Gigabyte’s WindForce Has The Fury In It: First Custom PCB R9 Fury Revealed

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Honestly this version of the Fury I would consider, however, the thing that makes me want the sapphire line more is the fact that those cards ONLY have displayport and HDMI, this makes them also better for what i have planned in the future which is a water cooled rig with maybe two of these puppies. That single slot will allow for lots of other pci devices or more cards. Other than that it looks to be a good card.
 
To me the biggest downside to the Fury X is the inability for OEMs to customize it. I think they could easily come up with a superior air solution for a Fury X.

Maybe we will see higher clocked Furys with custom PCBs and coolers but so far even the Sapphire Tri-X Vapor-X design and Asus Strix DCUIII have not pushed it much higher.
To me the biggest downside to the Fury X is the inability for OEMs to customize it. I think they could easily come up with a superior air solution for a Fury X.

This is true, although with a card of this caliber, I wouldn't think twice of putting a CLC on it in the first place. 95% of the time, it's going to result in lower temps and lower noise levels than any air solution.
 
This is true, although with a card of this caliber, I wouldn't think twice of putting a CLC on it in the first place. 95% of the time, it's going to result in lower temps and lower noise levels than any air solution.


Think about it.... 2 or 3 fans versus a single fan trying to the same job of the original three fans with just one; that single fan has gotta be spinning awful fast to provide the same cooling level as the original three.

Hasn't worked on CPUs, why on GPUs ? Let's look at the Noctua DH-15 and the Corsair H100i both of which have two fans. DH-15 is 33 dbA, H100i falls short by 2C in the thermal comparison but is at 68 dbA .... 12 times louder.

http://www.hitechlegion.com/images/cooling/swiftech_h220x/b2.jpg

Asus has a nice approach with their Poseidon. You get both air AND water cooling; so even after you add GPU water cooling to the GPU, you still get the advantage of air cooling. Wonder if we will see Asus make an AMD Poseidon
 


This is true but you have to remember that water and air have completely different properties when it comes to heat conductivity. Water is far more conductive than air and by extension, removes more heat. The H100i you mentioned I have had no experience with, but I can tell you that going from a Cooler Master Hyper 212+ EVO to a Cooler Master Seidon 120V, I had a 10c drop in temps and reduced my total fan count by 1 which lowered noise levels.

I realize this will not always be the case exactly, but you can't compare the conductivity of air to water, it's like apples and oranges.

 


True but there are other factors. Remember that 68dba was from a max fan speed. I have a H100i and I run it in quiet mode and my CPU never goes above 50c under heavy loads, which is pretty rare. There is also the size which for me is a plus. Some of the best heatsinks are pretty large and push into the RAM area with fans attached making you choose a specific set of RAM if you want it to fit.

I chose the H100i for the performance and the aesthetics more than anything.

That said, the Fury X probably does better on water cooling. We probably will not know for sure since they cannot make custom PCBs as AMD has not deemed it possible. Although there is a sort of idea since some of the Fury GPUs can be unlocked to Fury X (very rare) although this could also be problematic since they normally get binned to Fury status when they don't quite meet Fury X standards.
 


Well they are essentially the same cooler, however the fury has different placement of fins because it doesn't need contact with any GDDR5 memory chips. Other than that it should essentially be a similar design.
 


Totally irrelevant; you are missing the point. Water cooling is a multi stage process:

1. Heat is removed from the GPU and transferred to the copper block .... not the subject of discussion
2. Heat is removed from the copper block and transferred to the liquid coolant .... not the subject of discussion
3. Heat is removed from the coolant and transferred to the radiator .... not the subject of discussion
4. Heat is removed from the radiator and transferred to the air passing thru it .... BINGO, here's what we are talking about. This is a metal to air heat transfer, just like an air cooler.

Those 4 processes work in tandem and the cooling ability of the system is limited by the weakest link in the chain and that is step 4. There is no water involved in Step 4 so the extra heat transfer ability of water is meaningless in the context of this discussion.

What happens if the fan dies..... just how well is this cooler going to work ? Coolant temps will continue to rise to the point where the hot water will do very little cooling. The very foundation of water cooling is based upon the fact that air cooling heat sinks are necessarily limited in size. Water cooling works by transferring the heat from the CPU / GPU to a heat sink (radiator) that has a much larger metal surface are than is possible to reasonably fit on a CPU.

As for your experience, it must be recognized that while the Hyper 212 is a "great budget cooler", that does not make it a "great cooler". Put the Seidon up against a comparably priced air cooler and it doesn't stand up. And even against the Hyper 212, it's not all that impressive...especially when one is $89 and the other is $30

CPU_OC_typical.gif

fan_noise100.gif


In the preceding post's image you saw the H100i, generally considered the best performing (thermally) 2 x 120mm CLC out there, getting soundly beat by the cheaper Noctua DH-15.

Here we see the Seidon 240M w/ twice the cooling capacity of the 120 model, barely edging the Silver Arrow air cooler by just 1C. The Silver Arrow gets beat by many air coolers including the Phanteks PH-TC14PE. Noctua NH-D14 and NH-D15, Dark Rock Pro, Cryorig R1

5169_36_needs_edited_images_cooler_master_seidon_240m_water_aio_cpu_cooler_review.png


No matter how good water is in removing the heat from the CPU block, it still comes down to a metal to air heat transfer at the end (radiator of heat sink). And, at the end, it's all about metal surface area and moving air past the metal surfaces. In order to do better than the card's stock air cooler, a cooling solution must:

1. Provide more metal surface area and more fans
2. Provide more air flow (rpm - which comes with additional noise)
3. A combination of both.


True but there are other factors. Remember that 68dba was from a max fan speed. I have a H100i and I run it in quiet mode and my CPU never goes above 50c under heavy loads, which is pretty rare. There is also the size which for me is a plus. Some of the best heatsinks are pretty large and push into the RAM area with fans attached making you choose a specific set of RAM if you want it to fit.

I chose the H100i for the performance and the aesthetics more than anything.

It still doesn't quite work out when ya turn down the rpms on the h100i. Not saying that you can't be not satisfied with your system, just that air coolers still do better in this scenario. . Here's the H100i again compared (same source) at t he same noise level; a level obtained by lowering rpms to a point where all coolers stayed below the 40 dbA target.

b4.jpg


The NH-D15 air cooler still manages to deliver 5C better temps and 6 dbA lower noise when both have rpm lowered to the point when they can maintain < 40dBA. The Cryorig R1 air cooler still manages to deliver 6C better temps and 1 dbA lower noise when both have rpm lowered to the point when they can maintain < 40dBA.

If ya just like the way CLCs look, I think that's perfectly fine but any argument for a CLC over an air cooler on either a thermal performance, noise or cost (bang for the buck) basis is without a factual basis. OTOH, if it's the aesthetics, the reasoning for choosing CLC over an OLC is one I can't fathom given the marginal price difference of comparably performing units.

Other reasoning I have trouble grasping include heat sink size / weight arguments.

1. RAM height - Those RAM sticks with tall toothy heat sinks have just one "cooling function" and that is to "look cool". While DDR2 benefited from the extra cooling capacity, the efficiency of DDR3 and DDR4 makes these toothy things a mere novelty. There is no performance reason to choose a tall set if RAM as compared to a standard profile set..... manufacturers maintain these offerings only cause they look "less plain" in ads. It's pretty hard to "sexy up" a RAM stick

Furthermore, there are air coolers ... for example the cooler that beat the H100i by 6C above that let you put in any RAM you want

r1-universal_vs.png


2. Heat sink weight - I can not understand the concern of a 2 pound load from an air cooler heat sink weight, when peeps are applying 60 - 70 pounds of clamping force as per TIM application instructions.

I wouldn't argue the aesthetics thing but .....

- Size ? The H100i is much bigger than any air cooler; air cooler consists of block + heat sink ... CLC consists of block, pump and .... that giant radiator.
- Thermal Performance - the images above clearly show the best CLCs can not perform as well as cheaper air coolers.
- Noise Level - Again, the images above show the best performing CLCs to be noisier even when turned down to stay below 40 dbA and to get there, you lose 5C of performance
- RAM issues - these have to be *created* by making unwise and unjustifiable RAM module selections but if you must "go there", is still avoidable with coolers like the R1
- Safety - Air Coolers can't do this
http://www.tomshardware.com/forum/281843-29-corsair-exploded

That being said, I remain an avid water cooler. With modern CPUs / GPUs, there is very little, if anything, to be gained from a performance standpoint as on Intel CPUs / nVidia GPUs, you are much more likely to hit a voltage wall before you hit a thermal wall. I chose to do water cooling for noise and aesthetic reasons but CLCs don't fulfill any of those needs.

Noise - I have yet to see a CLC that is quieter than an air cooler when delivering comparable thermal performance. I have 16 fans in my water cooled system (6 case fans and 10 radiator fans) and the only issue I have with it is that it is literally dead silent.... I have oft sat down to a black screen and tried to "turn on" the system, thinking it was off.

By using 5 x 140mm of radiator, the fans run typically from 325 rpm like now when I am typing to 750 + rpm when running a stress test (Furmark) which throws about 680 watts (4 water blocks - CPU, 2 x GPU, MoBo) at the cooling system. Conversely, a 2 x 120 mm CLC will break 60 dbA easily with a CPU at just 4.4 Ghz. To get the performance it does, it must move a lot of air and when ya have just 2 x 120mm of rad to push it thru, that means pushing it thru very fast and that means noise.

Aesthetics - As I said, aesthetics is a personal thing but, for me, I abhor chaos. The sight of tubing running every which way at different curve radii is something I find unsettling :). Rigid tubes in straight, parallel / perpendicular lines gives the build a "pleasing sense of order" and since the box is used to design water treatment plants, it kinda serves as a representation of what we do. So if I am going to WC, it will be with rigid tubing.

Of course, this is not to suggest that what I find alluring would be or should be the same as anyone else's just conveying that aesthetics is different things to different folks. I will note that my build is OC'd to 4.6 GHz on the CPU (72C under stress test) and 26% OC on the 2 GFX cards (39C). We have done other builds that have clocked slightly higher on air tho GPU temps are more in the expected 75 - 80C) ranges and, yes, they are noisier.

With EK jumping into the market along with Swiftech with AIOs that are OLCs rather than CLCs, it's pretty obvious that the days of CLCs are numbered. Unlike CLCs ....

1. OLCs have copper radiators instead of aluminum
2. OLCs don't have mixed metals which promote galvanic corrosion
3. OLCs recognize the fact that corrosion and algae inhibitors lose their effectiveness over time and allow the user the option of changing coolant.
4. Can be expanded to add other blocks (MoBo, GPUs, etc)
5. Do not have weak pumps
6. Some have built in reservoirs
7. Are no more difficult to install than CLCs

A person who is a bit hesitant to jump into custom cooling, can purchase a OLC like the H240-X or Predator 360 for his system with an Asus Poseidon and, when ready, covert to GPU water cooling in a matter of minutes. Unlike a CLC type GPU solution, this retains the complete functionality of the card's original cooling system allowing them to work in tandem and w/o the worry of where and how to mount a second radiator. I would love to see more vendors take this approach, even using a BIOS switch which would kick up the factory OC once water was added.
 

While I like the rest you've got there, CPU cooler weight is very relevant. Compression and torsion stress are two radically different forces, and like most materials, mboards have much lower torsional strength than compression strength. Compressing a cooler onto a board, the force is spread out over the entire socket ( even more area if using a backplate ). However, torquing a cooler ( such as hanging it from a vertical mbaord like many computers do ) puts all that stress onto smaller contact points: the outer edges of the mounting bolts / backplate.

Under static gravity load, the board can take it just fine. However, if your system gets moved, the CPU cooler acts as a lever on the mboard and amplifies every bit of acceleration. A water-cooling coldplate is much lower and weighs less than a tower cooler, and puts much less torque on the mboard during transit or other times it gets moved. If your mboard is mounted horizontally, or if the computer never moves, you probably don't care as much.

It's like a balsa bridge. They can sustain thousands of pounds of straight on compression, but apply less than 100 foot-pounds of torque and they snap. The PCB layers don't even have to snap. All it takes is a tiny crack and the electrical pathways can get screwed up.
 
Status
Not open for further replies.