making a working cooling system out of a crappy design

SlightFall · 03-04-2004, 03:49 AM

alrighty, i have quite the problem on my hands here, the company i work for has hired a company (out of professional courtesy i wont say who) that, despite there numerous alleged "degrees" in engineering, cant seem to create a working liquid chiller for our systems... i guess i'll start with what there trying to accomplish.

an average system is running an overclocked AMD Athlon FX-51 and an overclocked N-Vidia 5950 Ultra on either an SK8N, K8V, SK8V, P4C800, or P4C800-E motherboard, in a custom LianLi chassis. and now quietly taking those extreme temperatures from both CPU and GPU (primarily the ram chips on the GPU of course) to an incredibly low temperature is there job.

-now with there cooling solution:

ok lets start with the radiator, (i'm sorry dimensions arent really my thing), its about 70mm long, and is 120mm wide as it is in height, with a 120 mm fan running at 2300rpm to "cool" it, and this unit is stationed right at the rear of the case. now there are a total of 4 tec's (running at about 74 watts per tech thus not really cooling anything) all of which are on the radiator itself with 4 waterblocks stationed ontop of them. the gpu has a "thermal transfer" plate with a cooling block mounted on top (on both sides of the gpu) and the cpu simply has a waterblock mounted. thermal transfer on the gpu is achieved by thermal pads.. and thermal tranfer for the cpu is achieved by arctic silver 5. now in the front of the case we have dual 120mm fans running at 2300 rpm and... well.. that just about does it, oh and instead of running water,, there running a glycol/water mix.

now, i may not have a "degree" in engineering, let alone a little peice of paper saying i know what i'm talking about... but bloody hell this things a disaster. these guys think they have it all figured out and refuse to change anything and after our last meeting, and 2 straight hours running a quieter air cooled system beside there liquid cooled (air cooled ran about 20 degrees cooler on the cpu), they've decided to drill a hold in the top of the case for yet another chiller of yet the same crappy design to "hopefully" get the temperatures they desire..

what i see is wrong is for one the tec's, there for one cooling the liquid itself then running to the GPU and CPU instead of tec's running directly on the GPU/CPU then to disperse the heat, via, the radiator. also the tec's are running on too low a voltage.

now heres where it gets interesting, my boss is totally down with these clowns despite him being unsatisfied with the temperatures, even though they have ACTUALLY said that "in NO WAY can there system be improved WITHOUT modding our case"... now say if you were given one of these chiller systems and ripped it apart, how would you mod it to make it work (general temperatures were seeing are about 70C for the CPU and near 90C for the GPU. and just to recap, ripping one of these systems apart will give you the following:
-4 tec's
-a 70mm x 120mm x 120mm radiator
-a 120mm fan
-4 tec waterblocks about 40mm x 40mm x 15mm
-1 CPU waterblock about 60mm x 60mm x 25mm
-2 GPU waterblocks....the top is about 35mm x 20mm x 15mm
the processor side is about 40mm x 40mm x 15mm
-as much thermal grease as you need
-as much tubing as you need

i'll get as much additional information if anyone requires it.. and i'll check back here soon.. thank you soo much for your help.

SlightFall · 03-04-2004, 04:01 AM

righto and i thought i should add what i'm trying to do right now... for starters i just want to leave two tec's on the radiator, and place the other two on both the cpu and gpu, both of which will have a copper coldplate added. second i'm wanting to run a higher wattage through the tec's. as for the liquid, i'm really not sure what else would work besides water,,, the glycol seems to be working well at transporting the heat, but its just well... cooling it down thats the problem. either way i look forward to any responses... thanks again in advance

krazy · 03-04-2004, 10:06 AM

I responded in the other thread about this setup as well, but this one has a little more technical information, so I'll toss down a few more ideas.

My number one suggestion is to fire this company and find someone else. They sound pretty worthless and aren't getting the job done the way it could be.

The major things I see needing to be changed include the following:

* I don't know what size that tubing is, but ditch it for 1/2" ID. It looks too small for the amount of heat you're trying to transfer.

* You need a real radiator. What you have now is not a radiator, but more of a heatsink attached to the hot side of some pelts, and heat has to go several inches through aluminum to get to air. With a heatercore, the heat has only one thickness of copper sheeting to go through, and there is a lot more of this better surface area. To be honest, you would probably get better performance from the current setup by attaching the hot side of those pelts to individual CPU heatsink/fan combos instead of that practically passive aluminum thing. With a "real" radiator, the pelts can't be used the way they are now, but I don't think you really need them. I'm cooling a modern AMD CPU, a hot viddy card, and a really hot chipset with a traditional watercooling setup and a nice big radiator. My CPU temps are between 28 and 35 celsius.

* Switch to copper parts. The aluminum parts in the system are going to cause nothing but trouble down the line. Besides being much less of a corrosion problem, copper has MUCH better thermal properties than aluminum. Why would this compant be using aluminum parts? Because they're cheap and easy to make. It's definitely not because they perform well, and if they are trying to tell you this, I see it as all the more reason to fire them. The waterblocks and heat exchanger would perform better and make the system last longer without getting gummed up if they were made out of copper.

* Redesign the blocks. Most of them look like a simple hollow water jacket. This is a very inefficient way to get heat off of the chips. Modern advanced blocks utilize carefully designed channels, jets, nozzles, and fins to transfer the most heat from the chip to the passing water as possible. If you can't effectively get the heat out of the chips and into the water, there's not much point of tweaking the dissipating end of the setup.

* Make sure the pump is capable of moving excessive amounts of coolant. If you have this much heat going through the loop and this many blocks restricting the flow, you need a pretty beefy pump to move the coolant through the loop quickly and with good head. In theory, flow doesn't matter as long as it is there, but the better waterblocks out there perform increasingly more efficiently with faster and more powerful flow (which creates more turbulance and thins the boundary layer down).

prandtl · 03-04-2004, 10:36 AM

Quote:

Originally Posted by SlightFall

righto and i thought i should add what i'm trying to do right now... for starters i just want to leave two tec's on the radiator, and place the other two on both the cpu and gpu, both of which will have a copper coldplate added. second i'm wanting to run a higher wattage through the tec's. as for the liquid, i'm really not sure what else would work besides water,,, the glycol seems to be working well at transporting the heat, but its just well... cooling it down thats the problem. either way i look forward to any responses... thanks again in advance

Dont even think using the TEC directly on the cpu, you will a hella more beefy TEC to applied directly to the cpu (something along the line of 170+W TEC). As Krazy said, I wouldn't even use the TEC and instead use an heatercore.

cristoff · 03-04-2004, 10:37 AM

U do know that both the AMD 64 and FX-51 have on chip memory controllers... so in that It can't overclock well... really at all... because the other things in the system like the pci/agp cards can only handle a certain overclock...

Anandtech.com - "The big surprise here is that none of the current chipsets for Athlon 64 have a working PCI lock, whether they claim one or not. This was a complete surprise and our first inclination was to throw out the results, since speeds on the Athlon 64 are derived in a very different way with Hypertransport. However, oscilloscope results and tests elsewhere pretty much support our findings. There does not appear to be a working PCI lock on any current chipsets for Athlon 64. Certainly, this contributes to the poorer overclocking that we have seen on A64 inasmuch as on-board component frequencies and AGP bus are related to PCI bus."

check this out...
http://www.anandtech.com/chipsets/showdoc.html?i=1968

most importantly this page
http://www.anandtech.com/chipsets/sh...tml?i=1968&p=4

hope it helps...

AntiBling · 03-04-2004, 11:17 AM

Even if you dropped the pelts completely the radiator is too small, IMO. A larger radiator and two slower fans in push/pull config would be quieter and more effective.

Zogthetroll · 03-04-2004, 11:59 AM

well, I replied in the other thread(which should be read as well), but i'll comment here too. problems,
1. unless you're using both copper and aluminum in a loop you shouldn't be using coolant with your setup, glycol is less effective at heat transfer than straight water, it is used primarily to stop corrosion or lower the freezing point of coolant(not an issue). most copper only setups use water with additives to stop organic growth.
2. Do not use those pelts on the cpu or gpu (well, maybe on the gpu, but that's iffy), they aren't powerfull enough.

comments in addition to the other thread, use a rad, and put a shroud between it and the fan, if possible a fan on either side of the rad. rad size is somewhat of a personal choice, one sized similarly to that heatsink would at least get you better results than air cooling. basically ditch the pelts and that pitiful excuse for a radiator, and tell them to come back when they've designed something thats less complicated and more efficient, in other words, an improvement, not a step backwards.
check out This article for some good design basics. I would also check out the articles section for some good reading. and I would second the flow issue. Higher flow through a loop is better. just a guess, but i would classify what you're getting with that setup as "anemic" try the old "bucket test" to see what kind of flow rates you're really getting (since mfg. max flow ratings for a pump are NOT what you're getting in a loop) also, take a look at some of the waterblock construction threads to get some basic ideas of how a reasonably effective wb is constructed. basically, in your place, i would be rightfully pissed at someone trying to pass this off as a quality watercooling system, especially at the implication that it "can't be improved upon" (although technically that might be the case if they want you to use those parts. garbage in, garbage out.) dare I ask what kind of cost they're trying to get out of it? for under $50 you can get a top of the line heatsink, and for around $100 you can get a thermaltake aquarius, which while not a good performer, is quiet and performs about the same as an air-cooling solution. I'd tell them to get lost regardless of price, but that's just me.

prandtl · 03-04-2004, 02:13 PM

Quote:

Originally Posted by Zogthetroll

well, I replied in the other thread(which should be read as well)

do you have a link to it? Cant find it.

Zogthetroll · 03-04-2004, 03:27 PM

yup, Here, this should do.

cristoff · 03-04-2004, 05:52 PM

I guess you guys dont get it....

No matter how good the cooling on the cpu, chipset, or video card a amd 64 or fx51 system will not overclock...

AntiBling · 03-04-2004, 05:59 PM

It appears that overclocking is not the goal, its to make it quieter.

SlightFall · 03-04-2004, 06:01 PM

Quote:

Originally Posted by cristoff

I guess you guys dont get it....

No matter how good the cooling on the cpu, chipset, or video card a amd 64 or fx51 system will not overclock...

actually we have the athlon fx-51 overclocked to 2.4 at 1.7v (totally stable) and the GPU overclocked from 475 to 525, and thats also working stable.... so here we have a totally stable system that just runs incredibly hot

SlightFall · 03-04-2004, 06:04 PM

Quote:

Originally Posted by AntiBling

It appears that overclocking is not the goal, its to make it quieter.

we've already accomplished the overclocking... didnt really take much.. but now were just hit with the problem of the noise, and heat. so yes your right.

SlightFall · 03-04-2004, 09:18 PM

alright... can anyone tell me what tec's i would need to produce good results, and maybe some links to where i can order them from. if so thank you very much... in addition to an earlier question in this thread, about the cost of these systems.. my boss is currently paying an excess amount of over $400-500 dollars canadian. and the "new"chiller that this company is going to try to build will cost an additional $100 canadian. total rip off if you ask me!!

03-04-2004, 03:49 AM	#1
SlightFall Cooling Neophyte Join Date: Mar 2004 Location: calgary Posts: 5	making a working cooling system out of a crappy design alrighty, i have quite the problem on my hands here, the company i work for has hired a company (out of professional courtesy i wont say who) that, despite there numerous alleged "degrees" in engineering, cant seem to create a working liquid chiller for our systems... i guess i'll start with what there trying to accomplish. an average system is running an overclocked AMD Athlon FX-51 and an overclocked N-Vidia 5950 Ultra on either an SK8N, K8V, SK8V, P4C800, or P4C800-E motherboard, in a custom LianLi chassis. and now quietly taking those extreme temperatures from both CPU and GPU (primarily the ram chips on the GPU of course) to an incredibly low temperature is there job. -now with there cooling solution: ok lets start with the radiator, (i'm sorry dimensions arent really my thing), its about 70mm long, and is 120mm wide as it is in height, with a 120 mm fan running at 2300rpm to "cool" it, and this unit is stationed right at the rear of the case. now there are a total of 4 tec's (running at about 74 watts per tech thus not really cooling anything) all of which are on the radiator itself with 4 waterblocks stationed ontop of them. the gpu has a "thermal transfer" plate with a cooling block mounted on top (on both sides of the gpu) and the cpu simply has a waterblock mounted. thermal transfer on the gpu is achieved by thermal pads.. and thermal tranfer for the cpu is achieved by arctic silver 5. now in the front of the case we have dual 120mm fans running at 2300 rpm and... well.. that just about does it, oh and instead of running water,, there running a glycol/water mix. now, i may not have a "degree" in engineering, let alone a little peice of paper saying i know what i'm talking about... but bloody hell this things a disaster. these guys think they have it all figured out and refuse to change anything and after our last meeting, and 2 straight hours running a quieter air cooled system beside there liquid cooled (air cooled ran about 20 degrees cooler on the cpu), they've decided to drill a hold in the top of the case for yet another chiller of yet the same crappy design to "hopefully" get the temperatures they desire.. what i see is wrong is for one the tec's, there for one cooling the liquid itself then running to the GPU and CPU instead of tec's running directly on the GPU/CPU then to disperse the heat, via, the radiator. also the tec's are running on too low a voltage. now heres where it gets interesting, my boss is totally down with these clowns despite him being unsatisfied with the temperatures, even though they have ACTUALLY said that "in NO WAY can there system be improved WITHOUT modding our case"... now say if you were given one of these chiller systems and ripped it apart, how would you mod it to make it work (general temperatures were seeing are about 70C for the CPU and near 90C for the GPU. and just to recap, ripping one of these systems apart will give you the following: -4 tec's -a 70mm x 120mm x 120mm radiator -a 120mm fan -4 tec waterblocks about 40mm x 40mm x 15mm -1 CPU waterblock about 60mm x 60mm x 25mm -2 GPU waterblocks....the top is about 35mm x 20mm x 15mm the processor side is about 40mm x 40mm x 15mm -as much thermal grease as you need -as much tubing as you need i'll get as much additional information if anyone requires it.. and i'll check back here soon.. thank you soo much for your help.

03-04-2004, 10:06 AM	#3
krazy Cooling Savant Join Date: Mar 2003 Location: St. Paul, Minnesota, USA Posts: 123	I responded in the other thread about this setup as well, but this one has a little more technical information, so I'll toss down a few more ideas. My number one suggestion is to fire this company and find someone else. They sound pretty worthless and aren't getting the job done the way it could be. The major things I see needing to be changed include the following: * I don't know what size that tubing is, but ditch it for 1/2" ID. It looks too small for the amount of heat you're trying to transfer. * You need a real radiator. What you have now is not a radiator, but more of a heatsink attached to the hot side of some pelts, and heat has to go several inches through aluminum to get to air. With a heatercore, the heat has only one thickness of copper sheeting to go through, and there is a lot more of this better surface area. To be honest, you would probably get better performance from the current setup by attaching the hot side of those pelts to individual CPU heatsink/fan combos instead of that practically passive aluminum thing. With a "real" radiator, the pelts can't be used the way they are now, but I don't think you really need them. I'm cooling a modern AMD CPU, a hot viddy card, and a really hot chipset with a traditional watercooling setup and a nice big radiator. My CPU temps are between 28 and 35 celsius. * Switch to copper parts. The aluminum parts in the system are going to cause nothing but trouble down the line. Besides being much less of a corrosion problem, copper has MUCH better thermal properties than aluminum. Why would this compant be using aluminum parts? Because they're cheap and easy to make. It's definitely not because they perform well, and if they are trying to tell you this, I see it as all the more reason to fire them. The waterblocks and heat exchanger would perform better and make the system last longer without getting gummed up if they were made out of copper. * Redesign the blocks. Most of them look like a simple hollow water jacket. This is a very inefficient way to get heat off of the chips. Modern advanced blocks utilize carefully designed channels, jets, nozzles, and fins to transfer the most heat from the chip to the passing water as possible. If you can't effectively get the heat out of the chips and into the water, there's not much point of tweaking the dissipating end of the setup. * Make sure the pump is capable of moving excessive amounts of coolant. If you have this much heat going through the loop and this many blocks restricting the flow, you need a pretty beefy pump to move the coolant through the loop quickly and with good head. In theory, flow doesn't matter as long as it is there, but the better waterblocks out there perform increasingly more efficiently with faster and more powerful flow (which creates more turbulance and thins the boundary layer down). __________________ Abit NF7-S ... xp1800+ ... Abit GF4ti4200 ... Maxtor 120gb SATA150 Currently gathering parts to water cool everything! (read about the project)

03-04-2004, 10:37 AM	#5
cristoff Cooling Savant Join Date: Mar 2002 Location: Mass Posts: 185	U do know that both the AMD 64 and FX-51 have on chip memory controllers... so in that It can't overclock well... really at all... because the other things in the system like the pci/agp cards can only handle a certain overclock... Anandtech.com - "The big surprise here is that none of the current chipsets for Athlon 64 have a working PCI lock, whether they claim one or not. This was a complete surprise and our first inclination was to throw out the results, since speeds on the Athlon 64 are derived in a very different way with Hypertransport. However, oscilloscope results and tests elsewhere pretty much support our findings. There does not appear to be a working PCI lock on any current chipsets for Athlon 64. Certainly, this contributes to the poorer overclocking that we have seen on A64 inasmuch as on-board component frequencies and AGP bus are related to PCI bus." check this out... http://www.anandtech.com/chipsets/showdoc.html?i=1968 most importantly this page http://www.anandtech.com/chipsets/sh...tml?i=1968&p=4 hope it helps... __________________ fr33t3chi3

03-04-2004, 11:59 AM	#7
Zogthetroll Cooling Savant Join Date: Mar 2003 Location: Ohio, U.S.A. Posts: 177	well, I replied in the other thread(which should be read as well), but i'll comment here too. problems, 1. unless you're using both copper and aluminum in a loop you shouldn't be using coolant with your setup, glycol is less effective at heat transfer than straight water, it is used primarily to stop corrosion or lower the freezing point of coolant(not an issue). most copper only setups use water with additives to stop organic growth. 2. Do not use those pelts on the cpu or gpu (well, maybe on the gpu, but that's iffy), they aren't powerfull enough. comments in addition to the other thread, use a rad, and put a shroud between it and the fan, if possible a fan on either side of the rad. rad size is somewhat of a personal choice, one sized similarly to that heatsink would at least get you better results than air cooling. basically ditch the pelts and that pitiful excuse for a radiator, and tell them to come back when they've designed something thats less complicated and more efficient, in other words, an improvement, not a step backwards. check out This article for some good design basics. I would also check out the articles section for some good reading. and I would second the flow issue. Higher flow through a loop is better. just a guess, but i would classify what you're getting with that setup as "anemic" try the old "bucket test" to see what kind of flow rates you're really getting (since mfg. max flow ratings for a pump are NOT what you're getting in a loop) also, take a look at some of the waterblock construction threads to get some basic ideas of how a reasonably effective wb is constructed. basically, in your place, i would be rightfully pissed at someone trying to pass this off as a quality watercooling system, especially at the implication that it "can't be improved upon" (although technically that might be the case if they want you to use those parts. garbage in, garbage out.) dare I ask what kind of cost they're trying to get out of it? for under $50 you can get a top of the line heatsink, and for around $100 you can get a thermaltake aquarius, which while not a good performer, is quiet and performs about the same as an air-cooling solution. I'd tell them to get lost regardless of price, but that's just me. __________________ My Machine (wip) XP1800, 256Mb Ram, DVD, 52x CD-R, 40Gb HD, GeForce4 MX 420 Clear Cube XP2000, 256Mb Ram, 32x CDR, 40Gb HD Last edited by Zogthetroll; 03-04-2004 at 12:04 PM.

03-04-2004, 03:27 PM	#9
Zogthetroll Cooling Savant Join Date: Mar 2003 Location: Ohio, U.S.A. Posts: 177	yup, Here, this should do. __________________ My Machine (wip) XP1800, 256Mb Ram, DVD, 52x CD-R, 40Gb HD, GeForce4 MX 420 Clear Cube XP2000, 256Mb Ram, 32x CDR, 40Gb HD

03-04-2004, 04:01 AM	#2
SlightFall Cooling Neophyte Join Date: Mar 2004 Location: calgary Posts: 5	righto and i thought i should add what i'm trying to do right now... for starters i just want to leave two tec's on the radiator, and place the other two on both the cpu and gpu, both of which will have a copper coldplate added. second i'm wanting to run a higher wattage through the tec's. as for the liquid, i'm really not sure what else would work besides water,,, the glycol seems to be working well at transporting the heat, but its just well... cooling it down thats the problem. either way i look forward to any responses... thanks again in advance

03-04-2004, 11:17 AM	#6
AntiBling Cooling Savant Join Date: Jan 2004 Location: orlando FL Posts: 147	Even if you dropped the pelts completely the radiator is too small, IMO. A larger radiator and two slower fans in push/pull config would be quieter and more effective.

03-04-2004, 05:52 PM	#10
cristoff Cooling Savant Join Date: Mar 2002 Location: Mass Posts: 185	I guess you guys dont get it.... No matter how good the cooling on the cpu, chipset, or video card a amd 64 or fx51 system will not overclock... __________________ fr33t3chi3

03-04-2004, 05:59 PM	#11
AntiBling Cooling Savant Join Date: Jan 2004 Location: orlando FL Posts: 147	It appears that overclocking is not the goal, its to make it quieter.

03-04-2004, 09:18 PM	#14
SlightFall Cooling Neophyte Join Date: Mar 2004 Location: calgary Posts: 5	alright... can anyone tell me what tec's i would need to produce good results, and maybe some links to where i can order them from. if so thank you very much... in addition to an earlier question in this thread, about the cost of these systems.. my boss is currently paying an excess amount of over $400-500 dollars canadian. and the "new"chiller that this company is going to try to build will cost an additional $100 canadian. total rip off if you ask me!!

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