Watercooled 290x

[MrK6]

According to hw.fr the 1GHz R290X is 9% faster at 1440 than the 980MHz GTX 780. R290X just needs to go higher, there is no other way to put it. 1250 on water for a setup that will cost much more than non reference 500-570 dollar GTX 780s that can hit 1300-1400MHz all day isn't going to cut it.

Some good GTX 780's can hit 1.3GHz 24/7 stable, but almost all cards will require new BIOS/volt mods to reach 1.4GHz as well as a watercooling setup, so none of your price arguments are correct. The thing is, if you actually do the research instead of guessing, you'll find that the 290X is actually much faster clock-for-clock than a 780. For example: http://www.overclock.net/t/1436635/ocn-gk110-vs-hawaii-bench-off-thread/500_100#post_21088075 . Here a 290X @ 1260/5800 is slightly beating a 780 @ 1411/7406. Roughly this is a 12% clock-for-clock advantage for the 290X. Also note that both setups required water cooling, as I mentioned above would be the case. Two other points worth mentioning are that here is a reference 290X vs. a cream-of-the-crop Classified 780, I imagine the 290X's lead will only lengthen once performance-oriented aftermarket models hit the shelves. Secondly, this is a nvidia favored benchmark/scenario, I'd also hypothesize that the 290X will pull further ahead in actual gaming, as BF4 and other stressful benchmarks are already starting to show.

Because BF4 Single Player is what everyone cares about, right?

Ouch.

Better get that Mantle going because they clearly aren't doing something right, like supporting DX11 features?

This is also incorrect, since, as mentioned, SWE also tested multiplayer. I would guess that whatever capturing software this polish review site used is negatively impacting the CPU/memory subsystem substantially enough to hurt AMD performance. It's well documented that FPS in BF4 are heavily tied to CPU and memory speeds. Nvidia has shadowplay, I wonder if they were amateur enough to use shadowplay on the nvidia GPU's and something more beefy on the AMD ones? In either case, these are incorrect results.

 

[BallaTheFeared]

Firestrike favors AMD, those scores aren't on hwbot, nor are they on 3dmark.

What do you think we should do with a test where AMD can disable tessellation on a post that has no official submission to verify settings used?


Let's give it to the R290X for Fire Strike. We'll say the score was legit, a water cooled GTX 780 needs 12% more clock speed to tie R290X in Fire Strike.

R290X is about 13% faster at 1GHz than a 980MHz GTX 780 in Fire Strike.

GTX 780 needs 12% more clock speed to match R290X in Fire Strike, however at stock it is only 9% slower in actual games.

So basic logic says that classified on water is faster than that R290X on water in real world gaming.

Given how power hungry R290X is, and the lack of aftermarket cards. It's pretty obvious that right now aftermarket GTX 780s are far and way better values.

http://www.newegg.com/Product/Produc...82E16814162142

GTX 780 aftermarket will easily outstrip reference R290X performance (not even accounting for noise). On water GTX 780 clearly still has a slight advantage in overall performance.

Whats the upside for R290X? In it's current incarnation it's slower than GTX 780 OC vs OC, either water or air. It costs more to do either, and doesn't even offer an acceptable cooler right now for on air overclockers.

 

[Aithos]

O__O You can put a whole system underwater for 150-250$, it just won't look fancy but will work the same.....Plus this includes the CPU

No you can't. Not with any reasonably decent level of parts anyway...

A bare minimum for a *decent* water loop is $500. Just a good pump (d5 variant) is $100, a good cpu block is $80-120, a gpu block is $100-150, a res is $50, tubing is $25, fittings are $50ish for compression, rads can vary wildly but two 240 will run you $150 easily. That's $575 right there and doesn't count the incidentals like: fluid, velcro, mounting supplies, tools, cleaning supplies. Oh and that doesn't count fans, if you want Noctua fans like the F-12 plan on $100 for just pull, or $200 for push/pull.

The point is that you can't do a cpu+gpu loop for $150-250, it's not even possible because of the cost of the gpu block for a new card like the 290x. Even if you could you'll get mediocre performance compared to what this post is claiming because I can tell you right now he isn't running a bare bones loop. Not to mention if you cheap out on everything it's a lot more likely to break down and need replacing.

Last but not least, for the extra cost a *decent* water loop costs...you can just buy a second GPU. I'll take two cards overclocked on air over a single gpu under water any day. The only situation I can see wanting to put a 290x under water cooling is if you ALREADY had a loop and just need to expand it (new block, a couple fittings and some extra tubing).

Edit: everyone suggesting an AIO cooler for one of these...just stop. Please. That's ridiculous and wouldn't perform any better than the stock cooler. You need to also cool all the other chips on there to significantly overclock, you can't leave them just exposed without even a sink or you're 100% going to throttle your memory. Putting a GPU under water isn't something you just do for fun, it's something you do if you already have a custom loop and you intend to overclock your entire system. If you are cheap, stick with stock cooling.

 

[FredNotFound404]

^ Post Wall of test, doesn't bother to read the rest of the thread, feel free to properly read it....

 

[schmuckley]

I highly doubt that. But, if you have the ability to machine your own blocks, more power to you.

It would be right @ $300 with a full-cover GPU block.

http://www.overclock.net/t/406256/the-official-bong-lovers-club-56k-fail/0_100

pump $44
tubing..eh..$20
barbs..Maybe $20 max
zipties.. $4
GPU block $40-$110
CPU block $60-could go used and cheaper..
Showerhead $3
48-qt cooler $25
Sewer pipe + wye $15
L-brackets $3
glue n stuff $10
==================
Total: $244 (gpu-only block) -$314 (full-cover block)
OOPs..I forgot RAMsinks: $14

I never said it would fit in a case or be pretty-looking
..but my res water is always below ambient after running for a while

 

[MrK6]

Firestrike favors AMD, those scores aren't on hwbot, nor are they on 3dmark...

Again everything you post is incorrect. People might actually take your argments seriously if you found sources to back up what you're thinking, much like I did, instead of trying to force your opinions by generalization (incorrectly at that). What games? What settings? From where? Where are your sources?

Learn to formulate a basic argument with references and then your posts might have merit.

 

[FredNotFound404]

3DMark is like any other game and it depends also on optimization, personally from all the benchs I'v seen this year, it does seem to favor AMD, the bench results don't always translate to real world situations, then again no one should be judging cards tro just that score alone, most of the times a card performance on a game tends to vary depending on optimization, the trick is to pick the cards the perform well all around. That's is my opinion anyway....

 

[Tempered81]

Again everything you post is incorrect. People might actually take your argments seriously if you found sources to back up what you're thinking, much like I did, instead of trying to force your opinions by generalization (incorrectly at that). What games? What settings? From where? Where are your sources?

Learn to formulate a basic argument with references and then your posts might have merit.

Good intentions, but didn't you hear? 3Dmark11 tess off: 1.9ghz Titan quad sli < 1.5ghz 290X Quadfire < HD 7950 < GTX 470 SLI

Warning issued for personal attack.
-- stahlhart

 

[BallaTheFeared]

Again everything you post is incorrect. People might actually take your argments seriously if you found sources to back up what you're thinking, much like I did, instead of trying to force your opinions by generalization (incorrectly at that). What games? What settings? From where? Where are your sources?

Learn to formulate a basic argument with references and then your posts might have merit.

It's an impossible task and you know it. Ironically when saying what I said was wrong and formulating basic "argments" to counter it your post lacked those easily found sources.

You aren't going to find a 1400Mhz GTX 780 vs 1250MHz review to site as a source.

Or are you asking me to post proof that any review running 3DMark would have shown, the performance advantage of the R290X in 3DMark isn't represented in the overall performance lead it obtains?

Roughly R290X is 13% faster in basic Fire Strike, a test which is run at 1080p. At 1080p the R290X in "uber mode" which only fluctuates slightly in a test that gobbles up 270+w between 990 and 1000 MHz it is only 7% faster than the 780 locked to 980MHz.

http://www.hardware.fr/articles/910-25/recapitulatif-performances.html

http://www.vortez.net/articles_pages/amd_radeon_r9_290x_review,13.html


If you do the math, you'd need about a 1337MHz GTX 780 to match a water cooled R290X @ 1250. Of course I have nothing to back this up as far as source, just piece meal based on observations and sketchy at best clock speeds that very few reviewers are actually giving for either product.

This is also incorrect, since, as mentioned, SWE also tested multiplayer. I would guess that whatever capturing software this polish review site used is negatively impacting the CPU/memory subsystem substantially enough to hurt AMD performance. It's well documented that FPS in BF4 are heavily tied to CPU and memory speeds. Nvidia has shadowplay, I wonder if they were amateur enough to use shadowplay on the nvidia GPU's and something more beefy on the AMD ones? In either case, these are incorrect results.

I missed this gem, so your scientific analogy of the phenomenon lead you to the conclusion that programs that don't even record FPS are what is causing a reduction of FPS when recording FPS?

So just to confirm I'm not misreading your properly structured and well backed scientific method confirmed observation, you believe when they record FPS numbers, they're actually recording gameplay to a drive?

 

[GaiaHunter]

I missed this gem, so your scientific analogy of the phenomenon lead you to the conclusion that programs that don't even record FPS are what is causing a reduction of FPS when recording FPS?

So just to confirm I'm not misreading your properly structured and well backed scientific method confirmed observation, you believe when they record FPS numbers, they're actually recording gameplay to a drive?

Who cares.

You have one review where the 290X is faster in single player and way slower in multiplayer.

Then you have a review where the 290X is faster in both.

The simpler solution is that the 290X is faster in both and something went wrong with one review.

 

[BallaTheFeared]

The care is what went wrong.

Was it driver or OS?

I'm sure if it turns about to be a Win7 vs Win8.1 deal then someone (thousands) will care.

Sure as shire had nothing to do with ShadowPlay or other recording software, lulz.

 

[GaiaHunter]

The care is what went wrong.

Was it driver or OS?

I'm sure if it turns about to be a Win7 vs Win8.1 deal then someone (thousands) will care.

Sure as shire had nothing to do with ShadowPlay or other recording software, lulz.

Lucky for us both overclockers.swe and the pclab.pl used windows 8.1.

 

[BallaTheFeared]

So what was the difference?

You seem to be willing to invest a lot more effort into this than I am, different driver? What was the issue?



Why are the results so different with the AMD vs Nvidia card?

AMD Radeon Catalyst 13.11 beta v7 is their listed driver.

 

[GaiaHunter]

So what was the difference?

You seem to be willing to invest a lot more effort into this than I am, different driver? What was the issue?

Same drivers.

It is a question of looking at the test setup of each review, hardly a big time investment.

 

[blackened23]

Blech. This is turning into the back and forth AMD fan vs NV fan garbage thread.

Personally if someone has water cooled performance data, that's cool. Specifically at what point in this thread did it turn into an epeen contest of which brand is better?

 

[BallaTheFeared]

Same drivers.

It is a question of looking at the test setup of each review, hardly a big time investment.

Far more than I'd like to give to a side topic.

Is there any reason to believe one result over the other?

Blech. This is turning into the back and forth AMD fan vs NV fan garbage thread.

Personally if someone has water cooled performance data, that's cool. Specifically at what point in this thread did it turn into an epeen contest of which brand is better?

Probably post #40 is when we got into the MHz measuring contest.


1GHz R290X is 7% faster than the 980MHz 780 @ 1080p and 9% faster at 1440p.

http://www.hardware.fr/articles/910-25/recapitulatif-performances.html

Which is where 1337 to match 1250 came from.

 

[Will Robinson]

Its hard not to smile reading this thread.
Glad to see R9 /GCN handing out a lesson in BF4.

 

[YBS1]

Fastest R9 290X (two gpu) Firestrike score posted so far on Futuremark

Too bad the user didn't fill in any details so we'd have a better idea of what to expect in terms of scaling.

 

[dragantoe]

Fastest R9 290X (two gpu) Firestrike score posted so far on Futuremark

Too bad the user didn't fill in any details so we'd have a better idea of what to expect in terms of scaling.

yeah that would be pretty disappointing if it was a 1270 mhz overclock

 

[Tempered81]

Fastest R9 290X (two gpu) Firestrike score posted so far on Futuremark

Too bad the user didn't fill in any details so we'd have a better idea of what to expect in terms of scaling.

he's posted details at ocuk:
http://forums.overclockers.co.uk/showpost.php?p=25189616&postcount=627


Edit:

1300mhz:

http://www.overclock.net/t/1436635/ocn-gk110-vs-hawaii-bench-off-thread/730#post_21101300

 

[BallaTheFeared]

1300 / 5972 @ 1.368V

Personally I'd give more weight to R290X in clock speed than that, do you guys think it might be suffering from error correction possibly?

Maybe a scaling issue in the driver, or even the 8350 limiting?

Seems low for the MHz is all.

 

[Silverforce11]

So what was the difference?

Well, you can't trust a review site which has GPUs running heaps faster in 64 MP than singleplayer, period. Its contrary to the user experience. Could just be a case of human error, they mixed up the numbers for the chart..

 

[dragantoe]

1300 / 5972 @ 1.368V

Personally I'd give more weight to R290X in clock speed than that, do you guys think it might be suffering from error correction possibly?

Maybe a scaling issue in the driver, or even the 8350 limiting?

Seems low for the MHz is all.

I didn't notice the cpu, my bet is that's the cause

 

[blackened23]

1300 / 5972 @ 1.368V

Personally I'd give more weight to R290X in clock speed than that, do you guys think it might be suffering from error correction possibly?

Maybe a scaling issue in the driver, or even the 8350 limiting?

Seems low for the MHz is all.

One, the AMD vs nvidia nonsense is really stupid. I think water cooled performance data is fine for either the 780 or 290X - if anyone wants to post that data, I think that's great. Cheerleading for one side, is really annoying - everyone that does it from either camp. There are certainly red guys doing this, so i'm not point anyone out specifically, but yes - annoying. Both sides. Two, you're comparing a custom PCB to reference. Not apples to apples - i'm not saying which is better or worse, but I think it's safe to say a reference GTX 780 PCB would not do all that great on water, while the classified PCB will. The same could easily be applicable to the 290X, and i'm sure you're aware of this.

We simply don't know yet. Then again, the 780 has custom PCBs available now while the 290X doesn't - which is really regretable because of both the cheap air cooler and because of this. Those who want the excellent custom PCB performance now have that option with the GTX while they don't with the 290X, which I am certainly aware of. But the main point here is that you can't do an apples to apples comparison of a custom PCB with 14 power phases to that of the 290X reference PCB. If you're fully aware of the limitations of GPU boost 2.0, how a reference GTX 780 PCB would do under water becomes all too clear - in other words, NOT good. The custom PCBs are way, way better for this sort of thing and that would easily be applicable to the 290X as well, but of course the GTX has this available now and 290X doesn't. Most people that water cool aren't Kingpins. They don't physically alter their GPUs to add VRMs and more power phases. That's what you do to make a reference GTX 780 good for water cooling. With the classified PCB, you don't have to.

That said, I think the classified PCB has more brute force ability than the 290X PCB does right now. So in terms of water cooling, I do think the classified will easily be able to brute force it's way to a performance lead because of the quality and features of the PCB. And like I said, GTX users have custom PCBs available now while 290X users don't - it really is too bad that AMD didn't push for aftermarket cards on day one, I feel like they should have especially with that stupid reference cooler. This can all change when the 290X has custom PCBs and what not, so the performance differences under water are fairly ambiguous at this point.

 

[Tempered81]

reference 780 PCB hit 1.55ghz on water, classy 780 it 1.7+ghz on water

aftermarket pcbs come in handy when tryin to get that last little bit out of the silicon