Watercooling an Intel 4670k/4770k/4670/4770 CPU

[Ghost]

This is a general guide for water-cooling the 4770k, 4670k, 4770 and 4670 CPUs but can be applied to most other CPUs as well. I will try to cover as many questions in this guide as I can and try to give you a basic understanding in choice of radiators as well as effects of overclocking on water cooling.

 

Apologies for any grammar/typo/spelling mistakes but I have yet to proof read this guide. It is also incomplete. I plant to add a few extra sections to it and also of course do an extra "section" concerning the 4670k.

 

Waterblocks/AIO Compatibility

 

The socket for the 4770k (LGA 1150) has the same mounts as the older socket used for Sandy bridge and Ivy bridge chips (LGA 1155). This of course means anything that fit the old sockets will fit the new socket. This includes AIO coolers such as the H100i, H80i, H220 and any air coolers and of course all water-blocks.

 

What you could expect is different jet plates that could be more optimized for the new chip but this will make  negligible differences in terms of heat.

 

Power Consumption

 

The TDP (Thermal Power Design) of the 4770k is 84W which is slightly less than the TDP of the 3770k which was 77W. What this means is of course; you need to dissipate more power (heat) coming from the CPU. This may not be very significant but if we take it as a percentage the 4770k draws 110% of the power of the 3770k which means 10% less cooling needed. Of course you can't add 10% of a radiator off but it does mean you will need to have higher fan speed at similar temps.

 

In terms of the new improved "low-power state", from my understanding this will not benefit the power-consumption much because it only activates when the computer is in sleep mode, at best it will mean your radiator fans can be almost switched off when the computer is in sleep mode but do not quote me on that. Please refer to my paragraph on overclocking for more info on the "low-power states" you get when the CPU is using less voltage.

 

Overclocking

 

I have not yet seen power consumption at overclocks so this section is TBC. That is all that's important during overclocking in terms of cooling. From school you may remember the formula Power=Current*Voltage . During overclocking you increase voltage to gain stability. and because P∝V when you increase the voltage power is increased. This is power you later need to dissipate while cooling the CPU.

 

Radiators

 

The amount of radiators you need is proportional to the TDP of the item you are cooling, in this case a 4770k. A problem with this is that radiators are not marketed by TDP. This is because power dissipation depends on the heat of the coolant and the air, the speed of the pump and RPM of the fans. And these variables affect the power dissipation of a radiotor a lot.

 

I am going to grab a quick example of a radiator power dissipation graph;

 

All credit for the graph go to martinsliquidlabs.com

 

Now this is a graph for a Alphacool NexXxoS ST30 360 Radiator. From the start you can see that if you are running the CPU at stock this radiator is (I hate to use the word but;) "overkill" for this situation. It could comfortably cool the 4770k with fan speeds of 500RPM.

 

I have to warn you not to take this as a "I need my fans to run at 500RPM to cool this CPU with this radiator". This test was taken with a ΔT of 10C between the water temp and the air and with a Laing D5 (MCP35X2) running at 40% of its speed and using Titan Kukri PWM fans.

 

Use this as a comparison only.

 

Unfortunately I am unable to source graphs for more relevant (120 and 240) radiators but there we go.

 

An important thing to note is that running you fans any faster than this is simply creating noise. At this point the radiator has cooled the liquid down back to room temperature. If you were to have a program monitoring your fan speeds ideally you would want to to match the air and coolant temp at the point of the coolant entering the CPU block with minimum fan speeds. But I have yet to see fan control software this advanced.

 

I've gone off on a tangent here but hopefully you have a better understanding in what is important when choosing radiators instead of just listing numbers as I have done before.

[Hyydrah]

Nice guide, you should include the 4670/4770 in case some people didn't know they can be used as well

[Ghost]

Nice guide, you should include the 4670/4770 in case some people didn't know they can be used as well.

If they are watercooling they will most likely have a k series CPU.

[Hyydrah]

If they are watercooling they will most likely have a k series CPU.

I know, and they should, but they can still use a non-k CPU and benefit from watercooling

[Ghost]

I know, and they should, but they can still use a non-k CPU and benefit from watercooling.

Fair enough, added it.

[felixbearpig]

tdp for 3770k is 77 and thanks for this 

[Ghost]

tdp for 3770k is 77 and thanks for this 

Thanks. Its fixed now.

 

I was wondering about that but turns out I misread my source.

 

The CPU TDP is 77W. The 95W on the box is just a specification for the thermal envelope the stock heatsink and fan need to be able to fulfill.

[PCBudgetSolutions]

Thanks. Its fixed now.

 

I was wondering about that but turns out I misread my source.

It's funny that despite it putting out less wattage than ivybridge it runs hotter.

IVY: 77(4.6/3.8)*(1.4/1.2)^2 =126 watts

HW: 84(4.6/3.8)*(1.2/1.2)^2=101 watts

 

They really screwed up somewhere.

[Ghost]

It's funny that despite it putting out less wattage than ivybridge it runs hotter.

IVY: 77(4.6/3.8)*(1.4/1.2)^2 =126 watts

HW: 84(4.6/3.8)*(1.2/1.2)^2=101 watts

 

They really screwed up somewhere.

Not sure about what calculation you did there :o

[PCBudgetSolutions]

Not sure about what calculation you did there :o

TDP is current clock/stock*current voltage/stock^2?

[Ghost]

TDP is current clock/stock*current voltage/stock^2?

The way CPUs work I am sure current changes with voltage because V=IR and the overall resistance should stay the same although I am not even sure of that. That's why I said P is proportional to V.

Not sure why the clocks would make a difference in Power calculations.