Introduction
 
This review has been a long time in the making, thanks to the large number of tests performed as we will get to soon. NZXT was kind enough to send a Kraken G10 and Kraken x41 (review here) for testing and by the time you are done reading this, I hope you agree that the time was well spent. The Kraken G10 was announced in November of 2013 and builds upon similar products that existed earlier (Dwood bracket, Red/Green mod etc) which provide a means to install an AIO CPU cooler on a GPU core for cooling it. There have also been (and continue to be) GPU SKUs that come with an AIO pre-installed too (AMD Radeon R9 295×2, HIS Hybrid IceQ series, Inno3D iChill series, EVGA Hydrocopper AIO to name a few) and also similar solutions with come with an AIO and heatsinks such as the Arctic Accelero Hybrid. Let’s not forget also what is possibly the one thing that will draw most comparisons to this- the Corsair Hydro HG10. This last product along with the Kraken G10 are GPU brackets which provide an easy method to re-use an old CPU AIO cooler on the GPU core while also attempting to provide cooling for the other major components that need it- the VRAM and VRMs in particular. Thanks to Mitchell from NZXT, we can now see how the Kraken G10 performs with the provided x41 cooler. I also want to thank James and George from Corsair for providing the HG10 as well as a Hydro H75 cooler, and also Linustechtips forum member “Faceman” for providing some heatsinks as well as some good information to build upon.
 
Let’s begin by taking a look at the specs courtesy the product page:
 

 
Now keep in mind that NZXT has attempted to make this as universal a solution as possible and the list of GPUs in here is not necessarily complete- the odds of it working with products not mentioned in here or yet unreleased/unannounced is pretty good. It is up to you to make sure it is compatible with your GPU before going ahead with this, and also note that some GPU manufacturers will do unannounced changes to PCB layout and components that can affect compatibility. There is a compatibility checker on the product page but again the same things apply to it also. Oh, and that last note about GPU warranty being potentially void when removing the stock cooler is completely true- some companies such as EVGA explicitly mention warranty will not be void as long as the GPU is not damaged by the user but others are not as willing.
 
Aside from this, we see that the cooler comes in multiple color options and all have a 92mm in common in the contents list. The fan has a sleeve bearing though which I am not a big fan of but we will come to that soon. The 2 year warranty ties in to the fan.
 
Lastly, the AIO compatibility list- if you are familiar with the world of closed loop AIO CPU coolers, you know that Asetek is the biggest OEM there is, followed by CoolIT Systems. A few others like Cooler Master and Swiftech design their own AIOs as well. The G10 is only compatible with Asetek products that follow the current installation system of the “teeth” on the cooler bracket mounted on a round cold plate. If you did not get that, please check out my review of the Kraken x41 and the Hydro H75 coolers. Now let’s go ahead and take a look at the product.
 
Unboxing and overview
 

 

 
The product box measures in at 8.5″ x 5.5″ x 2.5″ and comes covered in plastic wrap. The front of the box immediately tells you what you have inside, including the exact color of the G10 product you chose. For instance, you can see that I got the G10 in white color. The AIO pictured in there is not included of course but perhaps a small note saying that should have been included just in case someone thinks otherwise? You also see how the AIO tube management has been thought of in the product (this will depend on your particular GPU PCB).
 
On the back we see a look at how NZXT has color coordinated cases to go along with this should you so desire as well as their in-house test results comparing this with the stock Nvidia GTX 780 NVTTM cooler. While that’s all fine and dandy (and I can believe those results too), claiming a 40% performance increase on the box can be potentially misleading. If you look at the product page on the NZXT website, you see this instead:
 

 
Here it is clearly mentioned that the 40% performance increase in the cooling aspect of the cooling solution used. While Nvidia GPU Boost 2.0 technology does allow for GPU core frequencies to dynamically increase with lower core temperatures, a 40% increase here is bound by silicon lottery and overclocking. Again a minor point, but one that need not be brought up in the first place if there was no ambiguity.
 

 
On the side of the box we see the specs already listed out on the previous page.
 

 
Opening the box, we are greeted to some nice soft foam protecting the contents inside. Props to NZXT for continuing to set a good standard here as they did with the Kraken x41 cooler.
 

 

 
Removing the foam sheet on top reveals a note of caution in multiple languages warning end users that this is for experienced users and also that GPU warranty can be void by using it.
 

 
Below this is the first set of contents that come in handy for installation and usage:
 

 
You get a pouch containing installation gear, the backplate for the cooler, the installation manual (which can be found online here along with some more interesting information!), the warranty guide and some zipties, and the included 92mm fan.
 

 

 

 

 
We will get to these soon enough in the installation section. Let’s take a look at the fan now:
 

 

 

 
The fan is pretty much a 92mm x 92mm x 25mm one for all intents and purposes. It has a very universally fitting color scheme of white blades and black frame and will go well with just about any build where this will even be seen at all. In fact, as we will see soon, the fan frame will not even be visible once you have it installed. The blades (7 of them) appear to be static pressure optimized by the looks of it, and we will see soon how they fare at moving air without any real impedance to air flow.
 

 
The fan is rated for a maximum speed of 1500 RPM (+/- 10%) which is a change from the usual 80/92mm fans that tend to be high speed versions. But this fan is not pushing air through a radiator or a heatsink and perhaps 1500 RPM is enough with this design? RPM isn’t everything as I have come to realize with my testing of all the fans so far.
 
Another rating we see here is the max current draw of 0.15 A which corresponds to a power draw of 1.8 W on the 12 V rail. This differs slightly from the 0.18 A (2.16 W) rating on the specs listed on the product page. I spoke to NZXT support about this and was told that this could have been an unannounced change from the manufacturing factory (unknown to them too) and they would look into it. In practice, the max operating current I observed was a mere 0.081 A (0.97 W) so I am inclined to think the higher rating here accounts for start up boost also. Either way, you should be good to go with the fan on a standard 1 A header. Keep in mind that most CPU cooler fan headers are PWM (constant 12 V) whereas this is a 3 pin voltage controlled fan:
 

 
So if you end up using this with the CPU cooler’s fan headers (if applicable), then the fan will run constantly at full speed. If you wish to control the fan speed then you need to have this on a dedicated DC fan header. You can’t use the GPU fan header on the PCB either as (a) it is a PWM header and (b) it is not the standard size either! A Gelid PWM fan adapter can come in handy here for cable management but realize you will still be restricted to full speed. But, as we will see soon, that may not be a bad thing at all.
 
The major issue I have with the fan is the sleeve bearing used. It is no real secret that sleeve bearing fans wear out quicker if installed horizontally as opposed to vertically. This post on SPCR does a good job visually of showing why, and I have personally had sleeve bearing fans stop working completely within an year of being mounted horizontally before I knew what I was doing wrong. Having said that, not all sleeve bearings are made equally and this may be a long life bearing used here for all I know. But given how most people have a regular ATX layout case where this fan will be mounted horizontally, you may want to keep checking your VRM temperatures from time to time.
 
The fan cable is not sleeved but does use all black, ribbon style wires. The cable comes in at 16″ long and should be long enough as-is depending on your chosen mode of control/power.
 

 
Now we finally come to the bracket itself! It is also very snugly protected inside as seen.
 

 

 

 
Here you see the bracket in all its glory. The NZXT branding is on the side which usually faces the user in a regular ATX case layout, and you can see where the fan and AIO go. You can also see the multiple tie down points where one can use the provided zip ties for cable/tube management too. The fan section is smaller than the fan diameter itself, and is precise enough where the fan frame will not be visible but it does not block the fan blades at all. You also see fan mounting holes for a 92mm fan or an 80mm fan if you wish to use one.
 
Now let’s see how the installation process is.
 
Installation
 

 
For the installation demonstration as well as performance testing, I used a reference AMD Radeon R9 290 GPU. This particular one is from Gigabyte and uses the reference cooler and PCB to ensure 100% compatibility with the G10. Or is it? Several users found out that there is enough of a variation in the height and thickness of the GPU die (Especially the R9 290) to not allow for the default provided installation gear to work as intended. If you notice that there are any issues and you have followed the installation correctly, then please do NOT force the bracket/AIO on the PCB and instead contact NZXT for a revised screw/post set or get a shim like this one from EK. My card was well within the tolerances of a normal installation so I did not have to go through this process myself.
 

 

 
To remove the stock cooler, one needs to know what screws to remove. This of course depends on your particular GPU but a good source of information are waterblock manufacturers who list out instructions for a lot of GPUs. Given that EKWB tends to cover the most GPUs out there, I would suggest going to their website and checking out the instructions associated for the waterblock of your particular GPU (if applicable). For my card, I already knew what to do but the instructions listed out nicely (as seen here) did confirm what I was doing.
 

 
For this card, there are 18 screws in total including 2 on the rear I/O section and 4 on the GPU core bracket. Once done, slowly apply pressure sideways on the cooler till you can take it off:
 

 

 
If anything, do it just to replace the stock TIM which I have noticed is usually applied poorly and of a inferior quality compared to retail thermal pastes such as Arctic Cooling MX4, Gelid GC Extreme, IC Diamond, Prolimatech PK3 etc. Just switching to better TIM and applying it well can result in a big reduction in GPU core temperatures. Before removing the stock cooler, don’t forget to disconnect the fan cable from the GPU fan header.
 
Now use the provided 4 self tapping screws to mount the 92mm fan on the bracket such that the direction of airflow is inwards into the PCB:
 

 

 
Here is also where you would the 2 foam blocks as shown here by Legit Reviews. But in my case, the foam blocks provided must have been shorter than usual as they were not contacting the PCB completely. So I left them off, and I did not notice anything resulting from their absence as I did try out some pieces of foam I cut up myself just to be sure. I called up NZXT support and was told they don’t recommend using the foam pieces themselves. Why include them in then?
 
Next up, use the provided 4 posts, bolts and backplate as per instructions pertaining to your particular card. For instance, the AMD R9 290/290X cards meant having to use the “A” set of holes in the backplate:
 

 

 

 

 
Then slide it over the 4 holes around the GPU core on the PCB:
 

 
Now’s the tricky part (especially if you only have one set of hands for all this). Refer to my review of the x41 (for example) linked previously to see how to remove the stock Intel CPU bracket off the AIO cold plate. You then end up with this:
 

 
If you are using a brand new cooler, then now is the time to wipe off the pre-installed TIM. They were designed for a CPU IHS (and even so do an average job at best given how different CPU IHS have different sizes) and definitely not a relatively smaller GPU die. You don’t want too much die on the core, and you don’t want it outside the core either if possible. Best to apply TIM yourself here!
 

 

 
Navigate the coldplate through the “teeth” on the bracket keeping in mind how you are going to route the tubing and any other wiring on the cooler. Then turn it so the teeth on the cooler align in that on the bracket and pull the cooler such that they mate together. You need to have the pressure applied here till the bracket is screwed in completely and this is why having an extra set of hands can be handy:
 

 
Next up, while keeping the cooler in place on the bracket, orient it on the 4 posts such that they go through one of the 3 sets of holes in the bracket. I will refer again the Legit Reviews article as their sample had too thick a coat of paint to allow for the screws to go in completely thus necessitating some work on their end to figure out that the holes were a 3/32″ in size. They then ended up using a 3/32″ drill bit through the holes to remove the excess paint and make sure the screws went in just fine. I did not face any such issue myself but if you do- now you know what to do! Don’t worry if you notice that the posts do not allow for a lot of the thread to extend upwards from the bracket- it is designed that way.
 

 
Once this is done, use the 4 nuts provided to tighten the bracket and cold plate on to the card:
 

 

 
NZXT suggests tightening down in an X formation. Use your best judgement here but I always go with pairs of 2 diagonally opposite screws/posts and alternate between the two till there is no more thread to tighten down. With GPUs having bare dies, you really have to have a precise mounting mechanism and that’s why any variations in either the mounting or the GPU die itself is bad news. NZXT seems to have the mounting mechanism spot on, and have provisions for if your particular GPU is out of specs for whatever reason.
 

 
Now depending on your cooler, you can use zip ties to better manage how the wires and tubes come out of the card and towards the part of the case where the radiator will be installed. In a perfect world, you would now install the GPU in the motherboard’s PCI-E slot and the AIO’s radiator in the case. Plug in the AIO pump power, the fan wire and any other cables as necessary and that would be it. But this is not a perfect world…
 
Potential installation tweaks
 
Let’s go back where we left off:
 

 
Notice something? Maybe this will help:
 

 
The AMD R9 290/290x reference PCB uses two set of VRMs, and one (VRM 1) is on the left side of the GPU core. The thing about trying to have compatibility with as many GPUs as possible is that you can’t account for things like this. This is akin to a universal GPU waterblock which only cools the core and then depends on other things like heatsinks and/or fans to cool the memory and VRMs accordingly- especially if you do not have good airflow in your case.
 

 

 
As you can see the AIO coldplate makes good contact with the GPU core and the fan on the NZXT G10 helps cool the VRMs on the right side of the core (VRM 2) but it could be better. It also does not really cool the memory chips either. Time now for some heatsinks!
 

 
I am definitely not advocating this particular set of heatsinks over any other but these were provided free of charge and come in at a good price ($6.77 in the USA as of the date of this article).
 
These are mostly copper and come with a sticker on the underside which peels away to reveal some thermal adhesive. Fitting these on is a piece of cake really:
 

 
For the VRMs, I ended up using heatsinks and thermal pads from a cooling enhancement kit Gelid made specifically for this card:
 

 

 

 
Here’s how the card looks with all these added in:
 

 

 

 
The Gelid heatsink for VRM 2 just about clears the fan and ends up working great in my opinion. You can also go with a more universal VRM heatsink kit if you want to transfer everything from GPU to GPU.
 
Despite all this, there’s still no airflow over VRM 1 or the memory chips. Now memory chips will be fine with the heatsinks alone and some decent case airflow, but for VRM 1 I went ahead got this:
 

 
The Antec Spotcool is available for $11.99 in the USA as of the date of this article. This nifty little thing can be installed into any spare motherboard standoff in your case (or any 6-32 threaded hole for that matter).
 

 

 
This particular version uses an 80mm fan on a 3 pin cable (14 cm long, unsleeved) and can be potentially daisy-chained to the NZXT 92mm fan too (with both left at 12 V). There is a switch to choose between 3 fan speeds (specs list it as 2000, 2500 and 3000 RPM as seen here). The fan can be rotated around its axis so as to direct the airflow as necessary:
 

 

 
The fan support can also be bent around as well:
 

 
So you could potentially have this installed such that it is blowing air directly at VRM 1 (with or without heatsinks on). One thing that may deter you is that this has bright LEDs that can’t be turned off:
 

 

 
That’s as seen from the back and the front respectively with the light levels of course affecting the camera shutter speed (both were taken at f/4.0).
If this isn’t to your liking then there are always PCI slot brackets that allow you to install fans on it and you can then have this assembly below the G10/GPU assembly. This is contingent on you having free slots in your case and also if you are going with multiple GPUs spaced to allow for these as well.
 
I hope this section does not deter you from buying the NZXT G10. Many GPUs will be fine with the product as is, and all of these add-ons can be configured so that they can be re-used also. Some GPUs also come with cooling base plates that may be re-used as well. So it is up to the end user to decide how much of an initial cost to justify here. To help make that decision easier, let’s see how all of these play a role in cooling the card.
 
Performance testing

By which, I mean the cooling aspect here so there’s no confusion :)
 
Testing methodology
 
The provided 92mm fan was controlled using a dedicated fan controller (Aquacomputer Aquaero 6 XT) in power control (voltage regulation) mode. The controller also enabled RPM readout. Linear airflow was measured using an Extech 45158 Thermo-Anemometer 6″ away from the fan such that it measured the airflow in feet per minute. Fan noise was measured in an anechoic chamber of size 5′ x 8′ with ambient noise level ~19 dBA and a sound probe held 6″ away to measure the sound volume in dBA accordingly. This was also done for the Antec Spotcool at the three speed levels via the switch.
 
The GPU used for testing was the reference AMD R9 290 GPU. It was installed (as in, with NZXT G10 alone, and in different combinations of heatsinks and fans alike) on an Asus Rampage V Extreme motherboard in a PCI-E 3.0 x16 slot, with an Intel i7 5960x at 4.5 GHz core frequency to ensure that the CPU was not bottle-necking the GPU in any way. The GPU was powered using separate power cables for the 8 and 6-pin PCI-E connectors from an EVGA Supernova 1300G2, and the entire assembly was placed in a climate controlled hotbox at 25 ºC ambient. The results will still be reported in terms of a delta T (component – ambient) in ºC for ease of comparison. The ambient temperature should be nearly the same throughout the box with a +/- 0.25 ºC variability at most but still the ambient temperature sensor was held in front of the intake 92mm fan. Each test/scenario was repeated 5 times and the average value taken to account of possible issues with mounting and TIM. Speaking of TIM, Gelid GC Extreme was used on the GPU die. Unigine Heaven 4.0 in the extreme preset was used to apply load and GPU-Z (versiom 0.8.1) was used to monitor core, VRM 1 and VRM 2 temperatures as well as core frequencies. Measurements were taken at steady state (whenever applicable) to ensure that the GPU itself was as less of a variable as possible. With the stock cooler and the default shipping BIOS, the automatic fan curve meant that thermal throttling occurred and so an increased fan speed was chosen at the risk of higher noise levels. Since this is not a GPU review, I will not be faulting the stock cooler anymore than when it comes up as a comparison to the G10 w/AIOs used (Kraken x41 and Hydro H75).
 
Results
 
First up, the stock 92mm fan:
 

 
Well I kept hearing about how this fan was quiet even at full speed, and now I have the numbers to confirm that. In fact, I would suggest leaving it at full speed itself- especially if you are not using heatsinks. The RPM response curve is excellent as well, with the fan going from 1359 RPM (JUST inside the 1500 +/- 10% though, even after an hour of on time) at 12 V (100% power) to 252 RPM at 3.25 V (27% power) before turning off. As far as airflow goes, this isn’t going to win any rewards but that’s what you get with a 92mm quiet fan. I don’t have very many other 92mm or 80mm fans but if you take a look here, you will see another 92mm fan which goes all the way to 2800 RPM, more static pressure optimized and hits ~46 dBA when measured the same way. Is replacing the stock fan worth it? Not for silence. But if you are not going to use any heatsinks, then a higher speed fan may very well come in handy in cooling VRM 2 better.
Let’s get to the thermal testing now. The GPU core was set at 947 MHz (stock) and at the default Vcore under load of 1.18 V. I made sure there was no throttling from power limits, and then set about with the load testing as indicated in the previous paragraph. To compare the NZXT G10’s performance, I went ahead and added in the following:
 
1) Reference cooler with the blower fan speed manually controlled
 
2) Corsair HG10 which reuses the blower fan (again manually controlled)
 
3) Custom loop 1 (EK VGA Supremacy universal GPU waterblock, XSPC D5 PWM w/XSPC D5 dual bay reservoir, Hardware Labs Black Ice Nemesis 120GTX, EK Vardar F4-120 fans in push-pull) with the pump and fans manually controlled
 
4) Custom loop 2 (Swiftech R9-LE full cover GPU waterblock for the AMD reference R9 290 GPU, XSPC D5 PWM w/XSPC D5 dual bay reservoir, Hardware Labs Black Ice Nemesis 120GTX, EK Vardar F4-120 fans in push-pull) with the pump and fans manually controlled
 
The heatsinks described earlier were added on to the G10 and Custom loop 1 as applicable for sub-tests, with an EK Vardar F3-120 fan blowing over them in the latter case.
 
Let’s now see how these fared in the various permutations and combinations tested out:
 

 
Here we see a consistent trend. The NZXT G10 w/AIO is consistently better at cooling than the reference cooler, and by a large margin too. This is not a review of the GPU and its reference cooler of course, but in this case the G10/AIO assembly was also a lot quieter too. There isn’t much effect of the AIO used either- granted both were single fan radiator AIOs, and the only reason the Hydro H75 was cooler was because it comes with 2 fans as opposed to the larger Kraken x41 that comes with a single fan. But given my testing of the Corsair HG10 (review coming soon) with a large 280mm AIO and noticing very little difference in core temperature compared to these, I definitely recommend just going with the least expensive AIO and calling it a day. You don’t even need to have the AIO pump and fans at full speed either!
 
The Corsair HG10 provided a better mount for the AIOs with my particular GPU (keep that in mind) and thus came out ahead of the G10 here. Both custom loops also fared better as expected but they are using parts that are a lot more expensive also!
 
What about the VRMs? Keeping in mind that the reference R9 290 has two sets of VRMs (VRM 1 to the right of the core, and VRM 2 on the left), let’s begin with VRM 1:
 

 
I think it goes without saying that you should get some heatsinks with the G10. That little 92mm fan running at ~1350 RPM over naked VRMs won’t do much good. In fact, the reference cooler actually outperformed the G10 here albeit at higher noise levels. The Corsair HG10 with the customized heatsink/bracket worked out very well but once you add in heatsinks yourself, the NZXT G10 comes right back at it. As far as the custom loops go, don’t fret a lot about the performance of the Swiftech full cover block. If you read Stren’s roundup of the full cover blocks for this very GPU, you would see that this block is a very poor performer when it comes to VRM cooling relative to some of the other blocks. You can also refer that to see how other blocks performed at cooling the GPU core also.
 
Finally, VRM 2:
 

 
The story does not change much here. VRM 2 does not get as hot as VRM 1 but in my hotbox with the NZXT G10 having no direct airflow over it, it definitely showed. Simply adding in the heatsink helped a lot, especially since my hotbox has active airflow when needed to keep the constant ambient temperature and also partly from that fan on the other side. Adding in the Antec SpotCool also helps a lot as you can see. The Corsair HG10 again comes out on top here as-is, and the two custom loops perform about the same as before.
 
Notice how I did not mention the VRAM? Well I really did not notice appreciable effects of all these factors (fan, heatsink etc) on VRAM temperatures. To be honest, you would probably be just fine with no heatsinks of them at all and you can save on the money there to get heatsinks for the VRMs. Also, if you are wondering why I did not test for any effect of GPU overclocking then (a) remember that each test was done 5 times which adds up very quickly, (b) overclocking is totally GPU dependent and (c) I was pretty close to thermal throttling in one case and not very happy with the VRM 1 temperature being so high in another.
 
Conclusion
I have seen a lot of “G10 vs HG10″ topics and post in online communities, and it reminds me of “universal GPU block vs full cover block” topics. The way I see it, the NZXT G10 and the Corsair HG10 are not competitors, but rather complement each other by providing options for end users to get into liquid cooling a GPU easily and relatively less priced than custom loops. This is the big thing here- liquid cooling your CPU sounds great but does not provide much tangible benefit compared to good air coolers. But to be able to re-use an old AIO to keep the GPU cooler, help with overclocking while being potentially quieter than the stock cooler? I’ll take that any day!
 
The NZXT G10 has an MSRP of $29.99 in the USA for all the different color options but can often be found at lower prices directly or via rebates. I definitely recommend grabbing one of these if you are not inclined to go with a universal GPU block and setting up a loop each time. Spend some money on an older AIO such as the Corsair H55, a set of VRM heatsinks and perhaps also something like that Antec SpotCool and you really have no excuse left to not watercool a GPU anymore. If you are content with the performance of this assembly and don’t mind all the extra parts to be added in, then good for you. If you are looking to use this till you get into custom watercooling, then potentially also consider the Corsair HG10- especially if you have a blower fan that can be used as it does not come with one as of the date of this article.
 
Performance aside, installation isn’t the easiest thing in the world here and there were a few QC issues noticed by me (and others too). But the pros definitely outweigh these cons here. I hope this review has helped you out, and please do let me know if you have any feedback/suggestions.