Water-cooling solution brand Asetek, which is popular with the OEMs, demonstrated a prototype GPU water-cooling solution for the NVIDIA GeForce GTX 580. The prototype uses an off-the-shelf GTX 580 card branded by PNY (doesn't necessarily mean that Asetek is designing this for PNY). The cooler could be to be a pre-assembled GPU water block loop that replaces the aluminum channel heatsink in the reference NVIDIA cooler. The NVIDIA reference blower is retained to cool other components such as memory and VRM, probably at low speeds.

Asetek put its GTX 580 prototype to test, by overclocking the PNY GTX 580 Enthusiast Edition to 995 MHz core, 1846 MHz CUDA cores, and 1100 MHz (4.40 GHz effective) memory. At these speeds, the GPU was able to score 1285 points at average framerate of 51 FPS in Unigine Heaven 2.0 benchmark. Watch the video embedded after the break for details.

[---]


View at TechPowerUp Main Site

this is amazing . i like reference design of Geforce GTX 580/570 and i dont wanna change that . this gonna be solution cooling for those card . :D

Idk it's hard to believe they got over a 20c (over 30c @stock speeds) difference with that tiny radiator, plus these temps were based on the ambient room temps since it was on a bench and not in a warm case with the surrounding hardware.:rolleyes:

If someone has the 2/3 of these video cards then this cooling system is useless. As most cases wont accept 2/3 independent 120mm rads. I would better stick with a Koolance/Danger Den blocks for cooling my GPUs.:P

Waste of time imo, either go with real water or get the accelero extreme plus. This would take up too much room in a case.

would really depend on the price. if its priced the same as a gpu block, id say its good. Its like slapping an h50 on your gpu (in which H50 was doing ok for a cpu cooler). Would definitely be the best gpu cooler with larger surface to cool.

Does it work on a GTX 480.....?

btarunr995 MHz core, 1846 MHz CUDA cores, and 1100 MHz

disappointing, so close to a 1000 but they didn't push for 5 more mhz

i'd love too see w1z review this

failing 1000 we promptly increased the voltage, whilst the cooling solution held up all the way to 2v, the same could not be said for the card, which exploded

those shader speeds dont add up. 995 core on a fermi would equal 1990MHz shader domain or am i missing something.

Flibolitothose shader speeds dont add up. 995 core on a fermi would equal 1990MHz shader domain or am i missing something.

Maybe they mean 925 instead of 995? Even so, it shouldve been 923 :D

yeah i think 995 is a error probably. with good cooling a high binned chip should be able to hit that speed if the power protection doesn't interfere.

It had more sense what they'd presented back in '08 using a 8800GTX...







And after that, they used it on the HP Blackbird:



This is another (old) new thing for Corsair's "innovative" product line.:D

is it me or do the radiators look way too weak to properly cool a high powered system.

Hmmm, so this is a H50 for your GPU?
Not a bad idea, i can imagine a lot of people who want to branch out into watercooling, but don't have the time/motivation would be interested in this... But then again, i bought a CM Aquagate viva.. it was on offer for about $40 and was the worst purchase i ever made :(

Hey guys,

Just wanted to let you know that I am available to answer any questions =)

Cheers,
Stu

Oh and yes he mistyped up above. This was overclocked to 955 with the liquid cooler. =)

This will come preattached to the card which allows us to maintain a warranty.

All testing was done according to OFFICIAL NVIDIA testing specs for graphics cards. The use of an open bench for graphics qualification is part of that official NVIDIA test spec.

Cheers!
Stu

@asetek_stu,

What specifically are the settings you used in Heaven?

Thanks.

damricI want one for my 6850's.

what he said....

cheesy999disappointing, so close to a 1000 but they didn't push for 5 more mhz

i'd love too see w1z review this

failing 1000 we promptly increased the voltage, whilst the cooling solution held up all the way to 2v, the same could not be said for the card, which exploded

:laugh::toast:

Don't suppose the block it's self has cooling fins on it?

Would make sense since you have air blowing over it any-ways.

Arctucas@asetek_stu,

What specifically are the settings you used in Heaven?

Thanks.

I used default settings with AA turned up to 8x =)

asetek_stuI used default settings with AA turned up to 8x =)

OK, but the question I am asking is; what are the default settings?

Renderer= DX9, DX10, or DX11?

Mode= Resolution and Windowed or Fullscreen? (8X Anti-Aliasing)

Shaders=?

Textures=?

Anisotropy=?

Tessellation=?

I presume you saved the results or took a screenshot?

Thanks.

I want to see what's under the shroud, I love the idea of the sealed loop on the core with heat pipes with a smallish heatsink/cooling plate for the VRM and RAM as shown here.


Depending how it performed in reviews and of course what card the cooler is on i would take it over the stock cooler and most after market coolers.

It reminds me of wishing i could afford the sapphire toxic x1950xtx :laugh:


*edit*
Hmmmm that gives me an idea, hey asetek stu, you should get someone to contact sapphire and convince them to partner up with asetek for the next generation of toxic cards :D

t77snapshotIdk it's hard to believe they got over a 20c (over 30c @stock speeds) difference with that tiny radiator, plus these temps were based on the ambient room temps since it was on a bench and not in a warm case with the surrounding hardware.:rolleyes:

A single 120mm rad provided a 30C drop in load temps for me on an overvolted HD2900XT. It's very possible with a high flow fan.