Surround vs. Eyefinity

M

Matrices

Supreme [H]ardness
Joined
Feb 5, 2003
Messages
5,254
I don't think this is a particularly thorough review, but initial indications are that Surround handily outmatches Crossfire Eyefinity:

http://pcper.com/article.php?aid=946&type=expert&pid=4

http://pcper.com/article.php?aid=946&type=expert&pid=5


They are revealing GTX 470 performance tomorrow, but have teased that it is comparable to 5870 performance (but of course at a much lower price point).

It would be highly ironic if the one area where Fermi truly outshines Cypress is the area that ATI itself pioneered, but between superior SLI scaling and Crossfire+Eyefinity issues, I wouldn't be surprised.
 
It's not terribly surprising since Nvidia SLI has always been superior to crossfire.
 
Yes it's true that SLI scaling has always been superior to Crossfire but we are still talking about two card solutions whereas the other one will work with single cards.

But look at the results a bit more closely. 5870 numbers in L4D are almost the same as 5870 CF numbers! Also, Metro 2033 will of course be skewed toward Nvidia.

Nvidia wins again for best dual card setup and also fastest single card setup...but ATI comes out on top for single card multiple monitor performance.
 
That is impressive multiscreen performance.:eek:

I mean, I really like my EyeFinity, but if nvidia kicks that much ass in three monitor gaming, wow.
Just wow.
 
I'm wondering if by outputting the monitor on the second card they are saving time and bandwidth vs ATI transferring frames to the primary card.
 
The 5870's barely beat the 5970, that's strange. WSGF benched the 5970 vs. 5870 xfire before in eyefinity and it was very different from that.
 
Meh. Just doesn't seem right, not trying to be a fan boy at all.

Ill check back tomorrow, and compare their 470 vs 5850 results, with my personal 5850's.
 
Waiting for 470 vs 5970 results. I do not have the cojones for 480 SLI.
 
I agree that the L4D2 results make no sense - 5870 CF is no faster than 5870 there. In other games, CF is faster, but the scaling is still crap. I don't have other reviews to compare with though - is CF scaling in Eyefinity worse than CF scaling in regular mode?
 
Matrices said:
I agree that the L4D2 results make no sense - 5870 CF is no faster than 5870 there. In other games, CF is faster, but the scaling is still crap. I don't have other reviews to compare with though - is CF scaling in Eyefinity worse than CF scaling in regular mode?
Click to expand...

I will confirm from my own setup that scaling in eyefinity is generally worse than scaling with one monitor. It may or may not be true with gtx 480's as well - it's just that SLI scaling is better than xfire anyway. I may get my roommate to do some scaling comparisons with single vs. triple monitors with his SLI GTX 480 setup.

Regardless of scaling though, the 5870's should handedly beat the 5970 and not just barely edge it out.
 
Interesting results, but I think it might be a bit too late to matter. By now people have already bought what they want, and I don't see to many people jumping from Eyefinity to Surround Vision given the costs involved with that at this point. I'd like some more performance, but the way I see it my card is now, what, 9 months old, give or take? I'm just going to wait for Southern Islands, with 2nd gen Eyefinity - hopefully with improved Eyefinity CF support.
 
piscian18 said:
major grain of salt.

I'm not jumping until I see hard results.
Click to expand...

Don't think you need that much salt, clearly nVidia has a competitive product and while I don't have an Eyefinity setup to compare it to this setup is working MUCH better than I expected. The performance is amazing with these 3 480s, the only thing that bugging me right now is Dirt 2, I'm having issues cranking stuff up and maintaing a solid frame rate, had to drop down to medium which makes no sense. The benchmark FILES through at 90 FPS at 2xAA and everything on max. Not sure what's up there but these are beta drivers so hopefully that gets clear up. But it looks like most of the game I'm playing right now work and that's awesome.
 
looks like I don't need to look back to ATI for now, going to grab another 480 soon once I move in to my new apartment :D
 
heatlesssun said:
Don't think you need that much salt, clearly nVidia has a competitive product and while I don't have an Eyefinity setup to compare it to this setup is working MUCH better than I expected. The performance is amazing with these 3 480s, the only thing that bugging me right now is Dirt 2, I'm having issues cranking stuff up and maintaing a solid frame rate, had to drop down to medium which makes no sense. The benchmark FILES through at 90 FPS at 2xAA and everything on max. Not sure what's up there but these are beta drivers so hopefully that gets clear up. But it looks like most of the game I'm playing right now work and that's awesome.
Click to expand...

There are some seriously questionable performance numbers there. I have a hard time believing them until Hard does an overview on it , then I'll buy it. Testing methods are wildly different site to site.
 
Some of you guys are over thinking this:

http://hardocp.com/article/2010/06/29/nvidia_3d_vision_surround_experience/5
This is expensive technology to put into your home. That may or may not hold value for you, but make no mistake about it...NV Surround gaming with GTX 480 SLI will give you the best gaming experience you have ever had.
Click to expand...

And Kyle gave it a gold star so whatever numbers are to follow he just gave you the Cliffs Notes. Surround is the deal and [H] gave it its highest award.

As for the PC Perspective numbers, actually they make perfect sense, at least for Batman AA which I just played though pretty extensively. The numbers for AA come from the in game benchmark. I got an average of 136 using exactly the same settings which would seem to be low considering I’m running 3x SLI and the review is only using 2x SLI. However I noticed that the PCP reviews tops out at 142 FPS which is EXACTLY the same max I’m getting, this benchmark is apparently capped at 142 FPS. However when playing the game, there’s apparently a 62 FPS cap, the frame rate NEVER breaks beyond this.

All I can say is that Surround rocks, a few glitches here and there but overall the performance I’m getting with my sig rig at 5760x1200 overall is simply amazing and the bulk of games I’ve tried are working great.

Well done nVidia.
 
heatlesssun said:
Some of you guys are over thinking this:

http://hardocp.com/article/2010/06/29/nvidia_3d_vision_surround_experience/5


And Kyle gave it a gold star so whatever numbers are to follow he just gave you the Cliffs Notes. Surround is the deal and [H] gave it its highest award.

As for the PC Perspective numbers, actually they make perfect sense, at least for Batman AA which I just played though pretty extensively. The numbers for AA come from the in game benchmark. I got an average of 136 using exactly the same settings which would seem to be low considering I’m running 3x SLI and the review is only using 2x SLI. However I noticed that the PCP reviews tops out at 142 FPS which is EXACTLY the same max I’m getting, this benchmark is apparently capped at 142 FPS. However when playing the game, there’s apparently a 62 FPS cap, the frame rate NEVER breaks beyond this.

All I can say is that Surround rocks, a few glitches here and there but overall the performance I’m getting with my sig rig at 5760x1200 overall is simply amazing and the bulk of games I’ve tried are working great.

Well done nVidia.
Click to expand...

Was your 142 max fps done in surround at 5760x1200. I haven't done much testing but I saw a brick wall around 125 fps. I think perhaps that we might be hitting a limitation in the bandwidth needed to transfer frames from one card to another.
 
Kudos to Nvidia for finally delivering on the promised surround features. Looks like it was worth the wait. 480 SLI surround simply kicks ass performance-wise!
 
thanks ATI for limping me into multi montor gaming... I'll let Nvidia take it from here... ;) pretty much.. just picked up a pair of 470s.. :)

PS: my only real gripe with eyefinity now is the issues with crossfire and then also with crossfire and vsync.. which have never been sorted out properly after 7+ driver updates or something.. I'm really wanting more gpu power for my 5760x1080 gaming but not willing to sacrifice crossfire headaches and vsync for it.. looks like Nvidia might have finally come through here.. but I have enjoyed eyefinity a lot during the time I have been using it.. just wish it was easier to get more gpu power without a lot of headaches.
 
Last edited:
Those numbers seem like nonsense to me. From my Dirt2 experience with Crossfire and Eyefinity, there is almost perfect 100% improvement with Crossfire. Their figures suggest it barely gets 20% increase from adding a second 5870.

They say 40fps with a single 5870, which should be ~75-80fps in Crossfire, putting it right up there with a 480GTX SLI.

Edit: And then L4D2 as well. They suggest Crossfire is slower than non Crossfire. Sounds like they don't know what they are doing tbh. I'll wait for the H testing.
 
Finkus said:
Those numbers seem like nonsense to me. From my Dirt2 experience with Crossfire and Eyefinity, there is almost perfect 100% improvement with Crossfire. Their figures suggest it barely gets 20% increase from adding a second 5870.

They say 40fps with a single 5870, which should be ~75-80fps in Crossfire, putting it right up there with a 480GTX SLI.

Edit: And then L4D2 as well. They suggest Crossfire is slower than non Crossfire. Sounds like they don't know what they are doing tbh. I'll wait for the H testing.
Click to expand...

Does anyone have any Eyefinity and CF numbers else where to compare? Should be easy enough to point out the issues. As far as the Surround numbers they make sense to me in my testing of Batman AA and Dirt 2.
 
Finkus said:
Those numbers seem like nonsense to me. From my Dirt2 experience with Crossfire and Eyefinity, there is almost perfect 100% improvement with Crossfire. Their figures suggest it barely gets 20% increase from adding a second 5870.

They say 40fps with a single 5870, which should be ~75-80fps in Crossfire, putting it right up there with a 480GTX SLI.

Edit: And then L4D2 as well. They suggest Crossfire is slower than non Crossfire. Sounds like they don't know what they are doing tbh. I'll wait for the H testing.
Click to expand...

I don't know about that. We're talking about performance across 3 monitors so the gains would not be as great when you have to split performance 3 ways. Also, I am not sure about how strong crossfire performance is across 3 displays. For one display, sure it should be strong, but with 3?
 
vjcsmoke said:
I don't know about that. We're talking about performance across 3 monitors so the gains would not be as great when you have to split performance 3 ways. Also, I am not sure about how strong crossfire performance is across 3 displays. For one display, sure it should be strong, but with 3?
Click to expand...

Eh I wouldn't say it's quite that simple as dividing by 3. You aren't splitting performance amongst 3 monitors, you're rendering an image 3 times as big. Even if you hit memory bandwidth caps because of the large image, 2 cards will have a theoretical 2x performance boost over 1 card, because you are rendering twice as often. The complication arises in whether there's enough Crossfire/PCI-e bandwidth available for the cards to not get hung up.

The difference between 1920x1200 and 5760x1200 may be dramatically worse due to memory bandwidth constraints (it's not btw - the 3x larger resolution gets about 2/3 the fps not 1/3), but the difference between 1 card and 2 cards should aways have a max of double the performance.


The pcper review isn't *too* far out of line with this:
http://www.widescreengamingforum.com/wiki/ATI_Radeon_5870_Eyefinity6_-_Featured_Review_-_Page_6
But WSGF did have a bit better results. If you sub HL2 for L4D2, Source engine scaled pretty bad in either review when going Crossfire Eyefinity. But you also have to realize the PCper article is comparing 2 $500 cards ($1000) to 2 $380 cards ($760) when looking at the larger fps gap.


For what it's worth nVidia does have a possible technical advantage which is, when sending display output between the 2 cards, nVidia can send a fraction of the frame every frame instead of ATI's method of sending the entire frame every other frame.
 
Very interesting, I am saving up to build a Eyefinity setup but after seeing this maybe it's time to wait to see how things turn out. Maybe with performance like this, it's what Nvidia needed to turn things around.
 
noquarter said:
But WSGF did have a bit better results. If you sub HL2 for L4D2, Source engine scaled pretty bad in either review when going Crossfire Eyefinity. But you also have to realize the PCper article is comparing 2 $500 cards ($1000) to 2 $380 cards ($760) when looking at the larger fps gap.
Click to expand...

Actually that is an E6 model with 2gb ram. That's a $500 card x 2, not to mention $100 adapter or DP monitor. Considering 2 470's can beat a 5870 xfire in Eyefinity, I think the budget minded surround gamer will be going with a pair of 470's or lower.

Now as to single monitor gaming, the 5850 and 5870 have better price points and nearly the same performance. This round single monitor gaming goes to ATI, but it looks like Nvidia is going to steal ATI's eyefinity thunder due to driver issues. Nvidia simply has better drivers when it comes to multi-gpu multi-monitor gaming.
 
I don't know that it's a function of drivers. I think it is naive to blame drivers when there is a systematic problem with performance. You can't write code around the realities of hardware. It may be that there is a structural hardware issue (bandwidth) that affects Eyefinity Crossfire performance.

The real problem is that we have 50 reviews, one for each "new" brand of card with its 5mhz overclock boost, but no useful interviews with ATI or Nvidia engineers to discuss the specifics of their engineering or driver development process. I'd love to see a roundtable discussion about the constraints and successes of each company's attempt to incorporate new features.
 
I always got excellent scaling with crossfire with my 4850's. Always from 50-100% with most games closer to 100% than 50. Don't see how the 5xxx series could be worse than the 4xxx series was. Also don't see how triple monitors could impact the scaling efficiency like that either unless there is just a glaring eyefinity/crossfire driver issue.
 
Well, there obviously is a glaring eyefinity/crossfire issue, though again, it may not be due to drivers but due to architecture. Remember that when Eyefinity first came out, where was no Crossfire support at all.
 
Naieve said:
Considering 2 470's can beat a 5870 xfire in Eyefinity, I think the budget minded surround gamer will be going with a pair of 470's or lower.
Click to expand...

No, a budget minded surround gamer will be going with a single 5850 :p

Now as to single monitor gaming, the 5850 and 5870 have better price points and nearly the same performance. This round single monitor gaming goes to ATI, but it looks like Nvidia is going to steal ATI's eyefinity thunder due to driver issues. Nvidia simply has better drivers when it comes to multi-gpu multi-monitor gaming.
Click to expand...

I don't think it has anything to do with drivers. Eyefinity was unknown to all except the few who were working on it before launch - meaning most of the card was designed with no clue that it was going to be trying to render an image higher than 2560x1600. I wouldn't be surprised if something like the crossfire link is the bottleneck. Which, if it is, will hopefully be fixed in Southern Islands *crosses fingers* ;)
 
Matrices said:
The real problem is that we have 50 reviews, one for each "new" brand of card with its 5mhz overclock boost, but no useful interviews with ATI or Nvidia engineers to discuss the specifics of their engineering or driver development process. I'd love to see a roundtable discussion about the constraints and successes of each company's attempt to incorporate new features.
Click to expand...

I don't know if you've seen them, but Anandtech has 2 really good reads about the 4xxx and 5xxx series (including the story of how Eyefinity came to be):

4xxx series: http://www.anandtech.com/show/2679/1
5xxx series: http://www.anandtech.com/show/2937/1
 
kllrnohj said:
I don't think it has anything to do with drivers. Eyefinity was unknown to all except the few who were working on it before launch - meaning most of the card was designed with no clue that it was going to be trying to render an image higher than 2560x1600.
Click to expand...

Way more planning goes into things than that. The only reason Nvidia is able to add this so easily is its already been a feature in their Quadro line.
 
bexamous said:
Way more planning goes into things than that. The only reason Nvidia is able to add this so easily is its already been a feature in their Quadro line.
Click to expand...

Not for Eyefinity. Here, read about how Eyefinity came to be, and how few people knew about it: http://www.anandtech.com/show/2937/10

Also, I wouldn't say Nvidia has been able to add it "easily" - it took them how many months and missed how many promised dates? First it was supposed to come at launch, then in april, etc...
 
Finkus said:
Those numbers seem like nonsense to me. From my Dirt2 experience with Crossfire and Eyefinity, there is almost perfect 100% improvement with Crossfire. Their figures suggest it barely gets 20% increase from adding a second 5870.

They say 40fps with a single 5870, which should be ~75-80fps in Crossfire, putting it right up there with a 480GTX SLI.

Edit: And then L4D2 as well. They suggest Crossfire is slower than non Crossfire. Sounds like they don't know what they are doing tbh. I'll wait for the H testing.
Click to expand...

I'll back this up. Dirt 2 for me is probably the game that scales the best with eyefinity and xfire. Also, L4D2 does not show negative scaling in any way with my setup. Doesn't really make sense that theirs did.
 
Last edited:
kllrnohj said:
I don't know if you've seen them, but Anandtech has 2 really good reads about the 4xxx and 5xxx series (including the story of how Eyefinity came to be):

4xxx series: http://www.anandtech.com/show/2679/1
5xxx series: http://www.anandtech.com/show/2937/1
Click to expand...

Great articles there particularly the one on the R870. I have to say that I have been extremely impressed by the ATI 5000 series from a hardware point of view. Nvidia did take the single gpu performance crown back with the GTX 480, but at what price? It is a compromise of the product that they actually wanted to release, probably due to the numerous process issues they had with TMSC and 40nm yields.

ATI will almost certainly take back the performance crown when southern isles releases. But what's really crucial about the 5000 series, besides taking the performance title for 6 months, is that it gave ATI back their credibility and has given them market penetration in the mid to low range markets with DX11 capable cards whereas Nvidia hasn't even released anything to compete in those segments yet.

Plus judging from die sizes and defect rates alone, ATI is likely making significantly more profit per wafer than Nvidia. And that money can be cycled back in to develop newer and better gpus. The competition drives prices down, and consumers win. We've already seen some price drops on the 480 and 470. The biggest new ATI feature - Eyefinity, forced Nvidia to make available a comparable NV Surround feature for consumers that they had previously limited only to their Quadro line of professional cards. More gameplay enhancing features and options is a big win for us consumers.
 
kllrnohj said:
Not for Eyefinity. Here, read about how Eyefinity came to be, and how few people knew about it: http://www.anandtech.com/show/2937/10

Also, I wouldn't say Nvidia has been able to add it "easily" - it took them how many months and missed how many promised dates? First it was supposed to come at launch, then in april, etc...
Click to expand...

Wow I find that pretty impressive. Seems like gpu has some wacky display engine to just be able to get 3 active displays working, you would think it would have been trimmed down if only 2 active displays were required. I've never really looked at ATI cards, can you get 3 active displays with different mode timings (not in eyfinity)? Anyways, yeah pretty impressive performance doesn't blow.
 
bexamous said:
Wow I find that pretty impressive. Seems like gpu has some wacky display engine to just be able to get 3 active displays working, you would think it would have been trimmed down if only 2 active displays were required. I've never really looked at ATI cards, can you get 3 active displays with different mode timings (not in eyfinity)? Anyways, yeah pretty impressive performance doesn't blow.
Click to expand...

The entire Evergreen lineupt actually has 6 display outputs. 6 fully independent display outputs. 2 are capable of a DVI or HDMI signal, and the other 4 are DisplayPort only. Eyefinity runs on top of the 6 display outputs to make them appear as a single display to the OS.
 
kllrnohj said:
No, a budget minded surround gamer will be going with a single 5850 :p
Click to expand...

I agree, but I was talking about a relatively powerful setup. A single 5800 series card might be okay if you are running at 1680x1050x3, but if you are running 1080p+x3 you will be seriously scaling back on everything. Good multi-monitor gaming requires more than even a single 480 could supply. A pair of 5850's would be nice if they scaled like they do in single monitor gaming, but they don't.


kllrnohj said:
I don't think it has anything to do with drivers. Eyefinity was unknown to all except the few who were working on it before launch - meaning most of the card was designed with no clue that it was going to be trying to render an image higher than 2560x1600. I wouldn't be surprised if something like the crossfire link is the bottleneck. Which, if it is, will hopefully be fixed in Southern Islands *crosses fingers* ;)
Click to expand...

Yeah, but their drivers have fixed the scaling issues in some games. So it is at least partially a driver issue. As to the xfire bridge, I'm pretty sure they do in fact carry less than the sli bridge, and that forced ATI onto the pci-e bus for some cross card communication, which is that architecture issue you are talking about.
 
You must log in or register to reply here.
Back
Top