PDA

View Full Version : Radiator to cool 3 overclocked GTX 680!



ZCommander
08-21-2012, 04:53 PM
Greetings all,

I was wondering what's a good radiator for cooling 3 GTX680 FTW with some overclocking?

is a 360 radiator enough or should I go with a 480 ?

zalbard
08-21-2012, 04:57 PM
480 minimum.
360 will work, but far from being quiet.

ZCommander
08-21-2012, 05:02 PM
Hmm, then I should go with a cheap 480, maybe the swiftech, I can't afford another XSPC 480 in my case as it will make the case super heavy and will cost me some $$$$$$$$.

ZCommander
08-21-2012, 05:04 PM
Now wait, I have a better idea, how about cooling my 3 GTX 680 with my current RX480 and get a 360 to cool my CPU and mobo ?

zanzabar
08-21-2012, 05:27 PM
what cpu are you running, using a 360 just for the cpu and MB seams overkill. i would guess that you need 5 fans of rad overall for a 2011 chip and 3x 680 at a reasonable noise level. and why get 3 unless you are crunching 2 is more than enough and triple sli dose not scale well and adds in quite a bit of lag.

what case are you going to use, i would look at a 2fan and a 3fan, but if you already have a 4 fan you could just add in a 1 or 2 fan rad.

Xero99
08-21-2012, 06:18 PM
I'm running 3 overclocked Galaxy's on 1 RX360 and my temps never go above 56 degrees! I run them with 6 AP-14 fans in push/pull config its not loud at all. I find this set up gives me the option of
67 degrees @ 1000RPM (quite) or 56 degrees at 1450RPM.

Just my 2 cents.

Manicdan
08-21-2012, 06:36 PM
im with xero99

a good 360 solo should be more than enough

i would just toss it all onto your current 480 and see what temps are first. if its not good enough add in an extra 240 or 360 and share all the water.

El Mano
08-21-2012, 10:57 PM
im with xero99

a good 360 solo should be more than enough

i would just toss it all onto your current 480 and see what temps are first. if its not good enough add in an extra 240 or 360 and share all the water.

Common sense seems so scarce these days.
ZCommander, you should do as Manicdan says. Test your current equipment and if it doesn't work well enough there will be plenty of time to buy new stuff :)

ZCommander
08-22-2012, 02:59 AM
I am running a 980x @ 4.35GHz with 1.38v and an asus Rampage 3 extreme.

I want to use a separate loop for the graphics cards, I have an extra used DDC and a dual bay dual pump XSPC res.

SO all I need is some universal water blocks for the cards which I am leaning towards the EK SUpreme HF VGA or the swiftech MCW82 and a radiator.

bds71
08-22-2012, 03:45 AM
personally, i would put the quad on the 680's and put your CPU on the 360. you definately do not need a quad on your CPU, and it can be put to better use on the GPUs. just my 2c.

CrazyNutz
08-22-2012, 05:23 AM
Yeah. Put the Quad on your 680's. Then get a cheap 360, or a good 240 for the CPU

eRazorzEDGE
08-22-2012, 08:31 AM
I am running a 980x @ 4.35GHz with 1.38v and an asus Rampage 3 extreme.

I want to use a separate loop for the graphics cards, I have an extra used DDC and a dual bay dual pump XSPC res.

SO all I need is some universal water blocks for the cards which I am leaning towards the EK SUpreme HF VGA or the swiftech MCW82 and a radiator.

Just curious... why do you want to separate the loops?

I'm running two rigs; one for gaming and one for a file server/HTPC/community gaming, and both run a single loop.

My gaming rig runs a 3930K @ 4.7GHz (using a XSPC Rasa block) and 3x 7970's @ 1150MHz (using Swiftech MCW 82's), with 1x Swiftech MCP-655 vario @ "3", 3x Feser 360 rads (arranged so the surface area of all rads equals 360x360 square) with a custom shroud to encase all three rads and 3x Scythe Slipstream 120mm fans @ 1200 RPM. The loop is in a serial configuration using Ø0.5" tubing/fittings. During regular gaming, my air/water delta is less than 4°C with the air temp around 25-27°C, the CPU at around 50-55°C and the GPU's at around 42-48°C. When stressing the system or doing distributed computing, the air/water delta can reach 7°C and the CPU can get up to 70°C, while the GPU's only increase linearly with the added water temp.

My HTPC rig runs a 1090T @ stock (using a Swiftech GTZ block) and 2x 6970's @ stock (using Swiftech MCW 80's), with 1x Swiftech MCP-655b, 1x Coolgate 420mm rad, no shroud, and 3x Yate Loon 140mm fans @ 1000 RPM, and the loop is in serial configuration. During just regular use (web browsing, playing movies, etc.), the air/water delta is at 5°C, not sure exactly what the temps are of the CPU or GPU's. But during gaming, the air/water delta can reach 12-15°C. Increasing the fan speed to around 1500 RPM drops the delta by 2-3°C.

There is a big difference in my two configs with regards to heat input and heat dissipation, but it should give you an idea how much rad/fan you'll need for the temps you want to operate at.

Xero99
08-22-2012, 01:35 PM
bds71 is correct. Quad for the 680's and 240 for the CPU. What you will find with the 680 is just how little heat they put out compared to lets say my last set up of 3 overclocked 6780.
The 3 6780 was on the same 360 and would easily hit 77 degrees where as the 680's never go above 60 degrees and that's after several hours of gaming.

The new generation of cards from both ATi and Nvidia are a hell of a lot more energy efficient than the last gen.

ZCommander
08-22-2012, 02:17 PM
Hmm, I was thinking of using my current 480 on the GTX 680 FTW cards since they hit 80c+ on air !!!

I'm thinking an EX360 for the CPU with my old s-Flex fans @ 1600 RPM what do you think ?

zanzabar
08-22-2012, 02:27 PM
i think running one big loop will run cooler, and then a 2 fan rad should be more than enough to stay around 1000rpm

Conumdrum
08-22-2012, 02:28 PM
Yep, with the math you did figuring out heatload and expected Dt temps, why even ask?

Man, does anyone do math aymore?

ZCommander
08-22-2012, 04:53 PM
i think running one big loop will run cooler, and then a 2 fan rad should be more than enough to stay around 1000rpm

One big loop is out of the question for me as I will be using universal GPU blocks that are more restrictive than normal FC blocks.
I like my CPU temps where they are now, one big loop will for sure raise it up especially that I live in a hot country.

zanzabar
08-22-2012, 05:15 PM
One big loop is out of the question for me as I will be using universal GPU blocks that are more restrictive than normal FC blocks.
I like my CPU temps where they are now, one big loop will for sure raise it up especially that I live in a hot country.

you have 2 pumps, and i do not think that it will raise the temp.

ZCommander
08-22-2012, 05:19 PM
you have 2 pumps, and i do not think that it will raise the temp.

2 pumps in an XSPC dual bay res. they are separated and connecting them together will be a little awkward !

Xero99
08-22-2012, 06:27 PM
Ive always used one loop till I upgraded to the 680's. Figured I try a two loop set up for once. I know that there are post by experts that stated that one loop is better, but like Zcommander
I really like where my CPU temp is. I notice that under heavy gaming my CPU temps don't rise as high as it did with a single loop. I can only assume that the CPU is no longer getting flooded with the GPU's high water temp.

eRazorzEDGE
08-23-2012, 05:14 PM
One big loop is out of the question for me as I will be using universal GPU blocks that are more restrictive than normal FC blocks.
I like my CPU temps where they are now, one big loop will for sure raise it up especially that I live in a hot country.

Flow does not affect temperatures as much as you think. And besides, what's 5°C on water anyways? I'll tell you, a difference of about 5°C on your CPU or GPU... that's it. Living in a hot country is exactly the same as living in a cold country when dealing with electronics temperatures. The CPU and GPU are only going to get a certain temperature above the water temp, no matter if the water temp is at 5°C or 50°C.

Do you guys even have temp sensors for you water and air to be able to tell the difference?

lowfat
08-23-2012, 05:49 PM
Flow rate has almost zero affect on temps. You could drop to 0.2 GPM and still not see a measurable increase in temps. I agree with the others, one loop is the way to go. You'll likely get higher temperatures with two separate loops over a single.

Conumdrum
08-23-2012, 06:41 PM
Flow rate has almost zero affect on temps. You could drop to 0.2 GPM and still not see a measurable increase in temps. I agree with the others, one loop is the way to go. You'll likely get higher temperatures with two separate loops over a single.

Wrong. Don't forget your talking to the INTERNETS. A GPU loop can deal with lower flow rates. A CPU loop due to inpingment and the block properties NEEDS a higher flow rate than your stated low low flow rate.

Specify in the posts please. Don't give new folks wrong info folks.

Manicdan
08-23-2012, 06:52 PM
if your max cpu oc was done at up to 60c, and in gaming it reaches 50c, but if turned into a single loop while gaming becomes 55c, you would still be 5c under what your cpu can easily handle. even if you gpus are putting 1000w into the loop, the cpu temps are mostly going to be based on the load being applied rather than how cold the water being fed is.

for flow speed its not that bad if your go from 1.5gpm to like .8gpm, you might see a 2-3c increase. i literally run my pump at the lowest speed allowed possible, and it handles a restrictive cpu block, and gpu full coverage block and mb mosfet block.

lowfat
08-24-2012, 05:15 AM
Wrong. Don't forget your talking to the INTERNETS. A GPU loop can deal with lower flow rates. A CPU loop due to inpingment and the block properties NEEDS a higher flow rate than your stated low low flow rate.

Specify in the posts please. Don't give new folks wrong info folks.

Flow rate really has very little impact on CPU temperatures as well. It has even less on GPUs. Maybe in the day w/ when we were using CPUs w/o IHSs it was important but it really isn't now. I've done the testing myself (albiet w/ a crappy Koolance flow meter). I went down to about 0.12GPM using an Apogee XT, there was no noticable change in CPU temps. Same w/ my current systems, I run my MCP35x at ~ 1800RPM, with no noticable impact on temperatures.

Church
08-24-2012, 08:03 AM
According to skinneelab for cpu blocks flow 1<>1.5 gpm nets half degree, 0.5<>1 gpm ~ 1 degree, ~3-4 degrees worse at ~0.25 gpm (thus probably ~5C@.2gpm). If we hunt for the best blocks with performance of them often within degree among top ones, then imho 5C are not exactly to ignore. Might be acceptable for some though, but i'd prefer more flow then .2, at least .7-.8, which usually is easy to get even with single common DDC/D5 pump.

eRazorzEDGE
08-24-2012, 04:34 PM
According to skinneelab for cpu blocks flow 1<>1.5 gpm nets half degree, 0.5<>1 gpm ~ 1 degree, ~3-4 degrees worse at ~0.25 gpm (thus probably ~5C@.2gpm). If we hunt for the best blocks with performance of them often within degree among top ones, then imho 5C are not exactly to ignore. Might be acceptable for some though, but i'd prefer more flow then .2, at least .7-.8, which usually is easy to get even with single common DDC/D5 pump.

Exactly the point a few of us (I think) were trying to make. Having flow at around 0.5-1 GPM, you're talking about a 1°C rise in water temperature. Thank you for taking the time to pull numbers.

People just need to look at review sites, like skinnee's and martin's, and see what kind of pressure drop the rads, blocks, and fittings are producing and then ask themselves "how high is too high (for water temp) for what I'm trying to accomplish?" If you want to run your CPU on a single 360 radiator cuz you think a 1°C air/water delta helps you maintain your sandy bridge overclock at 4.0Ghz and 1.5v, then go for it. But don't expect a lot of us to tell you that you're doing it right.

ZCommander
08-24-2012, 10:39 PM
What is it I am doing wrong if I went 2 loops ?!?!!?

single loop or 2 loops, if temp. is minimal then cost comes in, again both will cost me the same.

So it won't matter going single or dual ?!

Church
08-24-2012, 11:18 PM
ZCommander: you can go dual loops of course if you wish to. It's just that single loop imho is preferable. You can check eg. Gabe's paper on single vs dual loop. Putting both pumps needed at dual loop in single one you get extra redundancy and not that much worse flow. +If your most common usage scenario is gaming instead of folding/stresstesting, then you might get lower temps with same rads in one loop. +You can save $$ on buying components per loop (eg. flow meter/temp probe/res/fillport/drainport or have impossible loop config if with dual independent loops - with one big rad or one pump for everything). +Single loop tubing routing is simpler (good for 1st LC). For these and other reasons vets advise single loop. In 95% cases it's win-win :)

ZCommander
08-25-2012, 03:45 AM
Ok so, single loop it is with one DDC pump.

Now which one do you think is better :

1- Pump> 360 > CPU > Mobo > 480 > GPUs > res
2- Pump > 480 > CPU > mobo > 360 > GPUs > Res
3- Pump > CPU > Mobo > GPUs > 480 > 360 > Res
4- something else

eRazorzEDGE
08-25-2012, 12:42 PM
I prefer number 3 above. Generally, you will not see even a 1°C drop in water temperature from the inlet side to the outlet side of your radiator. So it usually doesn't matter if you split up the "heat source"/"heat dump" as shown in numbers 1 and 2 above. You can, of course experiment to see if one option gives you better results that another.

I wasn't saying you're doing it wrong, but in the scenario I put forth (with running a sandy brdige CPU @ 4.0GHz at 1.5v, which is of course way more voltage than should be required), water temperatures do not matter. So I'm suggesting that you do some research on your own, after you've gotten our input, and see what works out best for you. I've done a lot of testing on my own to realize that for what I do on my computers, an increase in the water temps (by around 5°C) does not have a negative impact on my overclocks or stability.

Advice is great. Experimenting for yourself is better.

Conumdrum
08-25-2012, 03:52 PM
Res before pump. Tubing routing next. That's it.

With 3 680's you will see a bit more than 1C in the rad in/out temps unless yoou are wayy overadded, and you are just about there.

See my first line, go for it.

Some helpful stickies:

Read first:http://www.overclockers.com/forums/showthread.php?t=674257

Lots more good stuff:
http://www.overclockers.com/forums/showthread.php?t=628092

bds71
08-27-2012, 03:57 AM
Yep, with the math you did figuring out heatload and expected Dt temps, why even ask?

Man, does anyone do math aymore?

i'm curious - what's the math? granted, i've only been at this for a few years, but i would be interested in any formulas that can help determine how much rad to use. (i'm assuming for a given heat load in watts - something to do with the heat load calculators floating about in the ether?)

ZCommander: it's your loop - if you want to use 1 or 2 loops, do it however you want (informed or uniformed, neither is wrong). but, if your doing it because you think it might save a few degrees on your CPU, that might not necessarily be the case. and, even if it is - a few degrees won't effect your CPU, or your ability to OC said CPU, in the slightest. (there - now your informed :)) ----- but, it's still your loop!! if you wanna make two loops (just to see if you can, or to experiment!!) then go for it. experimentation is fun. it's how we learn all this stuff :) the great thing about WCing is you can never go too big. want to cool a CPU with a mora 3? sweet!! this is called xtreme systems. so go get xtreme and stop worrying about the "best" way to do something. and, if it don't work then change it up and try something else. you have all/most of the components you need already (2 pumps, quad rad, fittings, tubing, etc) so it doesn't sound like cost is really a facter, so just have fun with it.

Conumdrum
08-27-2012, 01:32 PM
Well, the math is extra simple. You use Goggle, forums etc for your max usage wattage. And account for load testing which pushes wattage usage.

You then use Martinsliquidlab.org, Skinnelabs.com, and forums that are worth a crap to decide if you have enough raddage for the wattage.

Or, you can build the system and if the temps on the parts are too high you increase fannage till you need more raddage.

Math, it's simple, understanding the wattage, noisage, raddage is just through reading a lot and building a system again and again over the years. Thankfully,

It's a hobby and not a job. New folks need to understand the idea of the math and why a 120mm rad won't cool a 690 and a 3970K.

Hope that helps.

bds71
08-28-2012, 12:36 AM
New folks need to understand the idea of the math and why a 120mm rad won't cool a 690 and a 3970K.

ahhh, i got ye now :) i was thinking there was some magical formula. well, i guess there is, but it's more mystical than magical hehe. read.learn.know. (and knowing is half the battle, eh?). just remember: you've never done ANYTHING...until you do it. so don't be scared - just DO IT! (it'll only break once - i promise :)