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

[Will Robinson]

From the conclusions...

As per the recent NVIDIA price drop announcement GeForce GTX TITAN pricing remains the same at $999. That makes GeForce GTX TITAN SLI cost $2,000. Compare that to AMD Radeon R9 290X CrossFire, at $1,100. Now go back through this review and watch which configuration provides the highest performance and the best gameplay experience. The Radeon R9 290X CrossFire configuration owned GTX TITAN SLI. There wasn't one situation where R9 290X CrossFire wasn't faster. It didn't just ride the line with TITAN SLI, it leaped well beyond GTX TITAN SLI performance at 5760x1200 on three displays and on a 4K display.
With the price drop to $499 on the GeForce GTX 780 it is very price competitive with the AMD Radeon R9 290X at $549. When configuring SLI and CrossFire, the difference is only $100 though. GTX 780 SLI will be $1,000, while R9 290X CrossFire will be $1,100. For only $100 more, you get gigantic gains in performance from R9 290X CrossFire over GTX 780 SLI.
We saw from 20-40% performance advantages with R9 290X CrossFire over GTX 780 SLI. These are real, large increases, in performance. That extra money spent, is more than going into more performance delivered and a better gameplay experience.
We are simply impressed with the dominance that Radeon R9 290X CrossFire is showing over the GTX TITAN SLI and GTX 780 SLI

That's pretty conclusive...

http://www.hardocp.com/article/2013...0x_crossfire_video_card_review/1#.UnNGehDW5R4

 

[Silverforce11]

"So much talk about smoothness, frametime, and the actual experience of multi-GPU video cards while gaming has been discussed lately. In the past, AMD has been highly criticized of having a sub-par CrossFire experience, and rightly so. There were major issues with smoothness where games would stutter or feel choppy, even though the framerates looked good. We've been telling our readers for years that CrossFire just didn't feel as good as SLI while gaming.

Those times have changed, at least on the new Radeon R9 290/X series. The new CrossFire technology has improved upon the CrossFire experience in a vastly positive way. Playing games on the Radeon R9 290X CrossFire configuration was a smooth experience. In fact, it was smoother than SLI in some games. It was also smoother on the 4K display at 3840x2160 gaming, and it was noticeably smoother in Eyefinity at 5760x1200.

AMD is using its Frame Pacing technology, but that is a software technology. That, combined with the evolved CrossFire technology is providing a noticeable benefit to the gameplay experience compared to the previous generation cards. Once you've gamed at a large resolution on CrossFire with R9 290X, you likely won't go back.

It is safe to say, from a real-world gaming perspective, stuttering is no longer an issue one has to worry about."

+Faster
+Smoother
+More Efficient! (For real! 30-50% performance gap at high res for 100W extra system power when its already 650W...)

-Hotter
-Louder

 

[BrightCandle]

At 4k and eyefinity resolutions the 290X just stomps on the 780 and Titan, its not even funny or close. Pcper.com did some 4k testing with FCAT as well which is worth looking at:

http://www.pcper.com/reviews/Graphi...deon-R9-290X-CrossFire-and-4K-Preview-Testing

Reading through the results on HardOCP and that conclusion it strikes me as odd that they considered the 290X to be smoother. The results from pcper.com show the exact opposite, they show the 290X with quite a bit of frame variance in the FCAT results when in crossfire. There are no runt frames or other craziness like there was before but its still varying more frame to frame than the Nvidia cards at 2560x1440 and 3840x2160.

Maybe what HardOCP is seeing is the higher fps and that is mattering more at the settings they are choosing. Consistently throughout the review that they are choosing settings resulting in higher average fps on the 290X and that might well have a lot to do with their conclusion. So it may be that the extra performance is what is giving them the impression its smoother.

But its measurably the case that the 780/Titan is smoother but slower. But at these resolutions the frame rate difference is so vast it doesn't matter that the 780/Titan is smoother, the fps is the bigger problem and the 290X thus feels better despite its higher variance. That is my theory why Kyle is saying what he is anyhow.

 

[DSey]

Maybe due to a difference in driver used? PCPer used V5, [H] used V6. Dunno if that could account for that.

 

[x3sphere]

At 4k and eyefinity resolutions the 290X just stomps on the 780 and Titan, its not even funny or close. Pcper.com did some 4k testing with FCAT as well which is worth looking at:

http://www.pcper.com/reviews/Graphi...deon-R9-290X-CrossFire-and-4K-Preview-Testing

Reading through the results on HardOCP and that conclusion it strikes me as odd that they considered the 290X to be smoother. The results from pcper.com show the exact opposite, they show the 290X with quite a bit of frame variance in the FCAT results when in crossfire. There are no runt frames or other craziness like there was before but its still varying more frame to frame than the Nvidia cards at 2560x1440 and 3840x2160.

Maybe what HardOCP is seeing is the higher fps and that is mattering more at the settings they are choosing. Consistently throughout the review that they are choosing settings resulting in higher average fps on the 290X and that might well have a lot to do with their conclusion. So it may be that the extra performance is what is giving them the impression its smoother.

But its measurably the case that the 780/Titan is smoother but slower. But at these resolutions the frame rate difference is so vast it doesn't matter that the 780/Titan is smoother, the fps is the bigger problem and the 290X thus feels better despite its higher variance. That is my theory why Kyle is saying what he is anyhow.

I thought so too, but in Far Cry 3 the average frame rates are very similar and he says the CF experience is much smoother.

A very important note to make, while the performance shown there shows the framerate, it does not relate the smoothness and feel of the game. In this, AMD Radeon R9 290X CrossFire was a lot smoother in Far Cry 3. We found that with both TITAN SLI and 780 SLI there was an extreme choppiness and stutter to the game. Whereas, with R9 290X CrossFire, it was a smoother experience. So far, in each game, a pattern is forming favoring CrossFire for smoothness on R9 290X CrossFire.

 

[3DVagabond]

[H] plays games, they don't simply take measurements. I'm pretty certain it's the extra performance the 290S's offer that's making the gameplay better. As they state they've been saying all along SLI was better. They could tell by playing the games. Now they are saying it's not. They are judging exactly the same way.

 

[5150Joker]

At 4k and eyefinity resolutions the 290X just stomps on the 780 and Titan, its not even funny or close. Pcper.com did some 4k testing with FCAT as well which is worth looking at:

http://www.pcper.com/reviews/Graphi...deon-R9-290X-CrossFire-and-4K-Preview-Testing

Reading through the results on HardOCP and that conclusion it strikes me as odd that they considered the 290X to be smoother. The results from pcper.com show the exact opposite, they show the 290X with quite a bit of frame variance in the FCAT results when in crossfire. There are no runt frames or other craziness like there was before but its still varying more frame to frame than the Nvidia cards at 2560x1440 and 3840x2160.

Maybe what HardOCP is seeing is the higher fps and that is mattering more at the settings they are choosing. Consistently throughout the review that they are choosing settings resulting in higher average fps on the 290X and that might well have a lot to do with their conclusion. So it may be that the extra performance is what is giving them the impression its smoother.

But its measurably the case that the 780/Titan is smoother but slower. But at these resolutions the frame rate difference is so vast it doesn't matter that the 780/Titan is smoother, the fps is the bigger problem and the 290X thus feels better despite its higher variance. That is my theory why Kyle is saying what he is anyhow.

Agree but in addition I don't think HardOCP uses FCAT, they instead use FRAPS + FRAFS to approximate frametime measurements so in that regard, PCPer is more accurate and has better data.

Regardless, I'm happy to see AMD has improved so much, that is one area I had the most gripe with when I owned Crossfire.

 

[3DVagabond]

Agree but in addition I don't think HardOCP uses FCAT, they instead use FRAPS + FRAFS to approximate frametime measurements so in that regard, PCPer is more accurate and has better data.

Regardless, I'm happy to see AMD has improved so much, that is one area I had the most gripe with when I owned Crossfire.

They are using their eyes to determine which plays smoother.

A very important note to make, while the performance shown there shows the framerate, it does not relate the smoothness and feel of the game. In this, AMD Radeon R9 290X CrossFire was a lot smoother in Far Cry 3. We found that with both TITAN SLI and 780 SLI there was an extreme choppiness and stutter to the game. Whereas, with R9 290X CrossFire, it was a smoother experience. So far, in each game, a pattern is forming favoring CrossFire for smoothness on R9 290X CrossFire.

I've said it before when nVidia had an advantage, I want to be consistent now, FC3 is a buggy, mess of a game. It shouldn't be used for testing. Although, the last sentence speaks volumes.

 

[5150Joker]

They are using their eyes to determine which plays smoother.

Which is highly subjective depending on the person. For precise data like frametime variance, you need reliable and accurate tools. Subjective observations also has its place but is not a substitute for concrete data at the end of the day.

Also, I suspect NVIDIA will shortly close the gap at multimonitor resolutions + 4K with driver updates.

 

[BallaTheFeared]

I can't figure out if they were using Uber mode or if he uncapped the fan to allow up to 100% fan speed.

He seemed to imply he used up to 100% uncapped fan, hopefully he'll show up and let us know.

 

[BrightCandle]

Far cry 3 is kind of well known for being a poorly behaved game, its just not very smooth in general. That particular game is really the odd one out, its got lower average and lower minimum FPS, even the trace looks worse on the 290X CF and yet Kyle comments he prefers it. That is just weird.

 

[5150Joker]

I can't figure out if they were using Uber mode or if he uncapped the fan to allow up to 100% fan speed.

He seemed to imply he used up to 100% uncapped fan, hopefully he'll show up and let us know.

They ran in uber mode but also talked about 100% fan:

In fact, we personally recommend manually setting your fan to 100% capacity anyway if running reference cooler R9 290X cards in CrossFire. One thing you don't want to do is sandwich these cards together. You want a motherboard configuration that allows a gap in-between these, like ours. We have a 1.5" gap in-between the video cards, and this allows these cards to breath. All of our tests on the gameplay pages are in "Uber Mode."

How much noise do these generate at 100% fan? Must be deafening.

 

[Grooveriding]

I think it's mostly a case of with these skipped frames no longer an issue for AMD the situation is fixed. Measuring down to these very fine granular differences in ms of frametime variations is likely irrelevant to the end-user and what they can actually perceive. Now with that corrected it seems at this point frametimes/FCAT are an overly anal metric is what I am getting from [H]'s real-world gamer subjective opinion being that Crossfire is fine now and 290X CF is delivering a better experience than Titan SLI.

You can measure all kinds of things in this world and see minute differences that are not perceptible to have any actual difference.

 

[3DVagabond]

Which is highly subjective depending on the person. For precise data like frametime variance, you need reliable and accurate tools. Subjective observations also has its place but is not a substitute for concrete data at the end of the day.

Also, I suspect NVIDIA will shortly close the gap at multimonitor resolutions + 4K with driver updates.

This isn't some random forum poster. We all can choose to use whatever measurements we want to. If you would rather own a setup that [H] says offers worse gameplay because it measures better, that's fine. I'll trust Kyle's observations more though.

 

[BallaTheFeared]

Nice, this is great for multi screen users.

Personally gave up on it when I noticed all my performance was going towards rendering more walls and grass/trees that are just blurred in your vision.

Hopefully he puts out a single 1440p overclocked review soon :thumbsup:

 

[5150Joker]

This isn't some random forum poster. We all can choose to use whatever measurements we want to. If you would rather own a setup that [H] says offers worse gameplay because it measures better, that's fine. I'll trust Kyle's observations more though.

Kyle isn't infallible and his preferences and definition can be vastly different than anyone else. His or Brent's opinions aren't universal and never were, even when they thought SLI was much smoother. However, given that even PCPer's data shows the variance difference being so small, it is basically a moot point now. I just want to see AMD continue to improve upon this.

 

[wand3r3r]

I can't figure out if they were using Uber mode or if he uncapped the fan to allow up to 100% fan speed.

He seemed to imply he used up to 100% uncapped fan, hopefully he'll show up and let us know.

He said they used uber mode, however he'd recommend 100% fan even though it may not go to 100%, just to allow it if it needs to. If you read it it's pretty clear.

AMD recommends that if running CrossFire you run in the "Uber Mode." The uber mode will raise the fan capacity to 55% from 40%. If you want the most out of your CrossFire, and better heat dissipation with two cards close to each other, you want to run in Uber Mode and no less. In fact, we personally recommend manually setting your fan to 100% capacity anyway if running reference cooler R9 290X cards in CrossFire. One thing you don't want to do is sandwich these cards together. You want a motherboard configuration that allows a gap in-between these, like ours. We have a 1.5" gap in-between the video cards, and this allows these cards to breath. All of our tests on the gameplay pages are in "Uber Mode."

What I find ironic is the fact that the areas which have been criticized are all improved to best NV. The drivers were messing around on NV at 4k when setting the resolution. The smoothness (for whatever reason, the new crossfire or the higher fps?) was better on the 290x. The performance itself just stomps on the 780 (20-40% better for 10% higher cost).

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.

Now we just need the competition to resume so gamers win again. I guess complaining paid off.

 

[BallaTheFeared]

He said they used uber mode, however he'd recommend 100% fan even though it may not go to 100%, just to allow it if it needs to. If you read it it's pretty clear.

Yes yes, if you read the thread you'd know this was already answered.

 

[SolMiester]

Great news with the smooth XF, its about time, and this really shows what a great chip AMD have in Hawaii, however (as I havent read the link), i wonder how a noisy smoother gameplay provides a better experience than less FPS at a quiet level?...

What is needed is higher fan speeds on the reference cooler. We honestly recommend running nothing less than a max fan cap of 100% fan if running R9 290X CrossFire. The fan won't always run at 100%, but it will have the potential to if it needs to. If you want the best CrossFire performance, and video cards that will last longer with that much heat next to each other, go into CCC and increase the fan cap up 100% if you are using a reference cooler.

Unless you wear a headset.....but then again, the neighbours might complain.!

 

[BallaTheFeared]

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

Is that really a big difference in actual noise?

 

[SolMiester]

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

Is that really a big difference in actual noise?

Perhaps to the neighbours?

 

[Eymar]

Hardocp must run with vsync off on Crysis 3, because r9's is a stuttering mess below 60fps with vsync (ie. Fraps says 45fps, feels like 30 or lower with noticeable lag). With Vsync on and if FPS is at 60, no issue with smoothness (ie. 60fps feels like 60fps and not 45fps like it did with 7970cf). However, don't notice tearing (and I hate tearing) and with vsync off and much smoother experience across the board (ie. 30FPS and above) so my preferred setting. Night and day difference compared to 7970 CF eyefinity at launch.

 

[3DVagabond]

Hardocp must run with vsync off on Crysis 3, because r9's is a stuttering mess below 60fps with vsync (ie. Fraps says 45fps, feels like 30 or lower with noticeable lag). With Vsync on and if FPS is at 60, no issue with smoothness (ie. 60fps feels like 60fps and not 45fps like it did with 7970cf). However, don't notice tearing (and I hate tearing) and with vsync off and much smoother experience across the board (ie. 30FPS and above) so my preferred setting. Night and day difference compared to 7970 CF eyefinity at launch.

Reviewers don't test with vsync on.

 

[monstercameron]

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

Is that really a big difference in actual noise?

http://en.wikipedia.org/wiki/Decibel
sound measured in decibel is lagarithic so a jump from 50 to 60 is a huge

 

[wand3r3r]

Perhaps to the neighbours?

No need to read the review before bashing? ()

When we started up Windows on GTX TITAN SLI or GTX 780 SLI on the 4K display, it did properly boot up at the right resolution of 3840x2160. When we went to enable SLI, sometimes the screen would go black and stay black, with the display going into power save mode. It would take a reboot, and trying again, and then it would come around just fine with SLI enabled. It didn't always work the first time though. Whereas, with AMD R9 290X CrossFire we could disable and enable it at will on the 4K display, and it never got stuck. We did not have to reboot to see the changes, we could do it all on the fly. It just worked without issue.

Looks like NV has to work on their drivers.