Unlike CPUs - which are designed to be cooled by a wide range of third party coolers - video cards are much more limited in terms of cooling options. Different manufactures make models with either more powerful or quieter coolers, but there are limited numbers of third party coolers available that you can install yourself. Of those that are available, they are typically either very expensive or not much better than the better stock coolers made by the various video card manufacturers.
The Kraken G10 from NZXT tackles GPU cooling in a way that we at Puget Systems have always wanted to see, but have not had the time to make happen ourselves. Instead of providing a whole cooling solution, the Kraken G10 is simply a metal bracket and fan that allows you to mount various different closed-loop coolers. These coolers are traditionally used for CPUs and allow you to custom tailor your video card cooling much like you can for your CPU.
We are not the first to review this cooling bracket, but so far no review has performed any testing with the video card installed in an actual chassis. Every benchmark we have seen has been done with a test bench which is completely open air with no sides or top. This is perfectly fine for comparing different video cards with stock coolers, but in our opinion is completely inadequate for comparing the stock cooler on a video card to the Kraken G10 that has been paired with a closed-loop cooler. Especially from a noise standpoint, you will get vastly different noise measurements with the system on a test bench than you will with the system installed in a chassis. In a chassis, the case panels block a good amount of the noise that traditionally comes from a video card's cooling fan. But with the Kraken G10, the fan is mounted to the front, top, or rear of the chassis rather than being deep inside the chassis which means the noise from the fan will barely be reduced at all.
Since none of the reviews available performed their testing in a real-world environment, we decided to do our own testing using the Fractal Design Arc Midi R2 chassis.
To examine the performance of the Kraken G10 bracket, we used the following hardware:
While most of our hardware is what we use in our standard test setup, we had to make a few changes to accommodate the Kraken G10. First, we used the Fractal Arc Midi R2 chassis as it is one of the few in our line that can easily mount a 140mm radiator in the front of the chassis. Since mounting it in the front means we can use it as an intake and still use the rear fan mount for the CPU cooler, this is on of the best chassis currently in our line for this type of setup.
The one change we made to this chassis was to add a side fan by simply swapping the side panel with one from a Fractal Design Define R4 which includes a 140mm fan mount. We have found that adding a side panel fan for high-end video cards like the GeForce GTX Titan and Radeon R9 290X results in much better temperatures, but also makes the system quieter overall as well. This will also help with the Kraken G10 as it will provide fresh air to the 80mm fan that cools the VRM and other individual chips on the card.
The other change from our normal test setup is not using the Corsair H60 or H90 liquid coolers that we typically use, but instead using the NZXT Kraken X40 cooler. The performance should be very similar to the Corsair H90, but the tubing on the Kraken X40 is 4 inches longer than most other closed loop coolers. We found this extra tubing to pretty much be a requirement for mounting the radiator anywhere other than the rear fan mount on the chassis. Also, the Kraken X40 includes a thermal sensor that monitors the temperature of the coolant and automatically ramps the fan according to a fan profile which can be adjusted with the NZXT control software. This is essential for this setup since plugging the fan into the motherboard would make it ramp according to the CPU or motherboard temperature rather than the video card. We will be using two different fans to cool the radiator: the NZXT fan that comes with the Kraken X40 and the quieter, but lower flow, Corsair fan that comes with the Corsair H90.
All of our testing is done in a controlled office environment that stays at a consistent temperature of 22 °C. Acoustically, the ambient noise level stays at roughly 33 dBa, with a variation of about .5 dBa. To compare the noise level and cooling of the Kraken G10/X40 combination, we will take temperature and noise readings in three different situations: idle, gaming (using Unigine Heaven Pro), and extreme load (using Furmark).
Idle Temperature & Noise Levels
Temperature and noise readings at idle are not a great metric since most modern hardware is really good at ramping down when not needed, but it is still somewhat useful information if your system is going to spend much of it's time in a low-load state (like browsing the web)
At this load level, we saw a 10 °C drop in temperatures on the R9 290X and a drop of 8 °C for the GTX Titan with the Kraken G10/X40 combination. This is a great sign of what the Kraken G10/X40 is capable of, but the big question is if the lower temperature results in a noise reduction.
Interestingly, the Radeon R9 290X and GeForce GTX Titan are almost identical in terms of noise in our office environment. We know that this isn't actually the case since reviewers with access to much more sensitive equipment and quieter locations have shown that the GTX Titan is quieter than the R9 290X at idle, but in our system with our setup, the difference is so small that we can't reliably measure it.
Another interesting point is that even with the quieter Corsair fan, the Kraken G10/X40 cooling combination is louder than the stock cooler on both video cards. This comes down to what we brought up in the introduction: since the cooling fan for the Kraken is located at the front of the chassis instead of being inside the chassis like the stock cooler, the noise is not being blocked nearly as much by the chassis itself.
Unigine Heaven Pro 4.0 Temperature & Noise Levels
In gaming situations, we see a huge thermal advantage using the Kraken G10 with a closed-loop cooler. Going from 80 °C to 41-44 °C on the GTX Titan and 94 °C to 51-55 °C on the R9 290X is very impressive for the relatively small cost of the cooling system. These temperatures are not quite as low as what you would see with a more traditional full liquid cooling setup, but they are within a couple of degrees.
The one concern we still have with the Kraken G10 is how hot the other chips on the card get now that they are not being cooled by the stock heatsink. The Kraken G10 comes with a 80mm fan to help keep those components cool, but we frankly are not sure if the fan alone will be enough without the aid of a heatsink. Unfortunately, while AMD video cards have a couple temperature sensors on the VRM chips that can report temperatures, NVIDIA cards do not. So to get an idea of whether the chips are being adequately cooled or not we decided to use our thermal camera to get a look at the temperatures of the various chips on both video cards.
Since the bottom of the cards are covered by the bracket and fan, we instead took thermal images of the top of the cards. Even if the chips we really want to see are on the bottom of the card, we should still be able to observe any hot spots as the heat should "leak" through the PCB to the components on the opposite side.
As you can see in the thermal images above, the VRM and other chips on the video card are being more than adequately cooled with the Kraken G10 on both the R9 290X and GTX Titan. In fact, the Kraken G10/X40 combination does such an excellent job keeping the main GPU core cool that it actually keeps much of the surrounding card cooler than it is with the stock coolers.
Due to the fact that the reference cooler for the GTX Titan is so quiet, the Kraken G10/X40 combination is louder than the stock GTX Titan cooler even with the quieter Corsair fan. So if you are gaming with the GTX Titan you can use the Kraken G10 to get great temperatures, but at the cost of increased noise levels.
Since the R9 290X's stock cooler is one that would never be called quiet, it is no surprise that the Kraken G10/X40 is quieter even with the louder NZXT fan. If you go with a quieter fan like the Corsair one we used, you can reduce the noise levels even further.
Furmark Temperature & Noise Levels
Furmark is somewhat of an unrealistic test for most users since it puts a much higher load on the video card than almost any game. However, it is a good way to put a "worst-case scenario" load level on the card so it is very useful. Amazingly, the Kraken G10/X40 kept the GTX Titan 36-39 °C cooler than the GTX Titan's stock cooler. Similarly, the Kraken G10/X40 kept the R9 290X 35-39 °C cooler than the R9 290X stock cooler.
We know that Furmark puts a strain on the VRM, however, so the main test is to see how well the VRMs are being cooled with just the 80mm fan that is on the Kraken G10 bracket.
While the GPU core is certainly being cooled well in Furmark, the thermal images give us a great look at the problems associated with removing the stock heatsink. In short,there is a hot spot right on the VRMs for both the R9 290X and GTX Titan when using the Kraken G10. Our thermal image doesn't quite show the exact temperature since we are looking at the back of the card rather than the front, but we are still able to see that the VRMs are running at dangerous levels. In fact, GPU-Z is able to report thermal readings for the VRM on the R9 290X and reported temperature as high as 112 °C in Furmark! For reference, the same sensors on the R9 290X reported temperatures of only 84 °C when using the stock cooler.
112 °C is an incredibly hot temperature for any computer component to run at, and frankly we are impressed that the card was able to survive these temperatures for even a short period of time. The GTX Titan is not nearly as bad, but based on the thermal images we estimate the VRMs on that card to be running around 90 °C which again is much higher than with the stock cooler.
In Furmark, we see slightly different results than we saw when running Unigine Heaven Pro. Starting with the GTX Titan, we again see that the Kraken G10/X40 is louder than the stock cooler, but it is by a slightly larger amount this time around. For the R9 290X, The Kraken G10/X40 is actaully louder this time with the NZXT fan, but is still a bit quieter than the stock cooler when using the Corsair fan.
After using the Kraken G10, we've come to the conclusion that it isn't quite as perfect as most reviews are currently making it out to be. Not only will you run into problems with the VRMs overheating when the video card is put under abnormally heavy loads, but the noise level of the cooler makes it not very attractive for some video cards.
If you are using a NVIDIA GeForce GTX Titan we honestly wouldn't recommend the Kraken G10 unless you are having problems adequately cooling your GPU without it. The Kraken G10 will make the GPU run cooler, but at the same time is louder than the stock cooler, and your VRMs will run hotter when under heavy loads.
The AMD Radeon R9 290X, on the other hand, is a completely different matter. Since the stock cooler on the R9 290X is so loud, using the Kraken G10 will both lower your GPU temperature by a significant amount while also lowering the overall noise level of your system. The big downside is the VRM cooling. The 80mm fan on the Kraken G10 is able to keep the VRM and other chips on the card cool enough while gaming, but simply isn't up to the task when the video card is under very heavy loads.
No matter what card you are using, if you are doing anything that puts an extreme load on the video card we do not recommend the Kraken G10. If you are running normal loads like gaming you should be fine, but need to be aware of the risk you are accepting by running your VRMs hotter. You can of course apply additional heatsinks to the VRM which is what NZXT recommends if you are concerned about the VRM cooling, but you would need to use thermal expoxy to ensure they are secure which means you likely cannot switch to a different cooler (or even back to the stock cooler) in the future.