R290X Crossfire V NVDA's SLI 780s/Titans from [H]

[BallaTheFeared]

I did a break down of each test.

Crysis 3

20% faster than Titan SLI
27% faster than 780 SLI


Far Cry 3

7% faster than Titan SLI
15% faster than 780 SLI

Tomb Raider

19% faster than Titan SLI
32% faster tahn 780 SLI


Metro LL

13% faster than Titan SLI
19% faster tahn 780 SLI


Overall

14.75% faster than Titan SLI
23.25% faster than 780 SLI

 

[Silverforce11]

The biggest downside is the power consumption, however I bet it's not far off the FPS/w in high resolution. Heat is also pretty high, however if it was throttling or not it still easily surpasses titan/780.

It's not a downside if you take into account the amount of performance gains it is above SLI Titan or 780, overall its 780W vs 647W, delta is what, ~20% extra power used, for ~40% performance gains at 4K res, and around 20-30% at Eye-infinity. In raw terms, the CF R290X system is MORE power efficient. ;

The only downside is the noise with blowers going above 55%, that setup is just for benching, because I would personally, not be able to stand gaming next to such a loud rig. Two custom R290X would use less power (GPU running cooler, wastes less power) and be much quieter.

I don't think it was throttling since he allowed it to go above 55% fan speed if the GPU needs it. Uber mode with a 55% fan speed CAP does throttle though.

 

[Silverforce11]

I did a break down of each test.

Look at the 4k results, mind blowing stuff.

 

[selni]

It's 65 db for CF vs 52 for Titan SLI.

Is that really a big difference in actual noise?

I think 6db is (very) roughly a doubling, so yeah it's big. Aftermarket coolers will hopefully improve this a bit.

 

[OCGuy]

290x is too hot and too expensive compared to 780. While it does have legs at extreme resolutions, it is jusy much too inefficient compared to 7990 or 690.....which are much better values for consumers not gaming on 3 monitors. 1.35kw PSU???

780TI will again change the landscape, leaving 290x without a home.

The 780 and the 290 pro will be the volume sellers for the top segment, imo.


Multi card performance is more about software than hardware.

 

[3DVagabond]

It's not a downside if you take into account the amount of performance gains it is above SLI Titan or 780, overall its 780W vs 647W, delta is what, ~20% extra power used, for ~40% performance gains at 4K res, and around 20-30% at Eye-infinity. In raw terms, the CF R290X system is MORE power efficient. ;

The only downside is the noise with blowers going above 55%, that setup is just for benching, because I would personally, not be able to stand gaming next to such a loud rig. Two custom R290X would use less power (GPU running cooler, wastes less power) and be much quieter.

I don't think it was throttling since he allowed it to go above 55% fan speed if the GPU needs it. Uber mode with a 55% fan speed CAP does throttle though.

He said their tests were ran in Uber mode (That's a 55% fan cap). They recommended using 100% fan cap. It's not the same thing.

 

[3DVagabond]

290x is too hot and too expensive compared to 780. While it does have legs at extreme resolutions, it is jusy much too inefficient compared to 7990 or 690.....which are much better values for consumers not gaming on 3 monitors. 1.35kw PSU???

780TI will again change the landscape, leaving 290x without a home.

The 780 and the 290 pro will be the volume sellers for the top segment, imo.


Multi card performance is more about software than hardware.

Why are you shifting the goalposts? This review is specifically about "extreme resolutions".

What about the 1350 Maxrevo PSU? You can't possibly be questioning it quality.

 

[x3sphere]

290x is too hot and too expensive compared to 780. While it does have legs at extreme resolutions, it is jusy much too inefficient compared to 7990 or 690.....which are much better values for consumers not gaming on 3 monitors. 1.35kw PSU???

780TI will again change the landscape, leaving 290x without a home.

The 780 and the 290 pro will be the volume sellers for the top segment, imo.


Multi card performance is more about software than hardware.

Too expensive...? That's why AMD has the 290 on its way. 290X is the top part, it's not going to be cheap, although I'd argue a $50 difference is insignificant when you are already spending $500 on a graphics card.

Also, the performance at high resolutions doesn't just matter to people with 4K monitors, particularly if you are downsampling or using SSAA.

 

[wand3r3r]

290x is too hot and too expensive compared to 780. While it does have legs at extreme resolutions, it is jusy much too inefficient compared to 7990 or 690.....which are much better values for consumers not gaming on 3 monitors. 1.35kw PSU???

780TI will again change the landscape, leaving 290x without a home.

The 780 and the 290 pro will be the volume sellers for the top segment, imo.


Multi card performance is more about software than hardware.

What exactly are you trying to say, it appears to be a big deflection.

Who cares about the volume sellers, in the context of 290x vs 780/titan? How is the 690 a value card? 2x290x demolishes 780's and titan at high resolution for $1100, the 690 at $1k will be considerably weaker and is definitely not a value card.

How exactly will the 780 ti do anything that the 290x isn't doing now, supposing it could beat the 290x by 10%, isn't that what the 290x does now to the current competition? How is it different from now, wouldn't the same be said for titan/780?

1.35kw psu? It was shown to use less than 800 watts which is pretty good for the performance beating it's giving to the 780/titan.

It's more about software? So the 290x has better software since it's beating the others? What are you saying.

 

[Eymar]

I definitely think this will be true for most (single card, single screen users which should be close to everybody). For those who have 3x 120hz monitors, i definitely like what I see from 2x 290x coming from tri-sli Titans. I'm still holding judgement on switching from the Titans to 290x, until I can get 120hz set and tested (currently running eyefinity at 60hz as I need to use the DP connection for 3rd monitor, I prefer how Nvidia allows to place each video connection on each card rather than single card). Of course my DP to DVI-D active adapter is mini-DP so waiting for DP > mini DP adapter.

290x is too hot and too expensive compared to 780. While it does have legs at extreme resolutions, it is jusy much too inefficient compared to 7990 or 690.....which are much better values for consumers not gaming on 3 monitors. 1.35kw PSU???

780TI will again change the landscape, leaving 290x without a home.

The 780 and the 290 pro will be the volume sellers for the top segment, imo.


Multi card performance is more about software than hardware.

 

[5150Joker]

Look at the 4k results, mind blowing stuff.

Too bad 4K is hugely irrelevant. By the time it is meaningful 290X will be a distant memory.

 

[boxleitnerb]

So another biased review (like Kitguru and Sweclockers):

Nvidia cards are tested at default (throttling down due to the 80C temp target)
AMD cards are set from default (quiet mode) to uber mode (allowing higher clocks)

Find the error.

 

[Subyman]

Nvidia doesn't have an uber mode.

Those aftermarket heatsinks can't come soon enough. I don't think I would crossfire 290x without a custom loop. 300 W on a rear blower ouch.

 

[wand3r3r]

So another biased review (like Kitguru and Sweclockers):

Nvidia cards are tested at default (throttling down due to the 80C temp target)
AMD cards are set from default (quiet mode) to uber mode (allowing higher clocks)

Find the error.

AMD says to use uber mode for crossfire (for everyone, not just reviewers). Does NV say anything about their benchmark mode and throttling?

 

[BallaTheFeared]

Nvidia gets lambasted for Fermi, praised for Kepler... AMD gets smart offers "silent" and "uber" bios and nobody puts much stock in the fact that these are the hottest, loudest, and most power hungry cards we've ever seen while posting the results.

AMD outsmarted the reviewers imo.

 

[boxleitnerb]

Nvidia doesn't have an uber mode.

Sure they do. It's called power and temperature target. Or do you think they exist for no reason? Apparently reviewers do, since they continue to ignore them but use AMDs uber mode.

• AMDs uber mode raises the fan speed to create more thermal headroom -> the card clocks higher under sustained load
• Nvidia's temperature target allows for higher temperatures to create more thermal headroom
  -> the card clocks higher under sustained load

It is exactly the same principle. Only difference is one is a hardware switch, the other one is a (more flexible) software switch. Pick your poison

AMD says to use uber mode for crossfire (for everyone, not just reviewers). Does NV say anything about their benchmark mode and throttling?

So because "AMD says so" the testing procedure is fair? To use the possibilities of one card but not the other? Nvidia states in their reviewer guide to test in a cool environment. Obviously they know the impact of temperature on their Boost 2.0 since they were the ones who designed it. While I disagree with this request since the inside of a computer case isn't cool, it's the same as AMDs request, just in another color. Both want the impact of temperature on performance to be minimized as much as possible for benchmark numbers as high as possible.

 

[Grooveriding]

So another biased review (like Kitguru and Sweclockers):

Nvidia cards are tested at default (throttling down due to the 80C temp target)
AMD cards are set from default (quiet mode) to uber mode (allowing higher clocks)

Find the error.

Considering how badly 290x is smashing Titan/780 SLI, even giving nvidia the handicap of having overclocked cards against 290x CF isn't going to help them do much better with a few more FPS from the bonus.

 

[BallaTheFeared]

Considering how badly 290x is smashing Titan/780 SLI, even giving nvidia the handicap of having overclocked cards against 290x CF isn't going to help them do much better with a few more FPS from the bonus.

Which is weird since Linus did 1440 reference 780 vs reference R290X OC vs OC and the 780 won every test.

AMD is clearly playing the reviewers like a fiddle, I imagine the reviews read similar to the marketing material that comes with their cards.

 

[x3sphere]

Nvidia gets lambasted for Fermi, praised for Kepler... AMD gets smart offers "silent" and "uber" bios and nobody puts much stock in the fact that these are the hottest, loudest, and most power hungry cards we've ever seen while posting the results.

AMD outsmarted the reviewers imo.

Well it's easy to overlook the heat and power aspects because the cards perform good.

Fermi barely beat out the 5870 in a lot of games and it was more expensive by a huge margin.

 

[BallaTheFeared]

Well it's easy to overlook the heat and power aspects because the cards perform good.

Fermi barely beat out the 5870 in a lot of games and it was more expensive by a huge margin.

Reviewers shouldn't overlook anything, that's what they're there to provide a buffer between us and the marketing team.

 

[Eymar]

Considering how badly 290x is smashing Titan/780 SLI, even giving nvidia the handicap of having overclocked cards against 290x CF isn't going to help them do much better with a few more FPS from the bonus.

Yup, that's why I don't oc videocards much. Even where it helps most (min fps), I do not see the difference subjectively until the difference is pretty large.

 

[wand3r3r]

Which is weird since Linus did 1440 reference 780 vs reference R290X OC vs OC and the 780 won ever test.

So are you basing 290x OC performance on a single cards lottery results? :awe:

 

[BallaTheFeared]

So are you basing 290x OC performance on a single cards lottery results? :awe:

You have more information to add? I'm like a sponge I'll soak it up.


It's not so much a lottery problem as it is a cooler problem, it's already 95C at 40% and 55%.

You'd have to ask Linus if they put in ear plugs to run the card at 5500 rpms to test viable overclocks.

 

[wand3r3r]

Reviewers shouldn't overlook anything, that's what they're there to provide a buffer between us and the marketing team.

Including NV's driver problems in 4k, what about the 2D performance smoothness difference? What about the fact the cards are smoother than the titan sli, and offer noticeably more performance? Seems kind of selective in what you don't want them to overlook.

Almost every review talks about the higher power consumption, as well as the operating temperatures. It's not like they leave them out.

 

[3DVagabond]

So another biased review (like Kitguru and Sweclockers):

Nvidia cards are tested at default (throttling down due to the 80C temp target)
AMD cards are set from default (quiet mode) to uber mode (allowing higher clocks)

Find the error.

They still throttle using Uber mode. 290X's that don't throttle are coming though. I wonder how they'll perform?