Help a watercooling noob with his i7 930 and GTX470

thadman

Member
Nov 23, 2009
25
0
66
I currently have a Gigabyte X58A-UD3R with an Intel i7 930 and a Corsair H50 cooler. These components are installed in a Corsair 800D case. I recently overclocked my processor to 4.2GHz, here are my system settings.

CPU Vcore = 1.2875V (level 2 load line calibration)
CPU clock drive = 800mV
DRAM voltage = 1.680V
QPI voltage = 1.375V
base clock = 200MHz
CPU frequency = 4200MHz (21x multiplier, turbo disabled, HT disabled)
uncore frequency = 3200MHz (16x multiplier)
RAM frequency = 1600MHz (8x multiplier)
RAM timing = 8-8-8-24-2

The BIOS is reporting that the DRAM voltage is 1.648V (although it is set to 1.680V), the CPU Vcore is 1.284V (although it is set to 1.2875), and the +12V rail is 12.365V.

Although I feel comfortable with this overclock, I am not comfortable with the heat. I ran IntelBurnTestV2.50 for 2 hours and the CPU peaked at 83*C and averaged 78-80*C. This heat will significantly degrade the life of my CPU and needs to be dealt with. As a result, I'd like to explore watercooling.

I currently have an EVGA GTX470. This thing gets HOT! It idles between 55-65*C and peaked at 93*C while running EVGA OC Scanner. However, it is currently being RMA'd. I'm going to try and upgrade it to the GTX470 Hydro Copper so I don't have to install an aftermarket GPU water block and void my warranty. I will be adding another GTX470 in the future.

I have a Corsair 800D case and I absolutely love it. It is a dream to work with and I'd really prefer not to modify it.

Would it be possible to use a single loop for OC'd i7 930, X58 chipset, and SLI'd GTX470s? I'm shooting for a 5*C temperature delta (Water Avg - Air In). I really think it would benefit from separate loops, but I'm not sure where I could place an additional radiator(s) without modifying the 800D. I'd also prefer not to use the Swiftech radbox. An external radiator would take away from the sleek look of the 800D.

I've read about every review / guide on skinneelabs and the stickies on XtremeSystems forums about 5x, but I'm still not sure what would be the best WC components.

http://www.skinneelabs.com/reviews.html
http://www.xtremesystems.org/forums/showthread.php?t=202394


Here is what I'm currently considering:

CPU water block: Swiftech Apogee XT extreme
http://www.newegg.com/Product/Produc...82E16835108113

I have heard that this water block is incompatible with some Gigabyte motherboards, specifically the EX58-UD3R. My motherboard (X58A-UD3R) is not mentioned on the incompatibility list, but I feel its layout might be quite close to the EX58-UD3R. Does anybody know if this water block will fit on my motherboard?


NB water block: Swiftech MCW30
http://www.frozencpu.com/products/20...ter_Block.html

Is this a good choice for a northbridge water block? I'd like to use all Swiftech water blocks if I can (uniformity in the design). The chipset gets extremely hot due to the poor air circulation properties in the 800D. When I had the i7 OC'd to 3.6GHz, the CPU reached ~55-60*C and the chipset reached ~60-65*C. This is too hot. I'd prefer if it were below 50*C at all times (specifically ~40-45*C).


Pump: Swiftech MCP355
http://www.newegg.com/Product/Produc...82E16835108097

Is this a good choice? I heard the Apogee XT extreme is a comparatively restrictive CPU block and believe the MCP355 may be better for these situations compared to the MCP655. Is this a fair assumption? Should I use multiple pumps? I heard these have reliability problems compared to the MCP655. Could anybody offer further insight on this?


Pump Top / Reservoir: XSPC Acrylic Reservoir DDC Top
http://www.frozencpu.com/products/72..._Fittings.html

Is this the best top? Where should I install this in my case? I'm absolutely anal when it comes to maintaining a clean case (the reason I purchased the Corsair 800D was for the cable management). Is this significantly better than bay reservoirs? Is there anyway to install bay reservoirs without removing the drive covers?


Radiator: XSPC RX360 (with 3x push Gentle Typhoon AP-15s) or Black Ice GTX 360 (with 3x push or 6x push / pull Scythe Ultra Kaze 3000s)
http://www.frozencpu.com/products/85...?tl=g30c95s161
-or-
http://www.frozencpu.com/products/53...?tl=g30c95s161

I live in a small apartment without central air. My room is ~100 square feet. I'm going to be purchasing a window mounted AC unit in the near future. I'm not sure how loud the AC unit is, but it will set the noise floor for my room. I also have two large fans operating in my room. I don't seem to mind them on high, but I'm also not sure how loud they are or how they compare to PC fans.

Are the Scythe Gentle Typhoon AP-15s the best fans for a low-medium flow radiator?

Are the Scythe Ultra Kaze 3000s the best fans for a high flow radiator?

I will of course be using a fan controller, but I'm not sure which one to get. Could anybody recommend a fan controller? How bad are the Scythe Ultra Kaze 3000s with a fan controller? Are they tolerable?

My Corsair 800D can fit a triple 120mm radiator at the top of the case. I can't decide between the XSPC RX360 and HWlabs GTX 360. I believe it will ultimately depend on the heat dissipation requirements of my system.


I used the CPU OC power calculator to determine power dissipation.

http://extreme.outervision.com/tools.jsp#cpuoc

At 4.4GHz (stock core voltage = 1.175V, OC'd core voltage = 1.3V), I'm getting 250 watts. However, I'm not sure how accurate that is.

The bit-tech review of the i7 930 shows it loves power (more than that calculator is suggesting)

i7 930 @ stock / idle = 124 watts

i7 930 @ OC (4.3GHz) / idle = 176 watts

i7 930 @ stock / load = 216 watts

i7 930 @ OC (4.3GHz) / load = 396 watts

http://www.bit-tech.net/hardware/cpus/2010/03/01/intel-core-i7-930-cpu-review/6

Are these numbers referring to system power or CPU power? What numbers should I go by?


I'm not sure how to determine the power dissipation of my GTX470. The [H]ardOCP review of the GTX470 shows the system pulling ~430 watts from the wall.

http://www.hardocp.com/article/2010/03/26/nvidia_fermi_gtx_470_480_sli_review/7

The PSU used in the review was the Thermaltake Toughpower 1200W PSU, which is 87% efficient. We might assume that system (GPU + other components) was receiving ~375 watts from the wall. I'm not sure how many of those 375 watts are being dissipated in the GTX470 compared to the other system components. Could anybody offer further insight on this?

Any thoughts on how to estimate the power dissipation of two GTX470s in SLI?


I have absolutely no idea how to estimate the power dissipation of the chipset.


I believe I must also take into consideration the heat dump associated with the pump I decide to use.


Of particular consideration is how we apply these estimates. We might simply sum the power dissipation of all the system components in the loop at 100% load and use that to determine the required radiator and fans. However, I'm not sure if this is representative of the real world. I believe all system components at 100% load is seldom observed.

I'm using this system for moderate gaming (Crysis), intensive multiphysics simulations (ANSYS, Cosmol, etc), Matlab, C programming, and miscellaneous engineering software. I'm currently using a Dell U2410 (1920x1200) as my primary monitor, so we might assume the system is going to be GPU limited (GPU @ 100%, CPU @ ?%) running games. Most simulation programs use the CPU for calculations, so we might assume the system is going to be CPU limited (CPU @ 100%, GPU @ ?%) in those application. However, once I familiarize myself with CUDA programming, I'll be using the GPUs for multiphysics simulations. I assume the system would be GPU limited in that application, but I am not sure of how significant the load on the CPU would be.

I will be adding two more monitors (considering the Dell U2311Hs) in the near future. I'm currently running two monitors. I'm not sure how that will effect the GPU load with traditionally CPU limited programs. I don't assume it will be a problem, but I'd like to cover everything.


With all of this in mind, how can we estimate system power dissipation in real world applications? I enjoy pushing my system as far as I can (within reason) using benchmark software (Prime95, IntelBurnTestV2.50, etc) to see what I can achieve, but its not THAT important.


A RX360 with 3x Yate Loon D12SH12 fans @ 1815RPM will dissipate ~280 watts before exceeding a 5*C temperature delta (this is what I'm designing for). It will dissipate ~580 watts before exceeding a 10*C temperature delta. I assume the performance of the Scythe GT AP-15s on this radiator is similar to the Yate Loon D12SH12s. Is this a fair assumption?

http://skinneelabs.com/xspc-rx360.html?page=4


A GTX 360 with 3x Scythe Ultra Kaze 3000 fans @ 2804RPM will dissipate ~580 watts before exceeding a 5*C temperature delta. It will dissipate ~800 watts before exceeding a 7*C temperature delta (empirical data does not exist beyond 800 watts) I'm not sure how much this will increase by adding 3 more fans and operating them push / pull. Any thoughts?

http://skinneelabs.com/hwlabs-gtx360.html?page=4

I don't think there is room for push / pull 38mm fans AND shrouds in the top of my 800D.


Do you think the RX360 would be sufficient for an OC'd i7 930, X58 chipset, and SLI GTX470s in a single loop?


I could probably install an additional 140mm radiator on the internal rear of the case (where Corsair recommends the H50 to be installed). Here is a photo (not my system, but it illustrates the idea).

0.jpg


However, I'm not sure if it will interfere with the airflow to the 360mm radiator at the top of the case. There is about 4.8 inches from the top of the 800D to where the H50 is currently installed and roughly the same (maybe a little bit more) to my DDR3 modules.

I dont't think I could install an additional 140mm radiator if I used the GTX360 with push / pull 38mm fans.


There are 3 140mm radiators available on frozenCPU that might work.

-Black Ice GTX 140
http://www.frozencpu.com/products/90...?tl=g30c95s929

-Black Ice SR1 140 (low air flow optimized)
http://www.frozencpu.com/products/10...?tl=g30c95s929

I haven't seen any scientific reviews (like the ones on liquidlab) of either of those radiators, but I assume their performance should be similar to their bigger brothers.

-Koolance 140mm High Flow Radiator (30FPI)
http://www.frozencpu.com/products/10...?tl=g30c95s929

This is probably very similar to its bigger brother as well (requires 1800+RPM fans)


If I went with a 360mm radiator and a 140mm radiator, should I put them in series or in separate loops? How would you distribute the system components amongst the radiators and how would you setup the loop order?


I read the review of the quick disconnects on liquidlabs. The Koolance VL3Ns look nice. However, I'd like to hear members experience using quick disconnects. Are they worth it? Have you ever had the liquid left in the disconnect spill and damage part of your system?

I'm sure I'm forgetting a lot of stuff. Please suggest something if you feel it should be considered.


Thanks,
Thadman
 
Last edited:

smackababy

Lifer
Oct 30, 2008
27,024
79
86
I'd go with the Thermochill PA120.3, I think that is still the best radiator out there, but requires 5/8G threaded barbs.

A good rule of thumb is use a power supply calc (http://extreme.outervision.com/psucalculatorlite.jsp), enter only what will be in your loop (CPU, GPU, Pump) then OC and what not and test it. That is around the heat you need that will be produced. Figure that into your loop. And modding your 800D to fit another res is not hard. I have another 120.3 in the bottom of mine, but I don't have a longer power supply.
 

thadman

Member
Nov 23, 2009
25
0
66
Thanks for the quick response:D I know this post is a bit overwhelming in length, but the topic of watercooling is a bit overwhelming in complexity.

It seems to me that we could get a definitive answer by complicating the problem a bit.

I haven't studied heat transfer significantly (I'm EE), but I believe a radiator will have a heat transfer with respect to flow rate relationship. I suspect that heat transfer peaks at a specific flow rate and if the source impedance (fan) was matched to the load impedance (radiator), we would have maximum power transfer.

Could the static pressure of the fan be taken analogous to voltage? If so, we might be able to define the fan's required static pressure for maximum heat transfer. We could then define the criteria for optimum WC loop as noise (maximum heat transferred without exceeding "X" dB) or total power. If noise, then we would start with a search for a fan that meets our criteria and then determine the optimal radiator for that fan. If total power, we would start with a search for a radiator, find the fans which meet the minimum static pressure requirements, and then differentiate amongst those fans by cost or noise.

With regards to pumps, we have two options
1) MCP355 (exhibits higher head pressure, but has a lower maximum flow rate)
2) MCP655 (exhibits lower head pressure, but has a higher maximum flow rate)

I believe pressure is indirectly related to heat transfer and flow rate is directly related to heat transfer. The pressure simply determines the flow rate. As a result, pressure doesn't really matter. We just need enough pressure to achieve the specific flow rate associated with maximum heat transfer. I believe if we modeled the system as a fluid circuit and optimized it (made sure series pressure drops did not drop below critical pressure), we would dissipate the maximum amount of power for a given radiator.

Any thoughts?
 

nitromullet

Diamond Member
Jan 7, 2004
9,031
36
91
With regards to pumps, we have two options
1) MCP355 (exhibits higher head pressure, but has a lower maximum flow rate)
2) MCP655 (exhibits lower head pressure, but has a higher maximum flow rate)

You have to take into account a few other considerations:
1) MCP355 - smaller, easier to fit into a PC case, but somewhat unreliable track record
2) MCP655 - big, difficult to mount, but built like a tank

Basically, If I was going to run DCCs I'd run more than one for redundancy, and put some sort of flow meter in my loop so I'd know if/when the pump stopped working.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,905
3,292
126
I believe a radiator will have a heat transfer with respect to flow rate relationship. I suspect that heat transfer peaks at a specific flow rate and if the source impedance (fan) was matched to the load impedance (radiator), we would have maximum power transfer.

Could the static pressure of the fan be taken analogous to voltage? If so, we might be able to define the fan's required static pressure for maximum heat transfer. We could then define the criteria for optimum WC loop as noise (maximum heat transferred without exceeding "X" dB) or total power. If noise, then we would start with a search for a fan that meets our criteria and then determine the optimal radiator for that fan. If total power, we would start with a search for a radiator, find the fans which meet the minimum static pressure requirements, and then differentiate amongst those fans by cost or noise.

With regards to pumps, we have two options
1) MCP355 (exhibits higher head pressure, but has a lower maximum flow rate)
2) MCP655 (exhibits lower head pressure, but has a higher maximum flow rate)

I believe pressure is indirectly related to heat transfer and flow rate is directly related to heat transfer. The pressure simply determines the flow rate. As a result, pressure doesn't really matter. We just need enough pressure to achieve the specific flow rate associated with maximum heat transfer. I believe if we modeled the system as a fluid circuit and optimized it (made sure series pressure drops did not drop below critical pressure), we would dissipate the maximum amount of power for a given radiator.

Any thoughts?

1. Yes the capacity of the rad is based on the fans used in relation to the flow rates.
The radiator will always be limited on the disapation side.
As you get close to ambient, your gains will drop as fast as a log graph.
Meaning its exponentially harder and harder to get closer to ambient.

Ambient meaning water temp in relationship to the air temp your radiator is pulling.

2. No a fans static pressure has nothing to do with voltage. Well minus more voltage = faster spinning fans = more pressure :p
I think your thinking of PWM.
But seriously this has more to do with the fan blades and in which it cuts air.
Also in the way it pushes air.
Static value is the value in which the fan can push air though restrictions.
CFM is in open scenarios.
Its true a good rad needs static, but as you get to low FPI rads, you want something quiet yet with good static numbers.

You wont need to look any further then a Scythe GT for a XSPC RX or Thermochill PA series rads. Fesser included as well.

3. no there is no concept of pressure. You are dealing with a closed loop.
Any form of pressure you get inside a closed loop are signs of problems.
Although ur res is prone to drop a bit when your pump is on due to suction and vacuum.
Water is not compressible, but the air in your res sure is.

4. good luck on your project.
I see that H5O kit only made you greedy and wanting more.

2 cents... if your going full blown custom, do not mickey mouse that H5O.
 
Last edited: