Keeping our gaming systems cool is just as, if not more important, than actually playing on them. Lets face the facts, how can we gamers play on a system that overheats everytime we play on our gaming system? An overheating gaming system can and will do one of the following or all of them; it can cause lock ups due to throttling (when our components down clock themselves in order to keep them from frying themselves), slow downs due to component throttling (not as drastic as lock ups but just as annoying), rebooting, crashing, artifacting (when we gamers start to see rather unusual lines, black triangles, or odd colors being displayed on our screens), to finally having complete and total system failure (this is when a component completely dies or gets fried). Water cooling has been around for just as long as there have been computers, and today we gamers can choose from a variety of different manufacturers, models, and makes for whatever we may need a water block for. Not to mention there are so many different ways we can run our water cooling loops, it is not even funny.
With the ability of us users to add even more and more video cards to our gaming systems (PC Only), adding even more heat inside of our gaming systems that can cause problems to happen more frequently (listed above) just from this extra heat we added. In a lot of cases the use of an air cooler is not an option, because some motherboards do not have the correct spacing to allow for the larger newer video card air coolers out today. The only option we have as gamers is water cooling. Now here comes the real question: what if you do decide on running water cooling on your gaming system, on multiple video cards, how do you run them? Well there are only 3 different ways we can run our water cooled multiple video cards, these are in Series (or daisy chaining), in Parallel, or we can run each video card water blocks in their own separate loops (sometimes this is not an option). In this article I am going to demonstrate two of the most common ways we can run our multiple video card water blocks, that is, in Series, and in Parallel configurations only. Now before I just jump right into the “results of my testing”, I need to show you what is the difference between running water blocks in Series, and in Parallel configurations.
The Different Configurations
These next few images are for references purposes only. I know, I should not give up my day job, because as an artist, I will be going awfully hungry.
This is a typical single water block configuration. Water enters the water block (blue) then goes through the water block (Green) then exits the water block (red) on its way to the radiator.
This is a Series configuration (or Daisy Chaining the water blocks). Water enters the first water block (Blue), goes through that water block (Green), exits the first water block and then enters the second water block (Purple), goes through that water block (Green Again), then exits that water block on its way to the radiator (Red). This configuration will give us the most restriction on our flow rates, because it gets reduced by the first water block, and then gets reduced again by the second water block.
This is a Parallel configuration. Water enters the blocks, then gets distributed across both water blocks evenly (Blue), goes through both water blocks (Green), and it exits the two water blocks evenly (Red) on its way to the radiator. This should technically keep our flow rate the same as if we were only using one water block because this configuration evenly distributes the water across both water blocks.
Time for me to put these three types of configurations into practice. The Water Blocks I am going to use are Koolance 697 for AMD’s 6970 reference video cards. The interconnects I am going to use to tie the two video cards together are none other then Bitspower Crossfire/SLI water block interconnects.
Testing with Two AMD Radeon 6970′s with Koolance 697 Water Blocks.
This portion of the article is me demonstrating on the two major configurations when we decide to run our multiple video card water blocks. I will be showing my “Base Line” configuration of just one video card water block. This is so I have an idea of any changes that running the water blocks in Series and in Parallel will have. All testing was performed with out a radiator, as always, your results may vary greatly from my own.
This is how I measured the flow rate of each specific water block configuration. I used a 300 GPH (Gallons Per Hour) pump and measured its actual flow rate by pumping 1 gallon thru it by using a stop watch. I did this three times and took the average of the three runs per configuration. There will be a small margin of error, typically this was roughly 0.03 seconds (or three hundredths of a second). The bucket you see here is a 3 gallon bucket, from line 2 to line 1 is exactly 1 gallon of water, from line 1 to the bottom of the bucket is another gallon of water. This extra gallon of water is to keep the pump from sucking in air during my “Highly Scientific” (well not really) testing.
Here is the “Base Line” configuration, this is just one video card water block being run.
What the water flow looks like coming out of the water block, as we can see, there is a bit of restriction from the Koolance 697 water block.
Now before I just jump right into the dual video card configurations ,I felt I should share a little water cooling tip. I opted to use the newer Bitspower water block interconnects for the two video card water blocks. This part is not an absolute need, but it does allow easier assembly of the video cards. I removed the two rubber O-rings from the threaded portion of the interconnect, then used some silicone heatsink paste and smeared it over the O-rings to keep these from drying out prematurely, and also it makes the plastic interconnect tubes slide in a lot easier. Then I put the rubber O-rings back in their respective spots of the threaded portion of the interconnects.
Time for me to look at how the flow rate would be once I hooked up two video cards in a Series configuration.
Lets see what the flow rate looks like. Wow, talk about adding a lot more restriction to the water cooling loop. Just by adding a second video card water block in a Series configuration has seriously hindered the overall flow rate of the water.
Turning our attention to a Parallel configuration with our water blocks. This time instead of entering one water block, then coming out of the second water block, we are coming in and out of the same water block. the restriction from the first water block will cause the water to enter the second water block and then it distributes the flow rate evenly over the two video cards. As we can see all I needed to add was an extra Bitspower interconnect to my configuration.
This is the end result of me running the water blocks in a Parallel configuration. It appears that my flow rate is about the same as I had when I ran just a single video card, and my flow rate is double that of when I ran the two video card water blocks in a Series configuration.
All testing was performed with black PVC 1/2″ ID x 5/8″ OD tubing.
Up first with our testing is the flow rate. This is the amount of seconds it took to pump 1 gallon of water through each one of the above listed items, or configurations. The lower the time it took to pump 1 gallon of water, the better the result. The pump testing was performed with just the pump itself, with a 2.5 foot of 1/2″ ID 5/8″ OD hose. It only took the pump about 16 seconds to pump 1 gallon of water. By hooking up 1 single video card water block to the loop, we can see that it seriously hindered the overall flow rate of the water cooling loop. Looking at the 2 video card water blocks in a Series configuration, really brought the flow rate of the water to its knees. Now the Parallel testing gave me a rather interesting result, I was expecting to have the exact same flow rate as I did with just one video card water block, but instead it increased the overall flow rate. Now I did run this particular configuration a few more times (full tear down and rebuilt it) to ensure my results were accurate, I kept getting the same result each time.
Now to get our gallons per minute we take how many seconds there is in one minute (60) then divide that by the amount of time it took took to pump one gallon of water (the key here is we have to use seconds). The configuration with the lowest time will naturally have the best GPM flow rate. Out of the three different types of configurations, we can tell the Parallel configuration gave us the best GPM out of the other two configurations of running just 1 water block, and while running the water blocks in Series.
This is here for chits and giggles, and for my amusement, all I did was take the GPM and then multiplied those numbers by 60 (how many minutes there is an hour).
These tests were performed with two Seriously overclocked 6970′s, the clock speeds used are 950/1450. These are the results of using a separate loop for the two video cards while having a Swiftech Triple 120mm radiator.
These are the results of my testing with the 6970 water cooled video cards. The ambient temperature was hovering on or around the 16C range (outside temperature of my room, typically ambient temperature is 21C) I like it a bit colder so your results will vary greatly. GPU1 is the primary video card that all three monitors are hooked to, as we can see each of the configurations performed nearly identical to one another. GPU 2 is the slave video card, for the single video card water block configuration, I did not have a slave card, so I could not include it. Now as we can see, while I ran the video cards in a Series configuration, my temperature increased by a couple of degrees. Normally I would not think anything of this slight increase. But When we look at the Parallel configuration, my GPU 2 video card was colder, this is the result of AMD’s power play feature that brings the slave video card to its lowest clock frequencies in full effect. The video card with all of the monitors hooked up will naturally be the hottest video card while we are in an Idle State, as it is still being in an active state. (Idle state meaning just operating within Windows 7)
Lets see how things fare now once I bring the two video cards to a full load state for 30 minutes using MSI’s Kombustor. MSI’s Kombustor is just as brutal to our video cards as Fur Mark, so your actual temperatures while gaming will be significantly lower. The configuration i am more concerned with is the Series setup, as we can see after running MSI’s Kombustor for 30 minutes. GPU 2 on the Series configuration is significantly higher than it is on GPU1, the reason for this is because the water is getting heated from the first GPU (GPU 1), then that heat gets transferred directly to the next video card (or GPU 2), warming it faster the more we work the two video cards. The temperature from the Parallel configuration gave me a lot better controlled temperature for each video card and kept the temperatures close to one another. Because now I am splitting the water evenly across both video cards, I am not transferring any heat from one video card to the other video card (GPU1 to GPU2).
We can have low temperatures while keeping our multiple video cards at their lowest possible temperature without the need of adding more radiators or loops to our computers. When it comes to water cooling, the biggest thing we all need to be aware of at all times (and cost), is our flow rates, if it drops too low, it can spell certain disaster for our gaming systems. Having too low of a flow rate, your components will produce more heat than what your cooling system can keep up with, even though you may have more than “enough radiator” to handle it. So by increasing the overall flow rate of our water cooling system, our radiator will be able to dissipate the heat each component produces in a more efficient way.