Originally Posted by Jakusonfire
It really isn't so much that 60% of cooling is done by rads and the rest by something else ( If my rads were only doing a bit over half my cooling I would want better rads ) its much more that people often over estimate just how much heat that components put into the loop to begin with.
As an example; The 240 testing I mentioned earlier with battlefield 4
This is with the CPU at 4.6 GHz and the GPU at 1200 Mhz
The power guages at the bottom measure the actual heat input from both devices using multiple water sensors combined with the flow sensor. The system has started to cool down slightly but you can see the max output recorded at the far end of the blue coloured arcs on each guage ... the blue arcs are the max and min recorded over the last 15 minutes.
As you can see ... about 210W for the GPU and 50W for the CPU. Sure each one can put out a bit more than that if running an artificial stress test but configuring a system based on that is just confusing.
These sensors aren't super accurate but do at least give an indication.
Perhaps it would be better said ... which I do sometimes but it's too much dang typing
.... that ya want at least 60% of ya "theoretical heat load" to be handled by the rads .... I use 6 heat sensors and power readings (at the wall) and I'm seeing like 645 watts under Furmark which only puts about 40 watts on the CPU.... that quickly drops down to 112 or so when ya stop.... so "doing the math" ....645 watts * 90 % efficiency says PSU putting out 580.5 watts . Take away the 40 watts for the CPU and say 20 watts for the pump and 20 watts for MoBo / RAM and I'm down to the cards are pulling ~ 500 watts at stock .... Guru 3D puts the power consumption at 254 watts for the GTX 780s so looks pretty accurate.
Yes, in your test, gaming, I don't expect that it would come anywhere near that level..... as for the confusing, the way I look at it is ya have to base everything on a standard .... typical usage is not a good one as everyone's typical usage is different. Calculating the theoretical heat load will always produce the same number no matter what you use. Then, it's simply much easier to my mind to use a "usage coefficient" to adapt to your usage patterns. I have been finding based upon what data I have collected that 60% works very well. I also like to use stress tests because it's nice to know the upset limit .... I pick power supplies based upon the fact that I will almost never see the max calculated load, but i want the unit to operate quietly and close to its max efficiency point (50% loading) .... so same thing here.
I design for 10C at the maximum load but I'm don't ever expect to operate there for any significant amount of time. I want my cooling to handle it at max fan rpm, but I don't want to be hearing my fans at max rpm day in and day out. If my 1250-ish rpm fans ever break 850 rpm outside if stress testing, I'm gonna be upgrading something.
BTW, I'm starting to collect data from various sources on rads, pump speeds, fan speeds, temps, etc at various loadings .... would you be interested in contributing to the database ?
Originally Posted by Wingwire
I've thought about what you guys have said, and made this plan using your suggestions
I decided to go with an EVGA 780 Superclocked, an XSPC Razor 780 GPU block and backplate and an XSPC RayStorm CPU block. What do you think?
I gotta say I am not a fan of the EVGA SC also as it's the only factory overclocked 780 that uses the reference PCB and VRM. As the VRM is the hottest component on a WC card, that may be the limiting factor in ya OC. The only thing that differentiates the EVGA SC series from the reference board is the cooler .... which you are going to remove.....so you are payting more for a component that you won;t use. So if ya want EVGA, might as well save money and stick with the reference card. The EVGA Classified and MSI Lightning have a custom PCB and VRM as does the Asus DCII, MSI N series and Gigabyte Windforce. The Classified and Lightning have a few extra features that the other's don't have hence their cost premium.
http://uk.hardware.info/reviews/4639/10/nvidia-geforce-gtx-780-asus-vs-evga-vs-inno3d-vs-msi-conclusionEdited by JackNaylorPE - 3/5/14 at 9:36am