name="tulanthoar" url="/t/1314452/mounting-screws-and-also-a-noob-question/0_100#post_18329918"]It depends on what kind of performance on what type of tasks you want. For me, I'm a gamer, so I almost never use my gpu and CPU at the same time. You can always try it out, and if you aren't getting the performance you need then buy a new rad. Rule of thumb is 10 degrees C for your delta t. Using that, you can calculate the thermal dissipation capacity of your radiator. People will tell you to use 2x11120 mm sections for your first block and add 120 mm for each additional block. I personally think this is overkill, as 4xXSPC RX120 (my planned radiator) should cool 800 watts at 10 delta t with 1500 rpm fans. If the benchmarks are correct, that is more than my PSU can even drive, so my 560 to sli and CPU won't even be capable of producing enough heat to need 4x120 mm sections. Assuming the benchmarks are correct of course. From what I've read, using indigo xtreme TIM makes a good 2-3C difference as well as almost guaranteeing a perfect mount. Much more cost effective to get the indigo xtreme first and then see if you need more rad. Assuming the benchmarks are correct of course
I apologize, but while I believe that this was intended to be helpful, I have to disagree with a ton of what you said...
First off, you are a gamer so you DO use your CPU and GPU's at the same time, but neither is ever close to fully stressed ("stress"being absolute, sustained 100% utilization) as pretty much only P95/LinX/IBT will be the only time that the VAST majority of enthusiasts PC's see full loading "bilaterally", although F@H can take a significant amount of power as well. So when you are gaming, you are utilizing likely between 20-90% of the CPU (depending on the engine) and actually probably similar numbers for the GPU's. The GPU's are generally MORE stressed than the CPU, but they are absolutely both under load.
Second, the "Rule of Thumb" is not "10C for dT"; instead, there are general guidelines about how a given loop should perform based on the intended use. In genera , anything BELOW 10C dT is considered "decent", but the whole point of liquid cooling (aside from those who use it solely for noise reduction) is to get AWESOME cooling. Thus, most people shoot for closer to 5CdT than 10 as it is simply better in every way (and often a ssmall amount of extra money, relative to the money already necessary for a good loop). It's when you get to the 1-2C dT level that the cost begins rising exponentially for performance gains that decrease exponentially, and generallyunless you are looking to set records, have a giant Trust Fund, or are legitimately going to be adding 2-3x MORE GPU's to the loop in the near future (as in you have at least 80% of the costs for the cards, but need an extra month or something), it simply isn't worth chasing.
Also, the advice given by most people to "buy 120x1 for each block plus (at minimum) another 1x120 for headroom" is actually about as specific as is ppossible to get when giving advice on cooling capacity, as it is almost always enough to get the loop to <10C dT. Sure, it doesn't apply to everyone, but if a person is concerned about the requirements of a certain system's cooling, that is why we have the always wonderful OCN Forums
*(I am not saying any of this to "call you out" or anything like that, so please don't misinterpret this post as I simply want to address what I feel to be insufficient and/or incorrect information, as I don't want anyone to go off of incorrect (or outdated) information when planning a loop.)
Now, I absolutely agree that it is important to figure out one's actual thermal output maximum, and there exist multiple ways to easily calculate a rougheestimat , but due to Murphy and his Law, it's best (in fact, unless you are quite experienced, almost necessary) to add extra rad space. This is in part due to the vast difference between some chips heat output and other's.
I personally think that it's best to put as much headroom into a loop as you can afford (or have space for), as overclocking introduces an inherently unpredictable variable. Also, many people will upgrade by adding more to the loop over time, and I believe it's better to have the extra headroom ahead of time. At the very least, having an extra 120mm worth of rad space is better than not.
The amount of power your PSU can produce is not a good indication of how much cooling one needs, as two loops using identical components on identical systems can perform very differently if one loop has 480mm of radiator while the other has 840mm, with the latter producing a significant decrease in Delta T and even doing so more quietly. Generally, 120mm worth of rad space is sufficient to dissipate 125W of heat, so if you're cooling 375W a 360 rad will not overheat your components but it also will not perform much if any better than extreme air cooling. However, if you have enough rad space that each 120mm section is cooling just 75W of heat, you will see a huge performance difference. I don't believe it is a good idea to base a loop's rad area exactly to match the heat of the components for this reason.
Thus, for a loop cooling a 95W TDP CPU and a 175W GPU, I would add 40-50% for overclocking (roughly), giving 150W + 275W for 425W. This seems like a lot, but it is always, always better to overestimate than underestimate the cooling needs. If one were to follow the 125W-to-120mm rule, we would have 120x3.5 + 120 of rad area, rounding up to 120x5. Now, because I have overestimated the output, it is certainly possible to run the loop on a single120x3 rad , but overclocking will be more limited and temps will not be great.
The same loop, but with a 480 or 360+240 (or 420/360+240/280) will produce better temps, and will allow for almost silent operation (you can go either way).
For a loop containing an overclocked CPU and GPU, using good blocks and a strong pump (MCP35X2>MCP35X>MCP355=D5Vario>D5-Basic=MCP350), I would recommend a minimum of an extremely high end 360 rad (UT60, Black Ice GTX, AX360, Monsta), but ideally, and to give morefflexibility in thickness of the rads, I would recommend a 360/420 such as an EX360/420, XT45, etc, and a second smaller rad such as a UT60 240, BI GTX 240, etc. I would also recommend push pull fans as well, as I feel that the benefits outweigh the detrimental aspects (cost and slightly higher noise). Have an MCP35X pushing the water, a great CPU block like the Raystorm or Apogee HD, an excellent GPU block (Heatkiller, Aquacomputer, etc), and most importantly good powerful fans (Bgears Blasters 140mm, Koolance 120x25mm or 120x38mm 2600rpm 120-size fans for all-out performance or AP15s for quiet with less sacrifice in performance than most of the so-called "quiet fans"). Going push-pull on both rads can add 10-to-30 percent performance depending on the fans and their speed, as well as the radiator(s) fin density.
I would also really recommend against using 4 separate 120mm rads, as a single 120 has the same amount of restriction as a much bigger member of the same rad type (within a standard deviation), which is 0.3psi (I believe, except for HWLabs Black Ice, which are extremely restrictive). So while a single 480 rad will have 0.3psi restriction, 4 120x1 rads of the same type will have 1.0-1.2psi of restriction, and in fact will perform noticeably worse! The bigger the rad, thebbetter it can dissipate heat, and the equivalent number of smaller ones will not do this.
Lastly, while good TIM is absolutely critical to getting the best performance fromyyour loop, you really don't stand to benefit so much as you may think. In fact, I would suggest that you will be lucky to see 2C if you have been using a high quality TIM up to this point. Personally, the liquid metal types of TIM are fascinating but the cost of $10 per mount is not worth the 0.893C difference from the absolutely incredible PK1 (which can be used for GPU's, too).