I have an i5 4670k overclocked to 4.5GHz on a Corsair H50 and it still runs hotter than I would like it to. I also have a single GTX 770, but I plan to eventually upgrade to dual GTX 780s in SLI. Would it be more beneficial to have my GPU(s) and CPU on seperate loops? Seems like they'd run cooler if they were. I've been looking at the XSPC 5.25" dual reservoir/pump combo(http://shop.xs-pc.com/xsp/XSPC-Acrylic-Dual-525-ReservoirPump-Combo-w-Two-DDC_36849.html). I've attached a hasty loop "sketch"
It depends on your usage scenario and how much radiator you plan to use. If you are a gamer for example while your video card is dumping heat into the loop your cpu won't be at it's hottest. If you do something like f@h or cpu/gpu mining then the loop has to contend with the heat of both while maintaning a decent delta.
The biggest advantage to dual loops is that video cards still see substantial benefits to watercooling even with a 20° delta, whereas the cpu will suffer with water temps that high. you also cant expect miracles from water cooling haswell. I'm heat limited at 4.5ghz on a high end custom loop with what would probably be a 2-3° delta when just my cpu is loaded. If I try 4.6 at 1.275 I hit 100° in prime and the chip throttles, unless I want to delid it. Though I suspect I got a worse than average chip, temp wise.
The are two good reasons to have dual loops. Ease of maintenance is one. As in if you would like to change gpu often or cpu. But quick disconnects are just as good here. (though likely more expensive than a second pump).
The second reason is if you are aiming for a completely different delta temp on different components. You could for example run a 240 rad on the cpu and reach a 10c delta t while running 3 or even 4 gpu in a 360 rad with a delta t of 20-30. For your scenario I think a single loop is best as you have more than enough rad space and not enough gpus to make a dual loop a benefit.
Also give us the rest of the specs of this build if you may OP. If you're just on SSD(s) you can probably hide those while taking out the HD trays and add more rads.
Also give us the rest of the specs of this build if you may OP. If you're just on SSD(s) you can probably hide those while taking out the HD trays and add more rads.
Yeah, the Corsair 750D has a couple of mounting points behind the motherboard tray for SSD's, but I was planning on going with at least one HDD as a backup drive.
My planned build is as follows (with some components coming from my current rig):
Yeah, the Corsair 750D has a couple of mounting points behind the motherboard tray for SSD's, but I was planning on going with at least one HDD as a backup drive.
My planned build is as follows (with some components coming from my current rig):
Nice components. My question to you is what retail options do you have? Live in the states or overseas, etc. I am asking because you linked us XSPC in euro pricing.
Ah ok. Well the shops we go to around here are Frozencpu (Biggest selection of parts), Performance-pcs and jab-tech (Cheapest price for mcp-35x pump)..
Up to you if you're looking for a kit or buying it part by part. We can assist you on the part by part if you'd like. I saw a few builds that fit a 360mm rad up top and 240mm on the bottom or front in that case.
Remember those three stores all have discount codes for up to 5% off depending in store. I don't have them on hand but do a search before you place your order.
If there is anyone out there who is overclocking their CPU and getting anything close to a 10c delta then I'd like to hear about it. I am running a i7 4930K overclocked to 4.3GHz, and in my single loop with the GPUs at idle here are my CPU temps while running Prime95.
I can only speak to my set up and performance, and I am not implying anything beyond that. Different CPUs and different loop configurations might produce different results. But I will say that my temps are fairly consistent with other IvyBridge-E owners that are overclocking on water cooled systems. The test runs are short, 15 to 20 minutes, but temps stay fairly consistent over longer runs. I let one run for 120 minutes and the highest temp I hit was 60C on one core. But I don't have graph data for that test run so I can't provide it. A few things that are noteworthy about the above data. Coolant tempertature is not affected much by CPU temperature. My CPU temperature isn't affected at all by flow rate (any flow rate over 1gal/min produces the same cooling performance in my system so there is no difference between pumps at 2900 RPM and pumps at 4700 RPM).
Now with GPUs it is a different story. The graphs below show my GPU temps while playing Farcry 3. I run with the settings turned up at a resolution of 5760 x 1080. I am running three GTX 780 Classified video cards in triple SLI, in a parallel water cooling configuration. My cards are overclocked beyond the factory overclock, though not by a lot. The graph below is a little crowded (sorry about that) but it shows that my GPU temps track with coolant temperature. The thick lines in the middle are fan speeds that are run by control curves (not really relevant to the discussion here).
Note that at max fan my GPU temps dropped about 6C because of a corresponding drop in coolant temperature.
It must be noted that these graphs are not the final word on performance. But they do suggest to me that a lot of folks overthink their rigs, myself included. The best water cooling component that went into my recent build was the Aquacomputer Aquaero 6 Pro. It is not without its faults but it is fantastic at giving you a window into what your cooling system is actually doing.
A few words on my water cooling set up. It is a single large loop with two D5 pumps. I use a Koolance CPU block and EK Acetal/Copper Classy Blocks on my video cards. My build originally had 3 radiators: two Aquacool NexXxos UT60 480mm and one NexXxos UT60 360mm. After running the system for a week, however, I decided to drain the loop and reverse the air flow direction on the 360mm rad, and while I had that radiator out of the system I accidentally dropped and damaged it. So I put the loop back together and have been running with two 480mm radiators ever since. I have a replacement 360mm radiator but I have not been in a hurry to install it since I don't think it is going to make much of a difference in my temps.
What does all of this have to do with the question of single loop versus dual loop? It just underscores why I prefer a single large loop. Depending on your CPU and your overclock you can hit the point of diminishing returns very quickly. With dual loops you have to buy two of everything, so it will typically cost more. In my rig the CPU temp seems to impact coolant temps very little, so there is plenty of reserve cooling capacity in a single loop to cool a GPU assuming you have enough radiator. My first build was a dual loop set up and it served me well, but based on my experiences I cannot see a compelling reason to run dual loops unless it just happens to be personal preference.
When we talk about a delta here in the watercooling subforum it is almost universally meant as the delta between water temp and ambient air temp. My 2700k had close to a 40℃ gap between water temp and core temp at 5.0ghz.
Also, radiators run into diminishing returns really fast. They scale pretty well to size so if your water has a 20° delta with 120.2, it will be 10° with 120.4, 5° with 120.8, 2.5° with 120.16. Your loop was massively over the top on rads so it isn't at all suprising that dropping from 120.11 to 120.8 was a small gain in temps.
I agree that I initially had more radiator than I really needed. 120.8 is more than enough for my CPU and three video cards. I read your post further up thread and you make some very good points about gaming vs mining. I don't do the latter. I have some high CPU workloads that I run but not while gaming. And gaming alone does not put much of a load on the CPU.
Yeah thats why i run single loop myself, im either gaming or encoding video, not both. Though right now i have too much rad in my system so i could probably do both and be ok, my cpu doesnt get really hot unless im running an avx2 synthetic benchmark on it.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
