I Need A Good Custom Loop For GTX 480's

[tinmann]

I have been running 2 GTX 480's in SLI and even though the temps have been manageable at Stock clocks running Battlefield 3 in Ultra everything on High bring the cards to temps in the mid to upper 70 degrees Celsius, I'd like to overclock both cards so I bought 2 Danger Den Water Blocks. Now the real question is just what else do I need to complete this project? I have a Corsair 800d case but I have a Corsair Hydro H100 CPU cooler mounted in the top of the case cooling an i5 2500K OC'd @ 4.3 Ghz with idle temps @ 30-33 degrees Celsius and in the mid to high 50's @ load. I'm pleased with it's performance so I only want to build this custom loop for the GPS's.
I'd like a good radiator that I could mount at the bottom of the case or where the rear exhaust fan sits. Would a 240mm Rad fit in the bottom of a Corsair 800d where the PSU sits or can I get by with a 120mm Rad? And which Reservoir fittings and tubing is best in my situation.

Corsair 800d
Asus P8Z68 V-Pro
Corsair Vengeance 1600 4X4Gbs
Corsair H100
GTX 4800 SLI
Seasonic Platinum 1000
Corsair Force 3 Sata 3 SSD

Do they make a radiator that fits a 140mm fan?

 

[Lutfij]

I was a little concerned about it but I have yet to hear of any issues from anyone else using this block.

google that issue in, and you'll see what everyone's throwing at it in rage.

 

[jonathann1818]

I was a little concerned about it but I have yet to hear of any issues from anyone else using this block.

google that issue in, and you'll see what everyone's throwing at it in rage.

Yes but do you mean the nickel plated blocks that were crapping out last year? I heard lots of bad things about those blocks but not much on the new nickel ones that have just been released. If you have any links regarding the new EN nickel blocks, please post them.

 

[Lutfij]

erm, you can google them - realraider i think is the thread name. or jsut google, EK nickel issue - the first link that pops up. but if you can't get it, i'll post back.

 

[jonathann1818]

erm, you can google them - realraider i think is the thread name. or jsut google, EK nickel issue - the first link that pops up. but if you can't get it, i'll post back.

The bulk of those posts for that what you mentioned are from last year, showing the flaking from last years blocks. I am not talking about those. I am talking about the new ones using the supposed EN nickel plating process. And yes it did see some posts about the new ones. I saw one video on youtube showing the flaking of the nickel within his mobo block. That video was posted just last month.

So yes I know all about the nickel flaking issue. But like I said I am not finding anywhere near as many complaints about the new EN nickel blocks other than what I just mentioned.

And I have had not one flaking problem so far. Keeping my fingers crossed. Im not saying that they wont flake though. Well have to see.

 

[tinmann]

Here are the pics.

http://imageshack.us/photo/my-images/638/dscf0839u.jpg/

http://imageshack.us/photo/my-images/809/dscf0833.jpg/

http://imageshack.us/photo/my-images/804/dscf0829a.jpg/

http://imageshack.us/photo/my-images/3/dscf0827sj.jpg/

http://imageshack.us/photo/my-images/69/dscf0825k.jpg/

http://imageshack.us/photo/my-images/855/dscf0822x.jpg/

http://imageshack.us/photo/my-images/855/dscf0819b.jpg/

http://imageshack.us/photo/my-images/23/dscf0818c.jpg/


These are the parts that I got:

Laing D5 Pump Vario (also called MCP 655, exact same pump)
RX360 Rad
RX120 Rad
Ek Spin Bay Actel Reservoir
3 x Koolance VIDNX580 GPU blocks
EK Supreme HF Acetal + Nickel CPU Block
EK X-Top D5 for Laing D5 Pump
Compression Fittings
4 x Scythe Gentle Typhoon 1850 rpm fans

A rad/pump/reservoir setup like this should allow you to water cool those 480's plus the CPU very nicely.

Here is a review that I used to select the radiators.

http://skinneelabs.com/xspc-rx360-v2/2/

Look at the top center thermal chart results (pink and blue one). That one will show you the heat load dissipated vs the temperature delta given different fan speeds and a constant flow rate. I would say estimate the actual DC power draw of the 480s and CPU from your power supply and then use this chart to get and idea of how much headroom you will have. A temperature delta of about 10C isnt bad at all (about average). But getting it lower than that is definitely better.

Here is a good review showing you waht your PC would likely consume from the wall AC.

http://www.guru3d.com/article/geforce-gtx-480-3way-sli-review/25

Apparently a fully loaded x58 system will use about 702 Watts from the wall AC when stressed. So about 90% of that would be DC so I figure that 630 Watts would be what your system is consuming. No take away the motherboard, drives ram etc. and you are probably down to about 600 watts for just your GPUs and CPU (you have LGA 1155 so its not exaclty this but close enough). That should give you a good idea for what rads and fans to use. I would still recommend using an RX360 and RX120 for this setup though.

I also use distilled water. Works great.

Hope this helps.

I like the RX360 Rad by XSPC and was really worried about the extra thickness of the radiator with fans but I see it's not a problem in the 800d. Do you use a pull configuration with the fans? I'd been looking at the Feser TFC X-Changer Triple 120mm but can't find many reviews.

 

[jonathann1818]

I like the RX360 Rad by XSPC and was really worried about the extra thickness of the radiator with fans but I see it's not a problem in the 800d. Do you use a pull configuration with the fans? I'd been looking at the Feser TFC X-Changer Triple 120mm but can't find many reviews.

Well all of the rads on the market are pretty well the same. Some may have a higher fin density and others lower. The reason I like the RX360 and RX120 is because they have a low fin density. Sure they may not dissipate quite as much heat as rads with a high fin density, but you can run fans pretty slowly and quietly because there is less to impede the flow of air through the rad. Besides the difference in heat dissipation capacity between different rads is minimal. So I would just go with XSPC Rads. Also, you will need to flush the rads before you put them in your loop because there are chemicals left behind in the rads from the manufacturing process. Some manufactures' rads have more of this stuff inside out of the box so it will require more to flush them properly. The XSPC rads dont have as much of this stuff as other brands. I dont know much about feser rads but if I were you, I would just go with what I know works. Up to you though. XSPC rads are fairly cost effective so keep that in mind.

Now with the fans, I am only using a push configuration for the RX360 and pull for the RX120. More fans will generate more noise and you probably wont see a huge difference in your temps. Only placing the fans on one side of the rad also allows you to hide them a bit as the gentle typhoons (which are great fans BTW) dont look that good because they are grey. You could probably do push pull though in the 800d. I think there is enough clearance so that the fans mounted to the bottom of the RX360 do not interfere with the mosfet coolers and CPU block/fittings/tubing. The bottom fans would come rather close to the ram banks though.

Anyways, what other components are you looking at getting for you loop?

 

[Motopsychojdn]

Try and grab an EX360, same performance as the RX series but same size as the Rs ones,
I'm having two EX480's for a couple of quid more than the price of a single RX480, more rad for your cash hehe

 

[rubix_1011]

I'm going to adjust your image files so they display...brb.

Edit: nm, you must have posted the links to your personal view...you'd need to replace them with your publicly accessible URLs.

I remember when I was selecting my parts, I was told by lots of people "just take 244 TDP for the 580 x 3 + 200 W for the CPU when overclocked and that is your power consumption". The reality is that when im really stressing the machine (ie Metro 2033 with 99% GPU usage ) the system will only consume about 930 Watts from the wall. Factoring in the AC to DC efficiency and other power consumption of the other components I was down to about 800 Watts for GPUs and CPU. Not the 950-1000 Watts I was being told. Now would more headroom be nice yes, but the system is working quite well.

So all Im really saying is that if you want to design a water cooling loop based on the TDP, you will have a lot of headroom and thats good but its also good to see how much power you are actually drawing.

This is very correct. Trying to determine actual heat dissipation based on a theoretical power draw isn't the most precise. You can often take 75-80% of that total and consider it relatively close to what a real-world system would actually be producing and need cooled. It's a little simpler to just go with power draw TDP values to make things a bit easier to calculate, but this does bring up a decent point...perhaps I should work on some calculations for determining an accurate power draw to heat output value for more accurate loop planning and design. This one might take me a bit of research.

That's if I can find time to stop playing BF3 if/when I'm not working all the damn time.

 

[jonathann1818]

I'm going to adjust your image files so they display...brb.

Edit: nm, you must have posted the links to your personal view...you'd need to replace them with your publicly accessible URLs.

I remember when I was selecting my parts, I was told by lots of people "just take 244 TDP for the 580 x 3 + 200 W for the CPU when overclocked and that is your power consumption". The reality is that when im really stressing the machine (ie Metro 2033 with 99% GPU usage ) the system will only consume about 930 Watts from the wall. Factoring in the AC to DC efficiency and other power consumption of the other components I was down to about 800 Watts for GPUs and CPU. Not the 950-1000 Watts I was being told. Now would more headroom be nice yes, but the system is working quite well.

So all Im really saying is that if you want to design a water cooling loop based on the TDP, you will have a lot of headroom and thats good but its also good to see how much power you are actually drawing.

This is very correct. Trying to determine actual heat dissipation based on a theoretical power draw isn't the most precise. You can often take 75-80% of that total and consider it relatively close to what a real-world system would actually be producing and need cooled. It's a little simpler to just go with power draw TDP values to make things a bit easier to calculate, but this does bring up a decent point...perhaps I should work on some calculations for determining an accurate power draw to heat output value for more accurate loop planning and design. This one might take me a bit of research.

That's if I can find time to stop playing BF3 if/when I'm not working all the damn time.

Well from what I have found there is an approximate proportion between actual power draw compared to TDP and as you probably know it depends on the heat of the chip (youre a guru so i think its pretty obvious you know). With more heat comes more resistance and therefore more power draw and more heat. This is the reason why the GTX 480 consumed so much damn power, apart from the architectural defficencies. When looking at the 580, the actual power draw is much less, sometimes by 30 Watts. This I am sure is the reason why my power consumption numbers were significantly lower than the theoretical numbers others gave me.

So I think the way to do it is to assign a realistic power draw value based on the realistic temperature of the chip and apply a typical reduction factor to it based on this.

Obviously the GPU and CPU usage plays a major role. I mean when do you actually fully utilize all of the GPUs and CPU? Apart from furmark?

 

[jonathann1818]

I'm going to adjust your image files so they display...brb.

Edit: nm, you must have posted the links to your personal view...you'd need to replace them with your publicly accessible URLs.

Is it just that no thumbnails are being displayed?

 

[rubix_1011]

Well, the URLs are all listed as 'my image' and if I click them independently, I can view, but they do not embed in the forum page. There might be a tag version to be used with public forums...this is what I use from Photobucket, but Imageshack might be different in that regard.

[quote]Well from what I have found there is an approximate proportion between actual power draw compared to TDP and as you probably know it depends on the heat of the chip (youre a guru so i think its pretty obvious you know). With more heat comes more resistance and therefore more power draw and more heat. This is the reason why the GTX 480 consumed so much damn power, apart from the architectural defficencies. When looking at the 580, the actual power draw is much less, sometimes by 30 Watts. This I am sure is the reason why my power consumption numbers were significantly lower than the theoretical numbers others gave me.

So I think the way to do it is to assign a realistic power draw value based on the realistic temperature of the chip and apply a typical reduction factor to it based on this.

Obviously the GPU and CPU usage plays a major role. I mean when do you actually fully utilize all of the GPUs and CPU? Apart from furmark?[/quote]

I appreciate the kind words, but you give me too much credit. I am just a guy that has been doing this for almost 10 years or so and have done a lot hands-on and a lot of reading. Most of what I have picked up over the years has been from forums, Skinnee/Martin and trial/error on my own. I don't consider myself a guru- just a guy that might be able to help newcomers figure out what they should do starting out instead of making mistakes and spending money on stuff they don't really want or need.

To your comment about actual power consumption at load- Furmark would be a good basis to determine as close to 100% GPU load whereas Intel Burn Test would be for a CPU. I know that between these two, you can isolate system issues and determine system stability. I also run a couple hours of BOINC with CPU+GPU at 100% allocation, just to see how it holds up. Also, some looped runs of 3Dmark Vantage or '11 on extreme should also point out any deficiencies.

 

[tinmann]

Best answer selected by tinmann.