Thoughts on SLI cooling

A

astolpho

Limp Gawd
Joined
Jul 12, 2004
Messages
209
I assume I'm not the only one to give some thought to this predicament. Two video cards..... probably very hot running video cards. I think I read somewhere that the video cards are putting out more heat than the procs lately, at least on the AMD side of things. Definitely don't want two GPUs and a proc on one cooling loop, unless maybe some water chilling is applied..... need a monster pump... probably don't want both GPUs on one loop..... The proc and a GPU on one loop and a GPU and Chipset on a second loop?..... Maybe 3 loops?... GPU loop, GPU loop, proc and chipset loop? Gotta look into a full tower case, thats a lot of pumps, reservoirs radiators, and tubing..... Its a nice problem to have, SLI, but have you guys got any ideas how you would handle it?
 
astolpho said:
I assume I'm not the only one to give some thought to this predicament. Two video cards..... probably very hot running video cards. I think I read somewhere that the video cards are putting out more heat than the procs lately, at least on the AMD side of things. Definitely don't want two GPUs and a proc on one cooling loop, unless maybe some water chilling is applied..... need a monster pump... probably don't want both GPUs on one loop..... The proc and a GPU on one loop and a GPU and Chipset on a second loop?..... Maybe 3 loops?... GPU loop, GPU loop, proc and chipset loop? Gotta look into a full tower case, thats a lot of pumps, reservoirs radiators, and tubing..... Its a nice problem to have, SLI, but have you guys got any ideas how you would handle it?
Click to expand...

I couldnt help it but I read the above to the rythem of Rain Man, well, Hoffman's portroyal of him anyways. :D

It will definitely be interesting to see what cooling can be applied to these cards since there will be a few clearance issues to discuss when manufacturing a water block at the least. I'd be interested to see if the AquagraFX 6800 blocks that I can get a hold of will fit if I ever decide to run a dual 6800 setup. Oh damn that would be nice.

-Dan
 
astolpho said:
I'm a very good driver....
Click to expand...


^^^^
WTF? another thing that we might have to look at is the PCI-e slots are so close together also... this means that there will be block thickness constraints as well.... as far as runing mor than one loop to cool the GFX cards it is very possible, I think in the long run we will just have to wait nd see what hapens :D
 
YOu can easily use two of these....http://snt-systems.com/catalog/product_info.php?products_id=137

And then use both 90 degree and straight fittings for a clean fit. I'd then make the pump an MCP650 or a D4 instead of an Eheim just so to compensate for the two GPU blocks and 90 degree fittings.

Separate loop for the CPU, although I think a nice heatsink is better... as there is more benefit overclocking the GPUs then the CPU for gaming.
 
Have I mentioned that I hate the snt website? Everytime I click on something I want more info on, it takes me back to the login page. If there were any other page in english that carried this merchandise I would be there, believe me.
 
I would almost bet that you would see less than a degree C difference between one big loop with everything in series, and separate loops for each thing cooled, assuming you are using one radiator. Water's heat carying capacity is so huge compared to the amount of heat we are dealing with, that it will all reach an equlilibrium no matter how many things are in series. (assuming you keep the flow up)

And remember you do not want to split the output of a pump as it will not be even flow between the loops (the path of least resistance will get the most water) so you would need separate pumps for each loop.

==>Lazn
 
I was thinking of 3 c-systems pumps, 3 120mm Radiators, and 3 of those cool looking aquacomputer reservoirs. Probably can go with 3/8" inner dia tubing with the loops being so short, and let's face it, 3 loops, I'll need to conserve space everyway i can.
 
astolpho said:
Have I mentioned that I hate the snt website? Everytime I click on something I want more info on, it takes me back to the login page. If there were any other page in english that carried this merchandise I would be there, believe me.
Click to expand...

I have never been able to experience this "error." What operating system and browser do you use? There should be no reason why you should be returning to the login page unless your browser doesnt read PHP well.

-Dan
 
At home it doesn't happen, must just be the firewall here at work not running scripts right. Not the website's problem.
 
I'm really thinking about this too much, but as cool as those aquacomputer reservoirs look, they are aluminium and I really don't want that mixing with my copper blocks and radiators, (the corosion factor). Maybe a T-line would be the way to go, saves even more space. I keep this up and I may get this in a mid-size tower
 
my opinion would be two loops, 2 c systems pumps, 3/8 id. the 2 video cards in one loop, and one loop for the cpu.
 
My WC system could handle 2 of the computers above and cool them adequately. I'd just have to turn the fans up.

You guys are really under-estimating the cooling power of a decent watercooled system. A mag 3 can push water through damn near anything, a dual pump system is overkill in this case. A double heatercore is a ton of cooling area, I could almost run it passive in my system and it would still cool better than stock cooling. Waterblocks arent a problem. You could run everything inline and not have a single issue.

You dont need 3 120mm radiators, you dont need 3 csystems pump when a mag 3 could probably take even them all put together, you also dont need to split the loops up, it doesnt run anything cooler overall. You would still be running the same heat to the same dissipation area.

This is a common misconception: People think that 3 radiators, 3 pumps, and 3 seperate heat sources on 3 seperate loops has a cooling advantage. In reality, all the water does is actively move heat from the processor to the radiator. It does not heat up significantly after each heat source, and does not cool significantly after the radiator. water temp is almost constant. Because of this, it would not hurt to concentrate cooling effort onto one loop with 3 times the cooling area on just a single line. Plus it would be cheaper, more reliable, and much easier to do (not to mention take up less space).
 
I understand that water in a loop reaches equilibrium and doesn't get significantly if at all hotter at anyplace in the loop, but it would seem to make sense that the more watts of heat you put into the loop the hotter that temperature of equilibrium would be. That's why I think seperate loops would be cooler than one loop for all. It may be that one loop with equivalent cooling area say an EVO 360 as opposed to a EVO 120 and a larger reservoir that would increase the volume of water being heated to the equivalence of 3 loops and a hell of a pump that could give you an equivalent flow rate to what you could achieve in 3 smaller loops would perform as well, maybe even better. I don't know, thats why I ask the hard questions and listen for informed opinions, or well thought out hypothesis. You are certainly right in saying one loop would be cheaper, easier to maintain and to route than 3 loops would be.
 
Lemme explain this again: the guy above my last post wanted to put in 3 loops with 3 different radiators. If you were to put all 3 of these radiators in the same loop instead of 3 seperate loops, you will get the same cooling capacity.

1 loop x 3 radiators = 3 loops x 1 radiator each = same temperatures
 
Wouldn't flow rate and overall water volume have some effect on that equation?.... Note, this is not an argument, this is a question.
 
A lot will depend on the power of the pump. The resistance of the cycle, in my experience, is determined by the most resistant part. So if one block lets through 2 gpm, the others 3 gpm, your total flow will probably be a bit less then 2 gpm in the system with a decent pump. It becomes problematic if you have a number of fairly restrictive blocks in the system, then the restrictions become additive and your total flow (and the cooling performance) will suffer unless you have a very strong pump (which in turn means more stress on connetors, hoses etc.).
 
The 3 c-systems loops sounds like the kinder gentler solution given that perspective
 
astolpho said:
I understand that water in a loop reaches equilibrium and doesn't get significantly if at all hotter at anyplace in the loop, but it would seem to make sense that the more watts of heat you put into the loop the hotter that temperature of equilibrium would be. That's why I think seperate loops would be cooler than one loop for all. It may be that one loop with equivalent cooling area say an EVO 360 as opposed to a EVO 120 and a larger reservoir that would increase the volume of water being heated to the equivalence of 3 loops and a hell of a pump that could give you an equivalent flow rate to what you could achieve in 3 smaller loops would perform as well, maybe even better. I don't know, thats why I ask the hard questions and listen for informed opinions, or well thought out hypothesis. You are certainly right in saying one loop would be cheaper, easier to maintain and to route than 3 loops would be.
Click to expand...

Volume of water has no effect on temperature, unless you have an open res with evaporation taking place. More water just means a longer time before equlibrium is reached.

And one good pump can often push more than three small pumps, meaning you get better flow through a single inline loop than you would through 3 loops.

Using the exact same radiators the exact same blocks, and the same flowrate (one bigger pump, or the 3 pumps spaced between obstructions) you will get the same temperature overall in a single loop as 3 separate loops... The difference will be that if one loop has more heatload than another it will be hotter and the others cooler than a single loop. But overall it would even out. (btw if you were planning it, I would suggest: don't watercool the northbridge, a large passive heatsink is best put to use for that)

==>Lazn
 
astolpho said:
The 3 c-systems loops sounds like the kinder gentler solution given that perspective
Click to expand...

Gentler? Perhaps. Effecient? No. A single MCP600, Laing D4, or Mag 3 will push harder through the tubes in a series loop then 3 seperate loops with CSP750s.

Although the restrictions of a gigantic series loop add up to a whole lot more than the individual loops of a 3 loop system, the weak pumps in the small loops will still provide lower flowrates then a big pump. You would really have to double the restriction of a already restrictive loop to match the flowrates of a D4 and a single giant loop (not that a loop with 3 blocks is giant or anything). This is very unlikely to happen in any case. A D4 or a mag 3 will beat 3 loops in nearly any situation.
 
You must log in or register to reply here.
Back
Top