now looks like my flow is not even. I will try to remember to take some snapshots as I go and will publish the finished product in a thread here for all to see. My goal has never been subzero cooling, just wanted to keep the CPU and NB at ambient or thereabouts. I could get some bigass power supplies and huge pelts to get cooler, but not worth it. These 5V pelts run fine off of a standard PC power supply and work well for this purpose. My electric bill went up $4/month when I put it in use. Cova, the flaw in your design above is that you are using the one loop. No way water can change temp. that fast. For these chiller blocks to work, they need to have a large amount of surface area exposed to the water in the cold loop. If setup properly, they can be effective. Last edited by JimS; 07-23-2002 at 06:33 PM.

07-23-2002, 04:00 PM	#3
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	The only way to make that work is to have two separate self contained water cooling units. Unit one is chilled by the cold side of the pelt and directly cools the CPU. Unit two cools the hot side o' the pelt. Caveats: Unit one should not have a radiator or the radiator would actually HEAT the water, if the water gets chilled below ambient. Further, unit two should, save the water blocks, be isolated completely from unit one to prevent bleeding. This is why it isn't a good failsafe system. If the pelts fail, you will now have no cooling except for the water on the CPU. Once that water reaches a certain temp, either the hoses will burst from increased temp or the CPU will fry. Either way you're screwed. In order to make it truly failsafe, you need a radiator in Unit 1 in case of pelt failure. Unfortunately, you're now cooling your room, but it'll give you slightly lower temps than straight water. Your rad will become a condensation magnet, though, so isolate it from your comp. For that matter, all hoses and EVERYTHING in Unit 1 will need insulated for that same reason. Talk about a PITA. Best case is with the rad inline, though it is much less efficient than direct pelt cooling. The benefit is that your system should never fry unless you lose both the pelts and the pump. Without the rad in Unit 1, you will essentially have a larger time window before your CPU fries if you lose your pelts. Does that make sense? I'm on dental drugs right now, so I'm not totally with it.

07-23-2002, 04:07 PM	#4
Cova Cooling Savant Join Date: May 2002 Location: Canada Posts: 247	I've got a whole 'nother set of ideas I want to experiment with at home airspirit that much more closely fits your 2-loop setup. Thats essentially a water-chiller setup. I've seen a lot of the dual-loop pelt-chiller setups - this is different and new. If you want a water-chiller setup you're better off running phase-change to cool the water (as in #rotors dehumidifiers). I know pelts create their own heat, as well as moving heat from side A to side B. I also know that pelts run more efficiently at lower voltages (create less of their own heat) while moving less heat from A to B. The theory in my above picture is to increase the efficiency of both the rad and the block (and I'm sure that it would), and if I'm right, that efficiency increase will be enough to offset the heat that the pelt creates in making them more efficient.

07-23-2002, 04:46 PM	#7
Cova Cooling Savant Join Date: May 2002 Location: Canada Posts: 247	I'm not sure what kev did wrong in that article. There is a site in the ProCooling links section - www.frostyfox.com (don't go there now, site is dead and you'll get popups from hell I just found out) of a guy that built a pelt-powered water-chiller, but using 6 pelts at 12V. With an additional 2 pelts right on the CPU he was running in the -60C range with no load if I remember correctly. In kev's article it looks like he did 2 experiements, one with 2 buckets of water, and one with his sink's - in the buckets of water experiment it appeared to work, and in the sinks it didn't. I think he likely made a mistake in one of them and didn't realize it at the time (eg. the chiller block had some air caught in it). Either way - I'm not building a 2-loop water-chiller here. As I said above if you want to chill water phase-change is a much more efficient way to do it. I want to use a single water-cooling loop, and potentially make that single loop more efficient through the use of a pelt in a different position than I've seen before.

07-23-2002, 07:20 PM	#10
Cova Cooling Savant Join Date: May 2002 Location: Canada Posts: 247	Hmm..., that likely would be a problem - without really long blocks on either side of that pelt the water likely wouldn't change temp by enough to make much difference. Hmm..., perhaps a solution to this will pop into my head while I'm trying to solve some other problem (thats how I got the original idea). The goal at the end is still to have a single cooling loop, but make it more efficient.

07-26-2002, 06:12 PM	#11
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	What was done wrong in the article is in no experiment was the hot side of the pelt isolated from the system. In all of them there was an avenue for the heat to bleed back in some way. If the hot side of the pelt is isolated from the main CPU loop and sufficiently cooled, you will have superior results to watercooling. I guarantee you will have better results with standard pelt cooling ... plus, you won't have to worry about your water lines freezing or collecting huge amounts of condensation (zappo).

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)