Why does 290/x best 780ti in multicard setups?

[Rvenger]

...

 

[GaiaHunter]

Single monitor performance doesn't translate to multi monitor directly does it? I heard there can be scaling issues that disappear when you run single monitor.

I'm under the impression that currently 4K monitors have a pair of display controllers and are seen as 2 monitors (4K/2 each).

 

[OCGuy]

I love to see these type of benchmarks, but they are really not useful for the majority of people.

Even people who could spend the money on that monitor setup choose not to. I know one of those people very well.

 

[tg2708]

I love to see these type of benchmarks, but they are really not useful for the majority of people.

Even people who could spend the money on that monitor setup choose not to. I know one of those people very well.

But that can also be said for anything, not everyone that has the money to burn will go all out but some will. The richer people in some cases are the most price conscious.

 

[JDG1980]

Someone has already posted in another thread: it's driver issues. NVIDIA is having a problem with SLI performance at 4k, has acknowledged it and is working on a fix. It has nothing to do with VRAM. It's also largely pointless because no one is gaming in 4k and no content has been created to natively run in 4k.

We're talking about expensive, cutting-edge video cards. It hardly seems surprising that users might want to pair them with expensive, cutting-edge monitors.

If this were a discussion of mid-range (~$200) GPUs, I would agree that 4K is irrelevant at this point in time. But at $1000-$1400 for a SLI pair of 780/780Ti, you should expect support for high-end monitors, not to mention some level of future-proofing.

 

[OCGuy]

But that can also be said for anything, not everyone that has the money to burn will go all out but some will. The richer people in some cases are the most price conscious.

Except look how deep the niche already is with these cards, without even getting to the resolutions:

- People who buy discreet cards

- People who buy discreet cards that are $500+

- People who buy discreet cards that are $500+ and use those resolutions


I am not discouraging publishing the info, I am just saying that it isn't very applicable to how the card performs for the majority of the niche market that buys it. Although at this point one could say that buying 2X 290x or 780ti cards to run @ 1080p means you don't read reviews at all 😛

 

[EliteRetard]

Edit:

I am not trashing any person or product, those of you who are need to chill.

The primary point of topic wasn't really even about Nvidia, but the great job AMD has done with these new cards especially in regard to CF. It appears Nvidia is having a driver problem right now. That is what allowed the AMD cards to stand out, it's not a physical advantage for AMD.

The main theme of my first post was the good performance and scaling of the new 290/x AMD cards, a pair of them can run almost any game at any resolution with good detail and FPS.

Because of AMD's lower prices you can also acquire a pair of them for around the same price as one of Nvidias fastest card. Based solely on price I wanted to compare an AMD CF setup and a single Nvidia card. Not much different than a discussion between 7970 CF and a single Titan a month or two ago.

Except now the major CF issues are basically solved and the performance is higher.
I don't see a downside to a 290/x CF setup...even for the games that don't multi-card.

 

[GaiaHunter]

Except look how deep the niche already is with these cards, without even getting to the resolutions:

- People who buy discreet cards

- People who buy discreet cards that are $500+

- People who buy discreet cards that are $500+ and use those resolutions


I am not discouraging publishing the info, I am just saying that it isn't very applicable to how the card performs for the majority of the niche market that buys it. Although at this point one could say that buying 2X 290x or 780ti cards to run @ 1080p means you don't read reviews at all 😛

But pretty much every user that has 4K for gaming will be using $500+ worth of GPUs, probably 2.

 

[x3sphere]

Except now the CF issues are basically solved and the performance is higher.
I don't see a single downside to a 290/x CF setup...even for the games that don't multi-card.

Still no DX9 frame pacing support even on the 290 cards. Probably not an issue for most but makes me glad I went SLI as I am replaying The Witcher 2 right now and running with Ubersampling is incredibly demanding still.

 

[cmdrdredd]

Still no DX9 frame pacing support even on the 290 cards. Probably not an issue for most but makes me glad I went SLI as I am replaying The Witcher 2 right now and running with Ubersampling is incredibly demanding still.

I always thought Ubersampling was a bit of a joke. You're right though, if you want to run it...very demanding.

 

[guskline]

Have there been any full water block tests of R9 290's?

 

[YBS1]

I understand that at the higher resolutions there may be a bug hindering nVidia's results in SLI, however even at the lower resolution the Crossfire 290X seems to be beastly, winning as often as the 780TI does. Even when losing it's not by much. Granted it's usually in Uber mode, but still. I don't think this fight is really going to be settled until we see both cards in their top shelf forms, the Lightnings, the Classifieds, the SOC, HOF, etc. I'm anxious to see just how long the 290X's legs are when paired with a more robust pcb and not crippled by that pos cooler AMD choose to go with, at the very least it will eliminate the biggest complaint everyone has about the cards, the noise. Bring on the customs!

 

[EliteRetard]

Again, all you people need to chill out. This isn't an argument over the 290x switch. They tested and listed both. I don't really see there being an issue there. I do think in a case such as this that they do NEED to list both and they did.

The real issue seems to stem from some confusion over Nvidias numbers. It was not made clear that they are having problems and it made the AMD cards look unusually good. Now that I (and everybody else in this thread) know that Nvidia may fix the issue and perform better eventually there's nothing to fight about.

Fact: Right now CF with the new 290/x looks better than the 780ti SLI numbers (at least in "4k"). That may change in the future.

What I think is interesting and I'd like to steer this thread towards is the great performance/scaling of the 290/x regardless of any issues Nvidia may or may not have.

We don't have exact numbers for vanilla 290 CF but I posted a theoretical comparison in my first post. If the 290 is 5% slower than the 290x non Uber then I can use those CF numbers and approximate. With most of the major CF issues fixed, and the high speed of a single 290 there doesn't seem to be a major downside to going 290 CF over a single 780ti and you can potentially get very large performance gains.

 

[Subyman]

Again, all you people need to chill out. This isn't an argument over the 290x switch. They tested and listed both. I don't really see there being an issue there. I do think in a case such as this that they do NEED to list both and they did.

The real issue seems to stem from some confusion over Nvidias numbers. It was not made clear that they are having problems and it made the AMD cards look unusually good. Now that I (and everybody else in this thread) know that Nvidia may fix the issue and perform better eventually there's nothing to fight about.

Fact: Right now CF with the new 290/x looks better than the 780ti SLI numbers (at least in "4k"). That may change in the future.

What I think is interesting and I'd like to steer this thread towards is the great performance/scaling of the 290/x regardless of any issues Nvidia may or may not have.

We don't have exact numbers for vanilla 290 CF but I posted a theoretical comparison in my first post. If the 290 is 5% slower than the 290x non Uber then I can use those CF numbers and approximate. With most of the major CF issues fixed, and the high speed of a single 290 there doesn't seem to be a major downside to going 290 CF over a single 780ti and you can potentially get very large performance gains.

What's different though is that Nvidia's problems are driver level and, as of right now, there is absolutely no solution to fix it.

 

[EliteRetard]

What's different though is that Nvidia's problems are driver level and, as of right now, there is absolutely no solution to fix it.

I don't know what to say about that.
It is an issue, and apparently what caused me to make this thread.
Probably should have been discussed/made clear in the review.

 

[cmdrdredd]

You are insulting buyers of the high end cards.
If somebody knows how to check FPS it's most likely that person would know where is the switch.

You think everyone who has the kind of money actually does research? This forum and others like represent the minority.

 

[cmdrdredd]

I don't know what to say about that.
It is an issue, and apparently what caused me to make this thread.
Probably should have been discussed/made clear in the review.

How can they make clear what is obviously speculation anyway? There is no way to know what the actual problem is. And there is indeed a solution to a problem that is driver related...updated drivers with a fix. Right now we don't know for certain it is drivers unless nvidia has said it was and I missed it.

 

[OCGuy]

Multi-GPU is about software, not hardware. 780ti is clearly the fastest single GPU on the planet, and SLI profiles will catch up.

 

[iiiankiii]

Multi-GPU is about software, not hardware. 780ti is clearly the fastest single GPU on the planet, and SLI profiles will catch up.

Traditionally, SLI seems to scale up a bit worst than Crossfire. Therefore, it's no surprise a 290x Crossfire is faster than a gtx 780 ti SLI eventhough a single GTX 780 TI is a bit faster than the 290x.

 

[OCGuy]

Traditionally, SLI seems to scale up a bit worst than Crossfire. Therefore, it's no surprise a 290x Crossfire is faster than a gtx 780 ti SLI eventhough a single GTX 780 TI is a bit faster than the 290x.

All the reviews I am seeing show the 290x Xfire only overtaking 780ti at 3XXX resolution....

 

[iiiankiii]

All the reviews I am seeing show the 290x Xfire only overtaking 780ti at 3XXX resolution....

You're right. AT's review is showing those cards trading blows at 1440p in mutli card configuration eventhough the 780 ti is faster than a 290x in single card configuration. Looks like Crossfire just scales up better.

 

[amenx]

As with almost all new card releases, SLI and CF performance is *usually* lacking out of the box. It improves with new driver releases. Only this time AMD did their homework and seems to have got it right from the beginning. Nvidia will almost certainly catch up in the weeks ahead to fix this. Meanwhile, the 99% of users who do not have 4k resolutions will not lose any sleep over it.

 

[Silverforce11]

As with almost all new card releases, SLI and CF performance is *usually* lacking out of the box. It improves with new driver releases. Only this time AMD did their homework and seems to have got it right from the beginning. Nvidia will almost certainly catch up in the weeks ahead to fix this. Meanwhile, the 99% of users who do not have 4k resolutions will not lose any sleep over it.

Just as 99% of gamers who aren't going to be buying 2x $700 GPUs... no need to worry about it...

 

[mak360]

Just as 99% of gamers who aren't going to be buying 2x $700 GPUs... no need to worry about it...

LOL Good one

Yeah, ever since Ryan (Maybe Anandtech) got back from Hawaii - been acting strange.

Just my observation

Edit: I kind of miss the Frame Latency threads.

AMD mentioned that their XFire scaling will reach 95% - I have no Idea how Nvidia will deal with that.

 

[boxleitnerb]

Simple explanation:
Possible throttling on the NV cards.

If someone with a 290 or 290X CF setup wants to compare multi-GPU performance with a Titan SLI setup, I'd be willing to run some benchmarks. The scaling that the AT review shows is definitely not normal.