bit-tech.net

Water-cooling the AMD Radeon R9 290X

Comments 1 to 25 of 40

Reply
Maki role 10th December 2013, 11:16 Quote
Must admit I simply love Aquacomputer's blocks, if only they covered more of the PCB though. I've ended up not using this one for a 290 as there is a lot of PCB, which doesn't look so grand really.
sandys 10th December 2013, 12:29 Quote
It covers the chip, the ram and both the VRM points, nothing else is needed, my EK blocks are much the same.

Might be worth noting that r290 and r290x use the same die and board so equally this applies to them for those who might not know.
Maki role 10th December 2013, 12:45 Quote
Quote:
Originally Posted by sandys
It covers the chip, the ram and both the VRM points, nothing else is needed, my EK blocks are much the same.

Might be worth noting that r290 and r290x use the same die and board so equally this applies to them for those who might not know.

Yeah my complaint isn't really about it being needed, just not looking as nice. EK sells blocks that cover a similar area, but they also sell ones that extend to cover almost all of the PCB itself, even if they offer no additional cooling (I opted for Titan XXL blocks for that reason in my rig). You can see what I mean here:

http://static.scan.co.uk/images/products/2283174-a.jpg
Panos 10th December 2013, 13:11 Quote
Guys, something is wrong with those benchmarks, and I would expected from bit-tech better job.

On 1920x1200 with 8AA the stock GTX780 has 2372 and the 290X under air 2176.
While few days back, at 2560x1600 with 0AA, the stock GTX780 had 2298 and the 290X under air 2157!

Which means that the PNY XLR8 OC or ZOTAC AMP! GTX780s which are 10% faster than the stock 780, beat here a water cooled 290X @ 1155?

(Before someone calls me a AMD fanboy, I have a PNY GTX780 XLR8 and i7 4820K, just stating the obvious)
Combatus 10th December 2013, 13:42 Quote
Quote:
Originally Posted by Panos
Guys, something is wrong with those benchmarks, and I would expected from bit-tech better job.

On 1920x1200 with 8AA the stock GTX780 has 2372 and the 290X under air 2176.
While few days back, at 2560x1600 with 0AA, the stock GTX780 had 2298 and the 290X under air 2157!

Which means that the PNY XLR8 OC or ZOTAC AMP! GTX780s which are 10% faster than the stock 780, beat here a water cooled 290X @ 1155?

(Before someone calls me a AMD fanboy, I have a PNY GTX780 XLR8 and i7 4820K, just stating the obvious)

Hi Panos. I'm not entirely sure what you mean but you do know that if you run Unigene at different settings, you get different scores? So if you run it at 1,920 x 1,200 with 8x AA, you'll get a different score in the same card as if you tested at ,2560 x 1,600 with 0AA. This is why a few days back we got different results, and in addition, the water-cooled tests were re-done on an entirely different system.

As such, you can't really compare like-for-like these tests in this article with all our other graphics benchmarks as it wasn't conducted in our usual graphics test system due to practical reasons. I hope this explains why the results are different!
Seb.F 10th December 2013, 13:49 Quote
I think what he's saying is that on the same settings, he was expecting the watercooled 290X to beat the 780, where at present, it isn't?
Combatus 10th December 2013, 14:10 Quote
Quote:
Originally Posted by Seb.F
I think what he's saying is that on the same settings, he was expecting the watercooled 290X to beat the 780, where at present, it isn't?

Ah right okay. Well it's a pretty close call - watercooling the card does add a lot of performance compared to both 100% and 40% fan speed settings especially. However, we're testing using entirely different hardware and settings to other graphics reviews but the point of the article, really, is just to highlight the impact watercooling has on the R9 290X - here we used the same card, the same settings and the same system so I can assure everyone that these results are not wrong ;)
Panos 10th December 2013, 14:54 Quote
That is great Combatus. Cheers.
TheMadDutchDude 10th December 2013, 15:19 Quote
I'm confused about the temperatures. How does the core get to 72c at various different settings, including slower fan speeds? There's absolutely no way a card will stay as cool at 40% as it does at 100% fan at the same clock speeds. I've heard a 290X and seen the temperature differences. Are you sure they're correct?
law99 10th December 2013, 15:20 Quote
be nice to see more wc testing... but I appreciate it is a ball ache what with leak testing etc all adding to the time and potentially expensive if you can't get the items as press items - let alone what ever you think your time is worth.
sandys 10th December 2013, 15:47 Quote
Quote:
Originally Posted by TheMadDutchDude
I'm confused about the temperatures. How does the core get to 72c at various different settings, including slower fan speeds? There's absolutely no way a card will stay as cool at 40% as it does at 100% fan at the same clock speeds. I've heard a 290X and seen the temperature differences. Are you sure they're correct?

The card throttles clocks etc to not exceed 95c max temperature any further, the graph is delta T to ambient so ambient is probably 23c
Seb.F 10th December 2013, 15:54 Quote
I was getting confused there too, didn't notice the delta!
Combatus 10th December 2013, 18:26 Quote
Quote:
Originally Posted by sandys
The card throttles clocks etc to not exceed 95c max temperature any further, the graph is delta T to ambient so ambient is probably 23c

Correct! :D
schmidtbag 10th December 2013, 18:34 Quote
I'm glad someone finally benchmarked these cards under water cooling. In a multi-GPU and/or CPU overclock setup, it seems like water cooling is a must have if you want a reliable and quiet system. I'm sure the water cooling companies are thrilled, since water cooling was almost deemed unnecessary for the last couple years, except in some cases.

I also particularly like how water blocks slim down the entire GPU, allowing you easy access to more PCI-e slots. Unfortunately, some of those GPUs still use both slot compartments due to the DVI or DP ports, but, if you have any PCI-e based SSDs, RAID cards, or maybe a GPU for PhysX, you could still use those since you don't need to reach the outside with them.
Kovoet 10th December 2013, 18:39 Quote
Damn I wish I knew how to water-cool as I'm too kak scared. Corsair H100i very easy but this and yet so damn tempting
schmidtbag 10th December 2013, 18:48 Quote
Quote:
Originally Posted by Kovoet
Damn I wish I knew how to water-cool as I'm too kak scared. Corsair H100i very easy but this and yet so damn tempting

I've never done liquid cooling myself but as far as I'm aware, there are fluids you can use that aren't conductive. Personally, I've considered doing a computer completely submerged in mineral oil (in a fish tank), but I'm not sure if you can use mineral oil in conventional water cooling systems. I'm guessing the viscosity might be a problem.

From what I've heard, the best way to do a water cooled system is to only power on the cooling system but disconnect everything else. Keep an eye open for leaks and if it happens to spray everywhere, you'll be fine. Contrary to what many people believe, water doesn't hurt electronics as long as they're uncharged. Every time I make an upgrade to my (desktop) computers, I take out most of the parts and clean them under a faucet and just wait a day or 2 for them to dry off. Its cheaper and sometimes more effective than air cleaners, and makes the stuff look like new when you're done.
somidiot 10th December 2013, 20:13 Quote
I'd be curious to see the GTX 780 or 780 Ti water cooled vs this same setup across the regular benchmark suite, see how far each card can be pushed on water vs. air.
law99 10th December 2013, 20:21 Quote
Quote:
Originally Posted by schmidtbag

From what I've heard, the best way to do a water cooled system is to only power on the cooling system but disconnect everything else. Keep an eye open for leaks and if it happens to spray everywhere, you'll be fine. Contrary to what many people believe, water doesn't hurt electronics as long as they're uncharged. Every time I make an upgrade to my (desktop) computers, I take out most of the parts and clean them under a faucet and just wait a day or 2 for them to dry off. Its cheaper and sometimes more effective than air cleaners, and makes the stuff look like new when you're done.

I've had the equivalent of a whole mother ****ing bunch of water all over my **** and all I did was dry it out. Leak test the bits separately outside the case, then move inside and leak test for a solid 12 to 24 hours powering only the pump using the paperclip method on a psu not connected to your PC... then after the green light, you can go right ahead and turn it on.

It is really simple, really, as long as you are methodical in testing and don't get carried away(add an extra trite really? Nah... ;) ).

Only downside is it is more effort to chop and change components that are in the loop. The temp difference is a god send though. Love not worrying about overclocking... really is night and day. Can't go back. Only those all in ones could tempt me, but I'm more worried about one of those exploding than stuff I tightened up and tested myself.

I was quite lucky adopting it at such a late point in my life also as it's all been figured out by other people by now. All the fittings, hose, pumps, rads etc.... simples.
Panos 11th December 2013, 11:36 Quote
Quote:
Originally Posted by somidiot
I'd be curious to see the GTX 780 or 780 Ti water cooled vs this same setup across the regular benchmark suite, see how far each card can be pushed on water vs. air.

With that need to put a modified BIOS also that unlocks the maximum voltage, that is hard set to 1.2v to all. (but EVGA Classified)
rollo 11th December 2013, 11:54 Quote
1.2v should be enough to max the clocks under water even. you would need more exotic cooling methods to push much passed 1.3v anyway.

Only thing that confuses me is with these benches how is the 780 on air faster than this is on water in unigene.
Gareth Halfacree 11th December 2013, 11:55 Quote
Quote:
Originally Posted by law99
I've had the equivalent of a whole mother ****ing bunch of water all over my **** and all I did was dry it out.
Tip for the future: if your precious electronics get water on 'em, wash them in isopropyl alcohol. It forms an azeotrope with water, meaning you can easily get rid of the water along with any mineral residue (more a problem for tap water, rain and the like than distilled water-cooling fluid, but still) without harming anything. It's a much more guaranteed fix than just waiting for the water to evaporate and hoping there's no conductive residue or corrosion anywhere.
sandys 11th December 2013, 13:54 Quote
Quote:
Originally Posted by rollo
Only thing that confuses me is with these benches how is the 780 on air faster than this is on water in unigene.

The benchmark and setup chosen (1080p) does not in away real way load up the gpus enough with that CPU i'd imagine, did a quick test myself and my 290 non X pulls the same score as a 780 in that benchmark with those settings at stock clocks, I have a stronger CPU (3770 @4.5)

Its all set a bit low to really stress it and so comes down to efficiency of the system and general variance.
Corky42 11th December 2013, 14:50 Quote
That's odd because i thought Unigine Heaven Benchmark was made specifically for testing GPU's, as in Accurate results due to 100% GPU-bound benchmarking.
sandys 11th December 2013, 14:53 Quote
don't think you get around Os/Driver overhead anyway you cut it, at such a low resolution despite 8xaa the GPUs such as these aren't having a hard time.
Platinum 11th December 2013, 14:56 Quote
72 degrees under load? What card have you got? my reference design one hits 95 under load at stock clocks?
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums