Arctic Builds 'Hybrid III-140' Liquid Graphics Card Cooler With 140 mm Radiator

Status
Not open for further replies.

alidan

Splendid
Aug 5, 2009
5,303
0
25,780
I never got the point of buying a GPU cooler for $120. Why not just spend $120 on a stronger GPU?

because if you have one of the stronger/strongest gpus this is it.
keep in mind, replacing the cooler on a gpu is more for sound than it is for headroom, my little brother uses a custom loop, there is literally no noise at all from the computer when its in use, above background noise... he has 3 b gear 100+cfm fans attached to it, though they go through a fan controller so they aren't making their full 35db of noise.
 

infinitemerald

Reputable
Aug 13, 2015
12
0
4,510
You could easily buy a HG10/Kraken G10 for $30-40, a [120mm ($40+)/140mm ($65+)/240mm ($65+)] AIO, and a few Copper heatsinks for the VRM for <$5.

$75 for 120mm, or $100 for 240mm.

For this, you're not paying extra for pre-fitted cooling as you still have to attach it to your block, and you're probably not paying extra for better cooling or sound either. This product serves next to no purpose now. A year or two ago, maybe but definitely not now.
 

rubix_1011

Contributing Writer
Moderator
I always laugh when I see these closed loop coolers smacked onto GPUs...no thank you. Either do it right with good air or do it right with good water...not a half-ass frankenstein so users can just say 'look, I have watercooling'. It's like calling your Honda Civic with a giant wing and stock engine a 'race car'.
 
I sigh :) .... the math just isn't there and nothing about CLCs makes sense.

1. We have 2 x 140mm CLCs that can't beat air on a CPU cooler sitting on an 84 watt CPU so how in the begezzes did they come up with the brilliant idea that a 1 x 140mm is good for GFX cards from 150 - 300 watts ?

2. The H100i, prolly the most well know on the CLCs, utilizes a pump capable of pumping 0.11 gpm. If you have ever looked at a radiator test on a half way decent water cooling site, the graphs show performance rising sharply at about 1.0 GPG, less on fill cover GPU blocks because of the large thermal mass.

3. Anyone with 11th grade chemistry under the belts (or anyone who has ever owned a boat, especially in slat water) knows what galvanic corrosion is. In these copper block / aluminum rad CLCs. the radiator serves as a sacrificial anode sending electrons and corroding away while the gunk collects on the copper fins in the block. Not to mention the thermal inefficiency of the rad in the 1st place.

If ya want all in one water, and don't want to go custom loop to cool ya 970 and CPU, grab:

Swiftewch H240-X
Full Cover EK Water Block
Pair of fittings and some extra tubing / coolant

Two 970's ? Grab an extra radiator and you're golden. At 1250 rpm, each 140mm of rad is good for about 135 watts of theoretical heat load at a Delta T of 10C


2 x (220 watt OC'd GFX cards) + 135 watts (overclocked i7) ~575 watts

575 / 135 = 4.25

So with two 2 x 140mm rads, you will have 540 watts at 10C Delta T or about 10.6C at 575 watts theoretical loading.





 

rubix_1011

Contributing Writer
Moderator
I agree with you, only problem is, the masses don't want to do the math nor do they care. Shiny package allows you to say 'I watercool!' to the uninitiated and uninformed. Yes, closed loop coolers are a form of watercooling, but it's the lowest common denominator and only allows you to claim it simply because it follows the same principles. In terms of performance, the pumps are dismal, the radiators are aluminum in most (as mentioned) and they're typically underperforming due to the poor Delta-T of the combined hardware.
 

Eximo

Titan
Ambassador
Having tried pretty much all of the solutions out there in various builds I can say that even a thin 120mm is enough for a single GPU even after overclocking. Really reduces the maximum temperature and gets it out of the case, so they are more then a gimmick. Expensive though.

Swiftech solution gets my vote as the easiest watercooling setup to do. It is what I am using currently.

Did just have the Arctic hybrid cooler on my old GTX580 crap out after 2ish years of use. Some form of pump failure, though the motor still turns, haven't taken it apart yet to see what the issue is. Other one is still working though. Now that old system has the original DCUII cooler on it, the arctic cooler mounted on a tri-SLI x58 motherboard in a top-mount PSU case that doesn't have enough slots for everything. Truly a masterpiece of second hand parts.
 

rubix_1011

Contributing Writer
Moderator
A 120mm rad is really only good for around 150-160 watts of dissipation - your typical enthusiast GPU is in the neighborhood of 170-250w, depending on card and chip.

So, yes, if you are OK with a poor delta-T, then yes, a 120mm radiator is fine to cool it. Considering load temps of GPUs are in the 70-80C range, seeing load temps in the 50-60s is an improvement.

The Swiftech H220-X and 240-X are great cooler solutions, if that's what you are referencing. Overall, Swiftech does make some decent stuff, although many of their radiators are a bit outdated.
 

Eximo

Titan
Ambassador
Didn't have any issues running my overclocked GTX580 with a teeny 120mm radiator. That was a 244W stock chip, and I overclocked it quite a bit. Max temperature reached was 65C so your estimate is spot on. The real benefit was over the axial coolers that were on there originally and getting the heat out of the case. Far better then the 100C temps and instant reboot.

 

rubix_1011

Contributing Writer
Moderator
Yes, GPUs have always benefitted from water/liquid cooling simply due to the amount of heat they generate and their knack for having a high thermal threshold to start with.

The real issue with water/liquid cooling is getting the end-user to understand the expectations of the chosen cooling solution based on the cooling delta they wish to achieve. If there isn't one, then any result lower than stock cooling is acceptable. Once someone begins to determine what they would like to achieve, then you begin to see the performance gap of some cooling solutions vs. those of others. There is definitely a market of 'novelty over performance value' for many products.
 
For $120 US you can get a full cover water block which would be superior. Or you could get an H60 and an NZXT Kraken G10 for around $80. Not really seeing the appeal at that price pint.
 

rubix_1011

Contributing Writer
Moderator
Not sure I follow what you're saying. This is about a closed loop cooler on a GPU, not on a CPU. A full cover GPU block is around $120, correct, but you would still require pump, radiator, tubing, fittings, etc. While the cost of closed loop coolers is attractive, their performance is not.

The term 'you get what you pay for' applies here.
 


1. That is in conflict with all available published lab testing.

It is plainly clear that 1 x 120mm is not quite up to "design targets". Easiest way to look at it is this. If it takes a H100i to do almost as good a job as an air cooler (i.e Noctua NH-D14 / Cryorig R1) with 2 x 120mm @ 2700 rpm for a CPU that generates only 84 - 135 watts overclocked, then how can we expect half that radiator area to handle GPUs with 1.5 to 2.0 times the wattage ?

In a well ventilated case, one should not expect any higher GPU overclocks than air cooled models. I have two Asus DCII 780s w/ EK Water Blocks and a 26% OC. A subsequent build w/ two MSI Gaming 780's (air cooled) did 27%. Obviously, heat is not a limiting factor here. Although, it must be said that using a non reference PCB with a beefed up VRM is a distinct advantage here which reference cards and even some non reference cards (i.e. EVGA SC which has no reference cooler but reference quality VRM) don't have. Taking the data from actual laboratory testing from martinsliquidlab .... "Busting out the science behind liquid cooling!"

https://martinsliquidlab.wordpress.com/2012/05/01/alphacool-nexxxos-st30-360-radiator/4/

Here we see that at 1400 rpm the 30mm thick Alphacool ST30 is able to dissipate 199 watts, 242 at 1800 rpm. Since the main reason to water cool these days is sound reduction, since performance gains are minimal, most folks will use 1250 rpm fans and keep them running below 850 rpm.

But even using the Swiftech's 1800 rpm fans @ 1800 rpm, that 242 watts for a 360mm rad works out to just 81 watts per 120mm.... just enough for a stock 4690k ... and no where near enough for a 970, 980, 980 Ti or the hotter running R9 series. Most 970's (Gigabyte's draws way more power then the competition) pull over 200 watts. All data below from Techpowerup for MSI cards, Gigabyte in ( ) when available

980 Ti = 279 - 281 watts (293 - 359)
980 = 205 - 207 watts
970 = 192 - 213 watts

MSI Afterburner, if memory serves, allows the following increase in Power Limit via afterburner

970 = 110%
980 = 122%
980 Ti = 120%

That makes the overclocked numbers ...

980 Ti =308 - 311 watts (322 - 395)
980 = 250 - 253 watts
970 = 230 - 256 watts

So at 1800 rpm, the 1 x 120mm delivers just 81 watts of cooling but we only expect the radiator fins too provide about 60% of the cooling for the following reasons:

-This are peak not average loads
-Additional cooling is accomplished by radiation from the block surfaces, tubing, fittings and even the radiator shroud surface (which is not measured nor included in the test data).

So taking the overclocked numbers from above x 60%

We need the rad to handle about 138 watts (230 x 60%) for the 970 and 1 x 120mm gives us just 59% of that.
We need the rad to handle about 150 watts (250 x 60%) for the 980 and 1 x 120mm gives us just 54% of that.
We need the rad to handle about 185 watts (308 x 60%) for the 980 Ti and 1 x 120mm gives us just 44% of that.

BTW, Delta T is not the difference between the GPU temp and anything... Delta T is the difference between the ambient air and the coolant. With a typical 80F (27C) indoor summer temp, this means your coolant w/ delta T of 10 will top out at 37. In the above instances, we are looking at 17 - 23C delta T. So at 27 C (80F) ambient, coolant temps are as high as 50C (122F) instead of 37C. Your GPU will be significantly hotter. The efficiency of the heat transfer is directly proportional to the delta T between coolant and GPU. So say you have a moderate load game playing and GPU temps are at 63C .... the heat transfer rate (GPU => Coolant) will be only 13C, about half as effective as it would be with the 10C Delta T coolant temp (26C).

And that's at a relatively noisy 1800 rpm. At 1250 rpm, still a bit of noise but bearable if using speakers. The 1 x 120mm gives just 60 watts of cooling (75%) and at a silent 850 rpm, just 43 watts (50%) .

On our 780 SLI rig, when coolant temps are 33C, the GPU runs at 39C (5 x 140mm of rad w/ 10 rad fans and 5 case fans) at 1200 rpm and 44C at 850 rpm which is where I have them set to not go any faster. The VRM however, even with the best water block for VRM temps is at 59C. We are pumping about 1.5 gpm while testing (max is 2.25) ... the CLCs do just 0.11 gpm.

Where CLC type water cooling typically fails is in VRM cooling.

HkEXMui.png


As you can see above the EVGA Hydrocopper runs 30C above the EK. At 63 above say a 27C ambient, that's 90C ... the Hydrocopper doesn't do well but at least it's a full cover WB, which the CLC solutions are not. You may have GPU temps in the mid to high 60s even but your OC will likely be limited by your VRM temps rather than your GPU temp here.

In short ....

- I don't know that the single 120/140mm fan is capable of transferring enough heat to actually lower GPU temps.
- The VRM remains air cooled and may or may not do as well a job as the original equipment, especially w/ the small 80mm fan
- Coolant temps will be high
- Noise reduction, again the one area where water cooling can have a significant impact, is not accomplished due to the high speed fans.

2. Not sure what you mean by "getting [the air] out of the case" but if you have the radiator fans exhausting air from the case, you have installed it contrary to Swiftech's recommendations (as well as Corsair and every other water cooling manufacturer as far as I have seen). The air will get out of the case quite quickly, w/ proper fan orientation it will be force air thru the rad, into the case and right out the rear grille. We utilize 6 thermal sensors (0.1C accuracy) on the test rig and the fog machine testing gives a great picture of air flow. There is zero concern about the air exiting the radiator heating up inside components as it is immediately exhausted before it gets there.

3. Water cooling modern GPUs does not generally result in increased performance, as long as you have adequate case ventilation. At least, on nVidia's side anyway, I have not been able to do any better than on air since the 7xx series. The reason is that nVidia has, both physically and legally with their vendors, limited what is possible to do voltage / power wise

The main reason to WC GPUs nowadays, assuming you are not facing indoor temps much above 85C, is to reduce noise . Though the higher end cards can push the thermal limits, especially in SLI configs where a fan on the back of the HD cages blowing air between the cards is recommended, I haven't had any throttling occur.

4. I agree that the Swiftech H240-X option is far superior to CLCs and just about ideal for a 970 / CPU loop. Let's take that 135 watts for the C'd CPU and 230 for the OC'd 970. With the 60% factored in that's 219 watts of rad cooling needed. You can d/l martins data extrapolated out for every rad he's tested here:

http://www.overclock.net/t/1457426/radiator-size-estimator

Looking at the Swiftech file, at the fan's rated speed of 1800 rpm, you get 98 watts per 140mm and 119 in push / pull

So 98 x 2 = 196 watts or 89% of what ya need for delta T of 10C .... Delta T would be 11.2C. You can mount one extra fan in pull (can't install fan on reservoir side) which would give you 98 + 119 or 217 watts or 99%. I begged them to do a H340-X but no luck. That would easily allow a CPU / GPU loop w/ any single GPU card and most cases that fit 2 x 140 also fit 3 x 140mm. SLI / CF would then easily be accommodated by adding an extra radiator in front or bottom.




H240-X ($150) + 970 water block ($120) = $270.

$120 GPU CLC + $110 CPU CLC = $230.

The 17% cost premium gets you:

a) All copper components instead of alum / copper mix
b) No galvanic corrosion concern
c) Reservoir
d) Ability to expand loop
e) 1.0 gpm pump instead of two 0.11 gpm pumps
f) Much quieter fans
g) Huge leap in performance
h) PWM fan PCB / controller

Edit: Forgot, you'll need to buy two fittings for about $3 each for WB.

Edit 2 : If responding I'm sure many following the thread would ask that ya don 't quote the entire post so as not to make reading subsequent posts cumbersome to read . There was a lot of info to present and hopefully, it will be helpful to some.... but reading or better said scrolling past it 3, 4 , 5 times can be burdensome :)

After hitting reply, you can edit / remove less relevant parts.



 

rubix_1011

Contributing Writer
Moderator
JackNaylorPE - by the way, I'm in the process of building a watercooling lab similar to Martin & Skinnee for our forum testing. Problem is, we need more of a following in our forum community.
 

Eximo

Titan
Ambassador
Neat write up, but I have operating cards with an Arctic Accelero Hybrid with a TDP greater then GTX970, GTX980 and comparable to 980Ti. Not sure what your target temperature is in your table.

Arctic themselves lists the original hybrid at 320W capacity, as compared to this new version at 300W, and I find it difficult to dispute thesemaximum capacities as my cards did not overheat despite being well over the wattage in some of the numbers you posted. Certainly ran warm at 65C for watercooling, but it was effective at 900Mhz at 1.15 volts if I recall correctly. Which would have been about a 10% power increase over stock. So about 270W.

Not saying it is ideal cooling, but can be a solution when you can't find any waterblocks and air cooling was stifled due to the thickness of the cards and PCIe slot spacing, as was my situation.

I don't like the idea of blasting warm air through the case so I generally have my radiators as exhaust. Were my motherboard also watercooled it might allow for lower temperatures overall, but I would rather keep the interior as cool as possible for drives and the like.

Would be neat to add EVGA's all in one GPU cooler to some sort of comparison if anyone is so inclined.
 

SylentVyper

Reputable
Dec 30, 2014
98
0
4,660


I fully agree with what he's saying. The 295x2 is running what amounts to a pair of 290x's cooled by a single 120mm AIO cooler, and is more than sufficiently cooled, even better than all the other single GPU cards that are air cooled.

http://www.anandtech.com/show/7930/the-amd-radeon-r9-295x2-review/17
 

rubix_1011

Contributing Writer
Moderator
I wouldn't say that 71C is running very cool, considering most people running watercooling loops with GPUs and a decent delta are seeing 35-45C load temps, depending on ambient.

Just having a radiator on a cooling solution doesn't necessarily mean you have 'enough radiator' for the cooling solution based on load wattage.

That's a terrible delta if 71C is your reported GPU core temps, especially if you're comparing to [stock] air coolers that aren't turning out temps a whole lot better.
 

SylentVyper

Reputable
Dec 30, 2014
98
0
4,660


Wasn't at all saying it's comparable to a custom loop, but showing that a single 120mm radiator is plenty to sufficiently cool a pair of 290x's. 71 is far below what would be considered dangerous. Again, not near a custom loop, but that wasn't what we were saying.
 
Status
Not open for further replies.