I'm a VERY experienced system builder (I was plugging 16k ram chips into 8088 motherboards before i could ride a bike) but am BRAND NEW to h2o cooling and still trying to get my head around it. Have been researching for a week or so and won't be pulling the trigger on components for 2-3 weeks.
There is an overwhelming amount of information out there, some of it contradictory, but I've yet to see my question addressed.
My new system is a 2600k with a 6870 (soon to be upgraded to probably a 570 or 580). Plan to overclock EVERYTHING to the max. I don't game much, in fact, these days my phone has enough processing power for my computing needs, but hey, it's what I do.
So watercooling has me fascinated and as of now I'm planning on externally mounting a 3x120 rad and hitting both the CPU and GPU to start with, eventually hitting chipsets maybe RAM (though mostly for fun and aesthetics if i have as much fun with watercooling as i anticipate).
Debating running a parallel loop, which i doubt ill need now but could come in handy when i sli.
Now here's my question.
You can get a single 120 rad pretty cheap: $35ish. Why not run a serial loop going first to the CPU block, then to the single 120 rad mounted internally, then to GPU, then external 3x120 rad?
It seems to me that the single rad would knock a few degrees off the water from the CPU before it hits the GPU block, right?
Am I missing something?
Any reason nobody has tried this? flow restrictions? inefficiency? poor price/performance? Just wholly unnecessary?
It really won't make a big difference in your temps...like you said, maybe 1-2C at very most...probably less in a single 120. Intake/exit ports on my two MCR320's I see about 2-3C difference between a total of 6x120mm worth of rad space. Just food for thought.
You really don't need RAM blocks- they are restrictive and DDR2/DDR3 both run very cool these days. Also, watercooling RAM blocks aren't really designed well to remove heat from the modules- almost all RAMblocks sit across the top of the DIMMs where there is very little surface area. You'd be much better off with good RAM with decent heatspreaders and/or an active cooling solution like those fans that clip over your RAM DIMMs...if you even need that.
Radiators are very low restriction, so running 2 or 10 in your loop should cause very minimal drop in flow.
Just FYI- If you are planning on cranking that 2600k, you're going to want a 3x120 rad just for that sucker...they start spitting out a ton of heat over 4.5 gig as you get closer to 5ghz. As for the GPU, it really depends on what you decide on. A 570/580 has a TDP of 220/244 watts, respectively. Overclocked, even more. I'm thinking a 320 and a 220 rad might be in order, here. If you go SLI, remember to add the TDP of the second card into your loop calculations. No needs for dual loops, necessarily. If you really want to, there are a couple alternatives to completely separate loops that will help temps.
Keep your loop components in serial, never parallel. The only allowable exception would be if you chose to use parallel SLI bridge connectors.
Thanks for the quick response rubix. Not looking like its worth it for 1-2 degrees...I'd hoped for a bit more than that but knew there had to be a reason people weren't doing it
I know active cooling isn't necessary on modern DRAM, even OCed...as I said in my original post looking todo it as much for aesthetics (with UV tubing) as anything else. (I've got a 17 year old nephew I'm trying to get into computing so have done some things with this build with a teenagers sense of aesthetics in mind).
But I'm taking a look at that swifttech apogee HD block with3 outlets. Figured I could run one to each GPU with3/8 or 1/2 and the third using1/4 to the chipset and RAM.
If I were to go that route and have 3 tubes coming off the CPU, could I then have each tube go into an internal res and then a single outlet from that res to the external rad?
Swiftech makes rads with 3 inlets designed for the apogee HD, but they're rather pricy.
If you'd told me in 1983 that 30 years later my build was being slowed by my ignorance of basic plumbing principles...
You really don't want to split flow like that. Those Swiftech blocks are decent for parallel loops in specific instances, but be aware that when you split flow like this, your flow rates between the split branches drops as much as 50%. While it might be convenient, it isn't performance-savvy to do so. You are much better off running every component in a single, serial-only loop.
As for the radiator order question from before- regardless what order you put components in, you aren't going to see more than a few degrees Celsius difference at any point in the entire loop. Loop order does not matter when it comes to temperatures, so feel free to design your loop to benefit simplicity when building and filling your loop...also to minimize tubing length. Usually, everyone will agree that keeping your reservoir -> pump is a good method to allow you to easily fill and prime the loop. Keeping it at least a bit higher than your pump inlet will help prevent your pump from pulling air through it.