Comments Locked

61 Comments

Back to Article

  • lavaheadache - Tuesday, March 11, 2014 - link

    I thought NZXT bought the design(idea) from d-wood? Funny how they get all the credit
  • sicyo - Tuesday, March 11, 2014 - link

    I immediately thought of d-wood's bracket too, as I have been using one for about a year now.

    I wouldn't look down on NZXT, they're making it more publicly available. d-wood couldn't keep up with demand and as far as I've seen, has been MIA. His last post on overclock was 9 months ago.
  • sicyo - Tuesday, March 11, 2014 - link

    Hah, wish I could edit that. NZXT is already out of stock. Hopefully they'll be able to keep up. There are a lot of gamers out there that want to liquid cool their GPU without spending a lot on a custom loop.
  • SodaAnt - Tuesday, March 11, 2014 - link

    I've had d-wood's bracket for almost two years now, and it works pretty well, but its nowhere near as pretty as this.
  • malnute - Thursday, March 13, 2014 - link

    Not sure what is so hard to understand I bought you and your designs hence you are now invisible.
  • jtrdfw - Monday, June 2, 2014 - link

    Well, the thing about buying the rights to a product is...
  • taserbro - Tuesday, March 11, 2014 - link

    Is there enough headroom under the 92mm fan to install some copper heatsinks for the VRMs and memory modules? I've seen a few reviews of this kit but none of them seem to address this.

    Some cards like the EVGA 780ti have been leaving the VRMs naked under the cooler and there have been documented cases where prolonged full load use resulted in hardware failure due to the lack of VRM cooling. While the prospect of having so much overclocking headroom on the GPU temperature front is quite attractive, I wouldn't want to instead fry the card's power delivery circuitry.
  • Yungbenny911 - Tuesday, March 11, 2014 - link

    Yes, there is more than enough room under the fan to install copper, or aluminum heatsinks for the VRMs and memory modules.

    Recent GPU's like the MSI Gaming 770 and 280x already have a passive heatsink cooler on the VRM's and memory modules. I left that on, and placed aluminum heatsinks on the passive heatsink. I also have dual side-panel fans blowing directly on them to prevent overheating. Works just fine for me, and i'm able to OC my memory to 2004Mhz on my gaming 770 without heat problems. http://i.imgur.com/viYQ6eL.jpg
  • kyuu - Tuesday, March 11, 2014 - link

    Could you please tell what the clearance is and, therefore, what kind of heatsinks could be installed for the VRM and memory? I'm quite interested in this, but I'm not too enthusiastic about leaving the VRM and memory naked and depending on just the fan to cool them.
  • Yungbenny911 - Tuesday, March 11, 2014 - link

    This is what my Gaming 770 looked like when i took of the Twin Frozr IV cooler.
    http://www.techpowerup.com/reviews/MSI/GTX_770_TF_...

    The black plate you see is the passive heatsink cooler i was talking about that already covers the V-ram modules and VC. I didn't take that off while installing my bracket and H55's, i just attached this type of Aluminum heatsinks to it (8 for each 770) http://www.newegg.com/Product/Product.aspx?Item=9S...

    There is still enough space between the bracket and aluminum heatsinks to allow air pass through.
  • kyuu - Wednesday, March 12, 2014 - link

    Awesome, thanks for the info.
  • Jioker - Wednesday, July 23, 2014 - link

    Hi mate, I was wondering if you could give me/us a more in depth "installation guide" of the heatsinks on the VRMs (as I got a bit confused)? It would very helpful! :D
  • Whatchagot - Tuesday, March 11, 2014 - link

    I thought the issue was that GPU liquid coolers need to cool VRM too. While they seem to perform well they don't cool the whole unit properly. Please correct me if I'm wrong.
  • E.Fyll - Tuesday, March 11, 2014 - link

    That is partially true. You see, it is not unlikely that the VRM of more powerful cards may require extra cooling, especially if the removal of the stock cooler means that the VRM will remain entirely without cooling. However, with the EVGA GTX 770 that I used for the review, the temperature increase of the VRM section was marginal, at best. It would appear that it depends on several factors, such as the design and the power of the card. I only had one card to test it on so I apologize for the lack of extensive data on the matter.
  • Whatchagot - Tuesday, March 11, 2014 - link

    Thanks for sharing the experience. Many people are interested in how to cool the GPU better and quieter.

    Any noise measurements or perceived improvements in noise?
  • E.Fyll - Wednesday, March 12, 2014 - link

    Well, that would depend on the AIO kit used, the speed of the fans, etc. I suggest that you read our roundup review of AIO kits, that should help you select the most suitable Asetek-based kit according to your needs.
  • Popskalius - Tuesday, March 11, 2014 - link

    why not just have longer hoses?
  • Yungbenny911 - Tuesday, March 11, 2014 - link

    It depends on the type of closed loop cooler you are using. I personally prefer shorter hoses to reduce the time it takes the coolant to go back through the radiator, and back to the GPU.
  • WithoutWeakness - Tuesday, March 11, 2014 - link

    I hate to say but the length of the hoses and their effect on the time it takes coolant to flow from the block to the radiator has next to no impact on the cooling potential of a water loop. The whole loop is full of liquid and the block always has water flowing through it. The length of hoses does not affect the flow rate of a water loop - the pump does. You could use 1 foot hoses or 10 foot hoses and the water would flow at the same rate (gallons/minute or liters/minute) on a given pump.
  • mpdugas - Tuesday, March 11, 2014 - link

    However, the longer hoses hold more fluid; that is a cooling benefit in-and-of itself.

    One thing I noticed about LC; the latent heat in the liquid keeps the CPU/GPU warmer, longer, than a good AC kit.

    The thermal shock is less, however, as the CPU/GPU warms more slowly and cools more slowly, as well.
  • owan - Tuesday, March 11, 2014 - link

    The increase in fluid is only a cooling benefit until the system comes to equilibrium, at which point it becomes meaningless
  • mpdugas - Friday, March 14, 2014 - link

    Of course, the CPU/GPU temperature is rarely at a steady, equilibrium state...
  • Aikouka - Tuesday, March 11, 2014 - link

    Before I built my first custom loop, I did a bunch of research into parts and such. One of the first things I stumbled across was a two facts about water cooling. (1) After a certain point, you don't gain much from increasing the flow rate. (2) It takes a large amount of coolant for the sheer volume to have a tangible effect on temperatures.

    The usual rule of thumb in water cooling is to use a LITTLE hose as possible. In *most* implementations, hoses are not designed to disperse heat unlike other items in the loop. So, unless you're one of those crazy types that uses copper tubing in their water cooling build, there isn't much of a point to excess tubing. Also, excess tubing can be just about as bad as too little tubing in that you can end up with some awkward bends. In my experience with AIOs, they aren't exactly known for using great clamping methods on the tubes. I recall reading complaints about leaks, and the users in questions had very tight bends, which I could see causing gaps between the rubber tubes and the barbs.
  • 1Angelreloaded - Tuesday, March 11, 2014 - link

    OK you are right, but wrong as well, see the pump supplies so much pressure per foot of available volumetric area, when we talk about pumps we typically discuss a statistic called pump head for that reason, you typically don't want 2 much head for a small loop otherwise the Bar pressure levels will be 2 high and most likely you will have leaks for most acrylic res containers, now if the head is 2 low you will have little pressure within the system and a low flow rate which is bad for a CPU cooler block that work better under a higher volume rate, 2 slow and the fluid will boil causing gas within the lines and catastrophe. GPU cooling is entirely different and because the Blocks are of a larger displacement they function better under lower pressure levels. Now how this would apply to an All in One cooler is beyond me but remember the original design is for a CPU not a GPU, a CPU can hover between 30-50 cel while a gpu can climb all the way to 90 cel, so I am not sure how I feel about this, My question would be more can This fit an EVGA 760 ACX SC 4GB and will it work with Corsair's Link software if it will an SLI 350D would be an interesting prospect.
  • tim851 - Tuesday, March 11, 2014 - link

    "One such limitation is that the fans of the liquid-cooling kit cannot be thermally controlled in relation to the temperatures of the graphics card."

    1) There is software, which can control fans and read GPU temps, like SpeedFan (at least for nVidia GPUs, last time I checked three years ago).

    2) There are adapter cables, which let you connect a standard 3-pin fan to the 2-pin mini connector of a graphics card.
  • Yungbenny911 - Tuesday, March 11, 2014 - link

    Or you can just get a fan controller for 20$. I have my 4 corsair fans in push pull config, and they are all controlled by my 600t's fan controller. MSI Afterburner's OSD is used to monitor the temp of the GPU as i am gaming, so i know when to ramp up the fans, or reduce them.
  • E.Fyll - Tuesday, March 11, 2014 - link

    1. The fan of the kit are not connected on the GPU, therefore they cannot be controlled by any such software. If the motherboard supports fan control on the header the fans may be attached to, then the speed of the fans may be controlled, which control will however be unrelated to the temperatures of the card.

    2. Yes, there are. These however are a) not included and b) the high amp draw of a 140 mm fan can easily fry the circuit of a card that was designed to drive fans 1/3 of that size. :)
  • thewhat - Tuesday, March 11, 2014 - link

    "If the motherboard supports fan control on the header the fans may be attached to, then the speed of the fans may be controlled, which control will however be unrelated to the temperatures of the card."

    SpeedFan lets you automate the speed of any fan* based on any temperature in the PC (CPU, GPU, chipset, HDD,...).

    *with the possible exception of the fans on Nvidia cards. AFAIK for now it only lets you read the temps, but not control the fan speed on the video card itself. With AMD cards, you can do both.
  • Aikouka - Tuesday, March 11, 2014 - link

    Aren't some GPU fans around 80-90mm? I believe my Gigabyte Windforce card uses 75mm fans, but that's also three fans compared to just one 140mm.
  • Nirvanaosc - Tuesday, March 11, 2014 - link

    SpeedFan for the win! (if you use Windows)

    I use it and it's even better than letting the card control the fans because you can customize the
  • Nirvanaosc - Tuesday, March 11, 2014 - link

    ... the fan speed / temperature curve. Just perfect.

    I have it controlling 8 fans + 2 pumps (6 GPU and + pump and 2 CPU + pump) in 3 PWM channels.
  • Streetwind - Tuesday, March 11, 2014 - link

    You claim that there is currently no competing product, but what about Arctic Cooling's Accelero Hybrid cooler? True, it's not just a bracket, but rather a complete solution. But still, it does the same thing, cooling the GPU with a closed water loop and the other onboard components with a small dedicated fan. It's been on the market since summer 2012.

    Overall though, I like the flexibility the mounting bracket offers. Users can pick and choose the liquid loop they want or need for their card and/or goals, which is important considering how much the various solutions differ in terms of price, noise and performance. And in general, I think that closed liquid loops are horribly overpriced and overspecced for CPU cooling without showing any real advantage over air solutions, whereas on a GPU their high potential can actually be utilized (as the temperature results in this test show).

    Now if we could just get an Asetek pump unit that was truly quiet in idle mode, even I might start to show genuine interest... ;)
  • SquattingDog - Tuesday, March 11, 2014 - link

    Any chance of testing additional cards, such as the 7970/280X as these require "shims" which do not appear to be included in the kit?! Also other higher power cards such as 780GTX, 290 and 290X would be great to see how it performs with them.
  • Folterknecht - Tuesday, March 11, 2014 - link

    http://www.tomshardware.de/r9-290-umbau-wasserkuhl...
  • BlakKW - Tuesday, March 11, 2014 - link

    In terms of what I might spend money on in the near-future, this is the most interesting mini-review + comments that I've seen on AT in long time... ty
  • Gigaplex - Tuesday, March 11, 2014 - link

    "While saying, "80 °C to 45 °C – oh, that's 44% lower temperatures" is technically correct"

    No, it's not even technically correct. When comparing temperatures like this, you need to either measure the delta from ambient (sensible) or measure from absolute zero (less sensible). 45 °C is not 56% of 80 °C, convert to Fahrenheit and try to make the same claim.
  • JarredWalton - Tuesday, March 11, 2014 - link

    Of course it's "technically correct", meaning that the temperature dropped 45C. But as we go on to explain, that's not really looking the right aspect. It's like arguing that saying "the sun is shining" isn't "technically correct" because if we get into the nitty gritty details, "shining" isn't really the right word. Good grief! There are all sorts of reasons to not simply compare temperatures, which is why E. uses the more useful C/W and explains why it's more useful.
  • JarredWalton - Tuesday, March 11, 2014 - link

    Maybe we should switch to Kelvin though? LOL
  • Rick83 - Wednesday, March 12, 2014 - link

    It's not technically correct, and NZXT should be called out for spouting such ridiculous marketing rubbish.
    It's just as incorrect as 4 inches being twice as long as 3 inches, because 0 inch celsius is equal to 2 inches. You cannot use linear scaling with fixed offset units. Doing so to get impressive marketing numbers is misleading, and should be punished. I for one will avoid NZXT, thanks to this ludicrous display of marketing mistakes, where the engineers weren't asked to proofread wild marketing claims.
  • E.Fyll - Tuesday, March 11, 2014 - link

    You are correct. Unfortunately my original text was phrased in a rather, um, "insensitive" manner, and thus an editor rightfully felt that he had to edit it. It is my fault, as I need to be more careful of how I phrase certain things.

    True enough, it is not technically correct. It just is not correct at all, in any way possible. :)
  • JarredWalton - Tuesday, March 11, 2014 - link

    Well, it's factually correct to say the temperature dropped 35C which is 44% lower, but since temperatures aren't a 0-based scale it messes things up and obscures a whole lot of relevant information. It's scientifically the incorrect way of discussing things, which is what I was trying to say with the "technically correct" bit. Hahaha....
  • mfenn - Tuesday, March 11, 2014 - link

    Small nitpick: liquid cooling is an adjective followed by a noun, not a hyphenated compound word.
  • errorr - Tuesday, March 11, 2014 - link

    Really? Who made you lord high pedant?

    Hate to break your bubble but a compound noun is exactly what 'liquid cooling' is. Compound nouns can be combined, hyphenated, or separate words. Perhaps you could look up the definition of compound noun when you have a chance.

    Hyphenation is a stylistic choice to enhance ease of reading. There are no universal orthographic rules per se that prohibit a hyphen there. Orthagraphy is a collection of conventions and not rules. Style rules are required to clarify common ambiguities but unless they actually increase confusion they are mostly harmless and at worst unnecessary.

    I will concede that over the last couple of decades hyphenation has been in serious decline as the internet has come to dominate informal written communication. I remeber the OED took out over 16,000 hyphenations in one of the short versions of the dictionary. That doesn't suddenly make them wrong. I doubt AT has its own style-guide so I can't determine where your authoritative declaration comes from.
  • JarredWalton - Tuesday, March 11, 2014 - link

    Sadly, the style guide on some things with AnandTech simply comes from Jarred, who is old and happens to like hyphens, dashes, and parentheses (among other outdated writing conventions). As you say, style is often very fluid, and mostly I just wanted to make sure all our instances of air/liquid-cooling" were consistent in the article. Which is not to say that another article might not use "liquid cooling". I do have to say that I am constantly amazed at the things people choose to complain about. Welcome to the Internet! :-)
  • SpartyOn - Tuesday, March 11, 2014 - link

    This article is pretty shortsighted as there have been custom LCS brackets available from Dwoods and others for at least two years, if not more. I applaud NZXT for taking this more mainstream, but that price is absolutely ridiculous for what you're getting. I would think the enthusiast marketplace is still the place to turn for these types of mods.

    I've had a Dwoods bracket on my GTX 770 4GB for the past 8 months running a Zalman LQ320 and a PWM Zalman 92mm fan. I've got another LQ320 in my machine for the CPU and it's all in a CM 120 mini-ITX box (I cut out the HDD cage) so space and thermals are at a premium. My temps are 23C at idle, 42-48C in gaming benchmarks, and 50-52C in FurMark torture @ 1350MHz core and 2000MHz memory. I do have an Antec SpotCool blowing on the back of the PCB since there are some RAM chips on the back.

    I think it's a little incredulous to say that there is no easy way to control the fans. As others have noted, SpeedFan is an option or you could do what I did and install PWM fans for both the VRM bracket and the radiator and use a powered PWM splitter from the mobo header. This works really well in a mini-ITX situation like mine as it keeps everything nice and quiet for HTPC use, but ramps up the speeds as needed when I start gaming. I'm sure I could get better fan curves if I did it manually, but for a ready-made hands-off approach, this works really well and cools just fine - even the VRM fan.
  • DukeN - Tuesday, March 11, 2014 - link

    Thank you for posting the review, very timely as just last night I was looking at this and the Accelero Hybrid.

    Sorry to burst the bubble but this was sold out at NZXT before the Anandtech story posted, so no AT effect :(

    This thing is basically not available anywhere ATM, so this could very well be d-wood 2.0
  • mwildtech - Tuesday, March 11, 2014 - link

    These brackets on the right card can be pretty beast. The vrm temps are worrying, especially since you can't measure the vrm temps via software on Nvidia Kepler cards except for a select few high end card.r
  • Subyman - Tuesday, March 11, 2014 - link

    I like how they say there is a 40% reduction in temps by playing with the arbitrary Celsius scale. Looking at the proper Kelvin scale, we note:

    363K down to 323K = A mere 15% temperature reduction.

    Obviously, I'm messing about, but throwing around dumb numbers is fun. I'm glad to see the author call them out on it :)
  • ExodusC - Tuesday, March 11, 2014 - link

    So for $30 I could move over my Corsair H80 off my CPU and onto my exorbitantly hot GPU, then just throw one of my perfectly adequate air coolers onto my CPU?

    I have to say that's pretty appealing. Reference coolers on GPUs are extremely loud and inefficient when under heavy load, I'd love to quiet my system down.
  • EnzoFX - Tuesday, March 11, 2014 - link

    While these units may be quieter than your average stock cooler. I would not consider them quiet by those truly seeking low audible levels. Pumps are inherently noisier than a quiet fan. I don't see this changing anytime soon.
  • Bansaku - Tuesday, March 11, 2014 - link

    Uhm, my Zalman CNPS20LQ (Asetek) pump is extremely quiet. In fact, no fan I have ever put in a case has been as quiet as most AIO coolers (from Asetek). I don't know where you get the idea that pumps are inherently noisier. Perhaps pumps from DIY kits.
  • joelslaw - Tuesday, March 11, 2014 - link

    Am I the only one who doesn't see a gallery on page 2? It says: "The installation of the Kraken X40 kit with the aid of the Kraken G10 on an EVGA GTX770 is depicted in the gallery below." but I only see 1 image below that.
  • Redmenace - Tuesday, March 11, 2014 - link

    Bottom line for the doubters...I had a 7950 in xfire on the top slot, overclocked and overvolted, which ran hot and 3000rpm loud and still hitting 80C. Installed this bracket with an X40 and just slapped the radiator on one of the existing 140mm case intake fans running at a constant 800rpm. Believe me, dumping two 92mm fans turning 3000rpm and replacing it with a 92mm fan (for the VRM's) running about 1300 is much quieter. Did not install any VRM heatsinks. Just adjusted the fan speed on the G10 fan enough to keep temps in the ballpark of the lower, cooler running 7950.
  • biostud - Tuesday, March 11, 2014 - link

    "Being nothing more than a steel bracket, the Kraken G10 essentially is a very simple product, yet it displays tremendous potential. One can wonder why no one else thought of this before (or at least, no one released a product like this until now)"

    Uhm, the Accelero hybrid which has been review on AnandTech in 2012 seems like a very similar product.

    http://www.anandtech.com/show/6531/arctic-cooling-...
  • samsp99 - Tuesday, March 11, 2014 - link

    I have wondered for a while why haven't the graphics card companies had any models with a built-in closed loop cooling solution? There are a couple of top end cards available with a water block, but nothing out of the box with a closed loop solution. The standard PC case design seems to be pretty bad from an airflow perspective for graphics cards, particularly for multiple cards. The cooling design seems to be one of the distinguishing factors between manufacturers, so having a closed loop option would seem to be a good idea. Fans on graphics cards seem to always be the noisiest and first to die in a build.

    Having a built in solution would allow for better fan/pump control so that they could adapt to the workload on the card.
  • Will Robinson - Tuesday, March 11, 2014 - link

    Probably should've tried it on GTX480...that would have been a real torture test.
  • NGR_ - Wednesday, March 12, 2014 - link

    1 - Can AIO watercoolers made for 90W TDP CPU reliably cool ~250W TDP GPUs ?

    2 - In a case like a Fractal Design Define Mini (http://www.fractal-design.com/home/product/cases/d... where should the AIO watercooler be placed ?

    3 - The G10 being out of stock, does anybody know if or when NZXT will be shipping new batches, or if other companies are preparing similar products (AIO watercoolers adaptors like the G10, or AIO GPU + watercooler) ?
  • E.Fyll - Wednesday, March 12, 2014 - link

    1. TDP is a meaningless number in such cases. Some of these kits have a very good °C/W ratio. For exact performance figures, please refer to our AIO kit roundup. We tested all kits in that roundup with a real constant thermal load of up to 340 Watts. It would be fairly easy to assess the temperature of the card you would like to install a kit on by checking the °C/W rating of that kit and the estimated energy consumption of the card. Then, simply multiply the figures. For instance, if the kit has a thermal resistance of 0.1 °C/W and the card has an estimated power drain of 200 Watts, then the temperature will rise 0.1 × 200 = 20 °C above ambient.

    2. Well, depending on which AIO kit you will buy...wherever you like. I cannot possibly make an assessment for every environment, every case "like the Define Mini" and for every kit possible.

    3. That I cannot answer, sorry.
  • JBVertexx - Wednesday, March 12, 2014 - link

    The just announced Fractal Design Node 804 seems perfect for a dual-gpu completely water-cooled configuration using this adapter, although that's a lot of dough needed for 3 CLCs plus 2 of these brackets.
    http://www.fractal-design.com/home/product/cases/n...
  • theNiZer - Tuesday, March 18, 2014 - link

    E. Fylladitakis : Nice review, but I don't get your point when you write: "One such limitation is that the fans of the liquid-cooling kit cannot be thermally controlled in relation to the temperatures of the graphics card. When installing such a kit on the CPU, the fans can be thermally controlled by the motherboard, but that's not an option here. Although fan control is partially possible for the kits with USB interfaces, such as the Kraken X40/X60 that were used for this review, this control is limited to the predefined settings of the user and does not adjust automatically depending on the load of the card."

    X40/X60 works the way you (and I) want - when the GPU heats the watercooling unit, the fans starts spinning more automatically. How is that not automatic and just like the way it works, had it been controlled by the motherboard??
  • StupidCupid - Tuesday, August 19, 2014 - link

    I have a GTX 770ti SC ACX. I am a little confused as to if this is the same card model that you used in the review, as it is sometimes referred to as just a "GTX 770." My card is not listed on the compatibility list, so I am nervous about buying the G10 for it. I think it may just require me to remove the VRM heatsink so that it will fit. Will this remove the factory warranty? Also, if I have to remove the heatsink, what size copper or aluminum heatsink will fit with the unit installed? Do you think the aftermarket, modular heatsinks will be adewuate for cooling my VRM with the factory overclock?

Log in

Don't have an account? Sign up now