Rumor: Radeon R9-290X-X2 to Bring Double the Hawaii?

[mapesdhs]

Please just say it has 4GB of visible memory, not 8GB total and then have
to caveat it by saying 4GB per GPU. Best to avoid confusion in the first
place. If it's not 8GB per GPU then don't even think of mentioning it as being
in any way an 8GB card. Tired of seeing manufacturers get away with this
kind of PR spin in GPU sales; toms should not encourage it.

Ian.

 

[childofthekorn]

now because these are pretty much the remake of tahiti chips wont the 290 x2 just be a glorified 7990?

False. the 280/280x are a rebranding of Tahiti. 290 is the new BVolcanic Islands architecture.

 

[quilciri]

Crossfire has better frame times than SLI in Bioshock Infinite:

Source: PC Perspective (and that's just the variance graph, the raw frame times are even more in favor of Crossfire).

The 95th percentile is not all that matters. The whole spectrum of percentiles matter.

Tom's Hardware disagrees with your conclusions. In Bioshock Infinite, for example, they point out that there's a difference, but not a noticeable one.

As for the serious issues Crossfire had in the past, or currently has on other cards, that's irrelevant to the 290, 290X, and the hypothetical 2x290X.

Your chart here conflicts with the review you originally linked. In your first, SLI's variance is better than AMD's in Bioshock.... At best you have conflicting evidence here, and It's certainly not enough to claim that AMD's frame pacing is better than Nvidia's in that game.
http://www.tomshardware.com/reviews/radeon-r9-290x-hawaii-review,3650-24.html

You're rehashing what I already covered. The 2.64ms variance that amd had @ the 95th in bioshock is acceptable...but you again are failing to address that they had 2, 3, and 4 times that in over half the games tested in that review....which is not acceptable.

In my earlier post...

There is 5 millisecond 95th percentile variance in Tomb raider, almost 9ms in Metro, 10ms in Crysis 3 and 11ms in Skyrim, the last of which is DX9 and unsupported by xfire (still!). 11ms is heavy microstutter. 5 is noticeable. that's a problem in 4 of the 7 the games tested.

Tomshardware called the 2.64ms variance in Bioshock infinite "not problematic, but notable". I can agree with that (though Nvidia's variance was less than 1ms there)

I'm a bit baffled at your insistence that the 95th percentile rendering is not a big deal. Maybe watching a ripple go down your screen 1 1/2 times per second doesn't bother you. Congratulations, you aren't bothered by microstutter.However, it's irresponsible to say that the frame variance problem @the 95th is not a big deal because there are lots of people (including me) who notice and are bothered by it.

 

[quilciri]

Crossfire has better frame times than SLI in Bioshock Infinite:

Source: PC Perspective (and that's just the variance graph, the raw frame times are even more in favor of Crossfire).

The 95th percentile is not all that matters. The whole spectrum of percentiles matter.
.

Also, your own argument is conflicting with you new image (which again, also conflicts with tom's review in the link you originally used). If you're saying the whole spectrum matters, xfire's frame variance is worse than SLI's all the way up to the 90th percentile.

 

[quilciri]

Call me crazy, but if I'm dropping over $1000 on a graphics solution, I'm not settling for "there are a few issues left". Even if they do eventually iron out the rest of the frame pacing issues on the 290's, the implicit F.U. to anyone who bought previous xfire and have been waiting patiently will lose them more customers than just me.

Even besides the frame pacing problem....the 290x already has high heat/poor reference cooling, which makes this proposed card even less of a good idea. I lol'd at the post saying "the first card with a reference waterblock"

 

[Novuake]

Call me crazy, but if I'm dropping over $1000 on a graphics solution, I'm not settling for "there are a few issues left".

Even besides the frame pacing problem....the 290x already has high heat/poor reference cooling, which makes this proposed card even less of a good idea. I lol'd at the post saying "the first card with a reference waterblock"

Oh, hi there Crazy!

 

[Sakkura]

Your chart here conflicts with the review you originally linked. In your first, SLI's variance is better than AMD's in Bioshock.... At best you have conflicting evidence here, and It's certainly not enough to claim that AMD's frame pacing is better than Nvidia's in that game.
http://www.tomshardware.com/reviews/radeon-r9-290x-hawaii-review,3650-24.html

You're rehashing what I already covered. The 2.64ms variance that amd had @ the 95th in bioshock is acceptable...but you again are failing to address that they had 2, 3, and 4 times that in over half the games tested in that review....which is not acceptable.




I'm a bit baffled at your insistence that the 95th percentile rendering is not a big deal. Maybe watching a ripple go down your screen 1 1/2 times per second doesn't bother you. Congratulations, you aren't bothered by microstutter.However, it's irresponsible to say that the frame variance problem @the 95th is not a big deal because there are lots of people (including me) who notice and are bothered by it.

No, the chart does NOT conflict with the review I posted. It just gives a more complete picture.

And it's you that claimed Crossfire had problems in Bioshock Infinite, when both reviews clearly demonstrates that it does not.

The 95th percentile frame variance is not crucial. The variance across all the percentiles is MUCH more important. The 95th percentile is just a single vertical slice of that graph from PC Perspective. And notice how SLI runs into worse frame latency in Bioshock Infinite at the higher percentiles than Crossfire does. That's what demonstrates that Crossfire performs better than SLI in that game. It's not a 2ms variance across half the frames that matters, it's the 5+ ms variance that in this case shows up more often with SLI than with Crossfire.

 

[Sakkura]

Even besides the frame pacing problem....the 290x already has high heat/poor reference cooling, which makes this proposed card even less of a good idea. I lol'd at the post saying "the first card with a reference waterblock"

That's a completely separate issue. And a very serious one.

 

[quilciri]

No, the chart does NOT conflict with the review I posted. It just gives a more complete picture.

You said

The 95th percentile is not all that matters. The whole spectrum of percentiles matter.

and the frame time variance for xfire is worse than SLI all the way up to the 92nd percentile in the chart you posted...the rest of that "whole spectrum" as you put it. That's a pretty clear conflict.

and that chart's evidence also conflicts with tom's review which shows better frame time variance for SLI in bioshock across the board

It may be that xfire has better frame variance in this game on a single 2560 monitor, while SLI has better pacing on tri-2560's....but who's gonna buy a pair of 290x's to play on a single monitor? (well, maybe a 4k )

Assuming there hasn't been a testing error, kudos to AMD for competing on a single game, but the extreme weight of evidence still lies heavily in Nvidia's favor.

 

[quilciri]

The point is your "not a big deal" is not everyone's standard. I notice a 1.5 times per second (or more in some of the games!) ripple and it bothers me.

Many people are sensitive to microstutter, and claiming it's not a big deal as a blanket statement is a poor evaluation.

 

[Sakkura]

and the frame time variance for xfire is worse than SLI all the way up to the 92nd percentile in the chart you posted...the rest of that "whole spectrum" as you put it. That's a pretty clear conflict.

It stays below 2ms all the way up past the 90th percentile, and SLI hits 2ms about as often as Crossfire. Above 2ms happens more often with SLI than with Crossfire.

Given that it takes at least around 5ms for it to be noticeable at all, the higher variance at the lower percentiles is not relevant to the user experience. What is relevant is how often you get noticeable spikes, and as the graph shows, they are more common with SLI than Crossfire.

 

[Sakkura]

The point is your "not a big deal" is not everyone's standard. I notice a 1.5 times per second (or more in some of the games!) ripple and it bothers me.

Many people are sensitive to microstutter, and claiming it's not a big deal as a blanket statement is a poor evaluation.

Yeah, and those ripples are more common with SLI than Crossfire in some games, like Bioshock Infinite.

I have not claimed that microstutter is not a big deal.

 

[quilciri]

Yeah, and those ripples are more common with SLI than Crossfire in some games, like Bioshock Infinite.
.

Incorrect. There was considerable microstutter in bioshock with my 7850's, among may other games, even with the 13.11 beta. I do not notice any with 670's.

 

[quilciri]

I have not claimed that microstutter is not a big deal.

That is in reference to the 290's. Their best case scenario latency in the 4 games over 5 ms is similar to FCAT for my 7850's in games where the ripple still bothered me.

That is what you claimed was not a big deal and what I refute. If that does not affect you then you are not as sensitive to microstutter. In that vein, you should not be applying a blanket statement that the latency on the 290's is not a big deal.

 

[Sakkura]

Incorrect. There was considerable microstutter in bioshock with my 7850's, among may other games, even with the 13.11 beta. I do not notice any with 670's.

Your 7850s are irrelevant since they use a different Crossfire implementation.

 

[Sakkura]

That is in reference to the 290's. Their best case scenario latency in the 4 games over 5 ms is similar to FCAT for my 7850's in games where the ripple still bothered me.

That is what you claimed was not a big deal and what I refute. If that does not affect you then you are not as sensitive to microstutter. In that vein, you should not be applying a blanket statement that the latency on the 290's is not a big deal.

It's not about best case scenarios. It's about how often it becomes noticeable to a human, and how often it gets how bad. And the 290X Crossfire beats the 780 SLI on both counts in Bioshock Infinite. It loses big in a game like Skyrim, but you don't need Crossfire/SLI for that anyway.

 

[quilciri]

Incorrect. There was considerable microstutter in bioshock with my 7850's, among may other games, even with the 13.11 beta. I do not notice any with 670's.

Your 7850s are irrelevant since they use a different Crossfire implementation.

The same FCAT performance = the same frame latency. Can we agree there?

If you have the same FCAT performance, you have the same amount of microstutter; doesn't matter how the multi-card is implemented.

Note that it also doesn't even matter if it's the same game. At a given percentile and fps, 5ms latency in battlefield and 5ms latency in bioshock translate to the same amount of microstutter in each game.

5ms frame latency on the 7850's is the same as 5ms frame latency on the 290's at the same percentile and fps. That makes them highly relevant.

 

[quilciri]

It's not about best case scenarios. It's about how often it becomes noticeable to a human, and how often it gets how bad.

How noticeable it is to a human is what I've been saying the entire time. Just because you are fine with a certain level of microstutter doesn't mean others are.

5ms frame latency on the 7850's bothers me....therefore 5ms frame latency on the 290's will as well.

 

[quilciri]

And the 290X Crossfire beats the 780 SLI on both counts in Bioshock Infinite. It loses big in a game like Skyrim, but you don't need Crossfire/SLI for that anyway.

You don't have consistent evidence to support The Bioshock statement. The tom's hardware review clearly shows the 780 SLI with better latency throughout. Your second chart from PC perspective shows the 290's with better latency after the 92nd, and worse the rest of the time.

The 290's were worse in all the games tom's hardware tested. The skyrim performance isn't an isolated incident;' two of the DX11 games had nearly as poor performance as the "unsupported" Skyrim. There are also some dx9 games that can use the sli/xfire horsepower, such as final fantasy realm reborn (though no blame to AMD for that at the moment - Squenix hasn't put xfire support in yet)

 

[Deleted member 217926]

Sakkura and quilciri you guys are having the longest, most polite argument ever! Keep it up!

 

[Novuake]

Sakkura and quilciri you guys are having the longest, most polite argument ever! Keep it up!

I noticed the same thing. Its actually constructive!

 

[Deleted member 217926]

I know right? I usually don't even keep up with news threads. This one is interesting though!

And how did you quote me and end up with rostrow416 as the person you quoted?

 

[de5_Roy]

I know right? I usually don't even keep up with news threads. This one is interesting though!

And how did you quote me and end up with rostrow416 as the person you quoted?

[strike]by mistaking[/strike] try changing/missing some of the digits from the quotemsg tag, tons of [strike]Not[/strike] fun for the kids!
for example:

Spoiler

I know right? I usually don't even keep up with news threads. This one is interesting though!

And how did you quote me and end up with rostrow416 as the person you quoted?

 

[quilciri]

Sakkura and quilciri you guys are having the longest, most polite argument ever! Keep it up!

I noticed the same thing. Its actually constructive!

Whoops. Sorry. I'll make it flaming....err I mean flame it up...umm how do you verb that?

 

[Sakkura]

The same FCAT performance = the same frame latency. Can we agree there?

If you have the same FCAT performance, you have the same amount of microstutter; doesn't matter how the multi-card is implemented.

Note that it also doesn't even matter if it's the same game. At a given percentile and fps, 5ms latency in battlefield and 5ms latency in bioshock translate to the same amount of microstutter in each game.

5ms frame latency on the 7850's is the same as 5ms frame latency on the 290's at the same percentile and fps. That makes them highly relevant.

The 7850s implement Crossfire differently from the 290s and will perform entirely differently. They are therefore completely irrelevant.