My system

Do-it-yourself and user modifications...
flungo
Water Pistol Grunt
Water Pistol Grunt
Posts: 1
Joined: Sun Aug 28, 2011 10:12 am

My system

Postby flungo » Sun Aug 28, 2011 11:38 am

Okay, so I really need liquid cooling. My PC is currently hitting 90 degree core temps when running at 100% which is not good. haha!

Problem is I am new to water cooling and don't know how to set this up. I have some ideas about working it out but my system is rather complex and if I am having water cooling I want to go full wack! I will basically be cooling everything with water.

Here is my set-up (I don't have all this yet but I want to have a good plan for the complete system and this is what my system is capable of).

  • Corsair Obsidian 800D Case
  • Super Micro SUPER X8DAH-+F
  • 2 Intel Xeon X5680 CPUs
  • 12 sticks of RAM (I currently have 2GB in each but when the prices come down they will be upgraded to either 4, 8 or if I come into money 16GB each)
  • ATI Radeon HD5770 - Might be upgraded to something like the ATI FirePro V9800

First of all I would like to know whether my calculations have been correct. I have been using a rule I learnt in chemistry for working out enthalpy changes to work out how much heat would be gained at each point. By doing so I have worked out the gain on each components to be as follows:

  • Each CPU - 0.2 degrees
  • 6 slots of RAM - 0.05 degrees
  • Radeon HD5770 - 0.16 degrees
  • FirePro V9800 - 0.35 degrees
  • Hard Drive - 0.02 degrees

I based these calculations off a flow rate of 2.5GPM which I got from the technical data for the PMP-450S with 14 feet of water flow.

Based on those numbers I don't think its necessaries to use a manifold and have messy insides with loads of pipes! However if I run it as two loops, I can half all those values.

So based on what I have worked out these are the two loops I want to run with the calculated accumulate increase in temperature in brackets:

Reservoir>Pump>3x120mm Radiator>CPU1(+0.20)>RAM1(+0.25)>CPU2(+0.45)>RAM2>(+0.50)
Reservoir>Pump>1x140mm Radiator>HDD(+0.02)>GPU(+0.37)

My main question is will those radiator set ups be enough? The 140mm Radiator is at the base of the case so very cool air if that makes a difference.

However if you have a better setup that I could use I would be happy to hear.

In case it makes a difference to anyone opinion, RAM2 is right next to a vent where another 140mm radiator can be fitted to then run the liquid to any other component if you think that would work better.

ENTROPY
Water Pistol Grunt
Water Pistol Grunt
Posts: 13
Joined: Fri Aug 05, 2011 6:25 pm

Re: My system

Postby ENTROPY » Mon Sep 12, 2011 10:58 pm

By water cooling components that really don't need it (including by your own calcs) such as hard drive and RAM, you are essentially wasting some of the cooling capacity that could be directed to where its really needed not to speak of the added expense and piping you are concerned about.

When it comes to those low heat loads, you are just shuffling heat around for little gain--to water--to air generally still in the case. The idea is to use the water to conduct heat away from the hot spots like CPU and GPU.

Rather than spend money on hard drive coolers -- get an SSD for your boot disk which will run a lot cooler. use a fan for Ram if you think you need to cool it. Ram use to generate a lot more heat than it typically does these days.

Izerous
Rain Maker
Rain Maker
Posts: 1185
Joined: Tue Sep 16, 2008 10:34 am
Location: Edmonton, Canada

Re: My system

Postby Izerous » Tue Sep 13, 2011 7:52 am

A Bunch of things

HDD Cooling:
Very little value in liquid cooling HDD's especially in the 800D. Put SSD's in the bottom and put your storage drives in the hotswap area.

Ram Cooling:
Your ram blocks out of everything you want are the most restrictive, you are better of running both banks of ram in parallel if you actually want to liquid cool them. Typically I would suggest against even bothering to do the ram however with the board layout the airflow of the 800D and other factors it probably isn't a bad idea.

Video Card:
The one you want to "upgrade" to is a workstation card, not as good for gaming as it is for rendering things in CAD etc. Also I don't think there are coolers for the FirePro's.

Chipset(s) you seem to have ignored:
Because of the case and components you intend to use, you should take off the current chipset coolers and measure them. Your going to see bigger gains buy liquid cooling the chipsets than you are from liquid cooling either the ram or the hdd's even if you only use the universal low profile chipset coolers. Things like your chipsets especially with a dual socket board is where things start to heat up and the board only has passive coolers.

Loop Order:
I have more comments about the loop, however while the pump does generate a bit of heat it is the physical heart of a liquid cooling setup and I typically suggest you go rad > pump not pump > rad for that very reason. CPUs/GPUs etc all have quite huge heat tolerances they can handle the little extra a pump adds to a liquid loop. The pump if you overheat it well your whole cooling loop overheats.

About Bottom Rad:
About Putting a rad in the bottom. I found with longer PSU that have a Rad in the bottom with the power supply is a pain in the butt. Because you still need room for the cables to come out of the back of the PSU. Modular PSU's are worse because you need enough room to plugin new cables when needed.

Conclusion:
In the end I would be tempted myself to do something more like the following for a flow path instead.
Res > Pump A > Pump B > CPU 1 > CPU 2 > RAM A + Ram B (parallel due to flow restriction) > Chipset(s) (parallel if possible due to size) > 3x120 rad in the top mounting position > GPU > 140rad Back Position > Return

The 3x120 is more than enough for 2 CPUs + Ram + chipsets however the 140mm rad is a bit on the small side for such a high heat video card. Going series overall balances out some of the overall cooling power.
Corsair 800D
2xPMP-450S > Amd FX8370 > Asus CH4-EX > AMD 6990 > AMD 6990 > 2x1080 rads > return
32GB Ram

ENTROPY
Water Pistol Grunt
Water Pistol Grunt
Posts: 13
Joined: Fri Aug 05, 2011 6:25 pm

Re: My system

Postby ENTROPY » Thu Sep 15, 2011 8:43 pm

I just have one qualification in re: to Izerous' reply.

If your pump is adding any appreciable heat to your system, it is because you are operating too close to shut off head and the pump is churning. Otherwise a centrifugal pump will not heat the water. The flow rates are far too high and the friction is far too low as the diameter through the impeller and volute are constantly expanding out of the eye to cause the result energy changes (kinetic to flow). You folks running these series arrangements with other less than optimal configurations may run into that problem. In series head loss is additive (it works just like resistors) causing pump flow to drop off due to the increased head required to overcome series blocks. THAT LOST FLOW IS WHAT IS TERMED SLIPPAGE--it slips back to the suction side across the impeller--slips from the pump discharge back to suction side--churning that is which causes a lot of turbulence, increased head loss in the pump, and instead of creating energy changes as designed, the pump ends up just expending work which with no where to go and nothing to do on a Saturday night, expresses itself as heat. This won't necessarily happen if you put two or three things in series but it could--it depends on how well the pump was sized to the intended load in the first place.

Having said that, our pump still takes a suction on the rad which receives the return from the blocks. Not because the pump would heat up the water --that is no concern whatsoever but 1)to help maintain the minimum net positive suction head on the pump and 2) to provide the max pressure (flow energy) available to the blocks where distribution is most critical. Centrifugal pumps are prone to cavitation due to gases coming out of solution in the impeller--therefore the need to maintain a minimum net positive suction head on the pump suction. The impeller is spun at a very high speed flinging water down the vanes--this causes a vacuum in the eye of the impeller which helps draw water into the impeller. The low pressure can cause gases to come out of solution without adequate head on the suction (they aren't the bubbles you might see in your system unless it is gross--they will collapse in the higher pressure of the pump discharge--the source of noise--the air bubbles will erode your pump quite handily however--leading to early pump failure and further decreases in performance until then. This is adversely affected by increasing temperatures of the water being pumped (in the pump)--which is a good reason to take your suction off the rad. The second aspect is you want your pump discharging directly to the blocks to supply as much pressure to the blocks as possible for good cooling characteristics.

Just a Note: Parallel flow using splitters is far superior to series because 1) the head loss is no longer additive (the blocks are the major source of head loss); 2) a max delta temp is provided as the water entering each block is essentially that leaving the rad (as opposed to the warmer water leaving the previous block); and 3) you can monitor the return temps of each and redistribute how much flow (cooling capacity) each block receives. If you water cool the three big sources that could really benefit: CPU, GPU, and MB; then the tubing is very manageable.


Return to “DIY and Modding”

Who is online

Users browsing this forum: No registered users and 1 guest