PekkerCadette
n00b
- Joined
- Aug 2, 2003
- Messages
- 34
Can anyone inform me if nVidias Surround requires the three monitors to run the same resolution or if I can keep my 20"+30"+20"? I'm hoping for the latter...
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
NVIDIA Surround Technology Performance Review
- Thread starter FrgMstr
- Start date
GoodBoy
2[H]4U
- Joined
- Nov 29, 2004
- Messages
- 3,143
I'm almost positive that the resolution and refresh rate capabilty needs to be identical across all three.
Using different sizes is probably possible so long as the above conditions are met, but wouldn't be as good of an experience imho.
That's already been covered in the thread. Though the adapter doesn't necessarily need to cost that much. I've got one that was $10.00. Unfortunately its only passive and doesn't work with resolutions over 1920x1080. If you shop around you can get the active DVI-DP adapters for considerably less than $100.00.
Yes, you can apparently get VGA adapters that are $20. And also yes, they apparently only work up to a certain resolution (1080p or 19x12?). But that doesn't obviate the issue some people have with Eyefinity regardless of their adapter - flickering on the Display Port connected-monitor.
To be honest I was expecting a bigger difference between the GTX480 SLI and 5870 CF, given a single 480 is more powerful to begin with and people say how much more awesome nVidia muli-GPU scaling is, I was expecting a pretty big difference. Not sure what sort of deals you guys can get where you are, but out here a 480 is around 28% more expensive than a 5870.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
http://hardocp.com/article/2010/07/20/nvidia_surround_technology_performance_review/9
You're joking right? I mean as HUGE a fan Kyle is of Eyefinity, or maybe was might be more correct now, for him to say Surround "completely owned" CF Eyefinity, what else do you want, cake?
EVERY site that's done this review has come up with CF EF getting its ass literally handed to it. It's the biggest blowout in between two first gen products (400s and 5000s are first gen DX 11) parts in years.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
I might try to get some done over the next week, I've not seen much floating around the web. But if you want the readers digest condensed version. hella fast, getting substaintly better numbers in all the games Kyle tested that I've played just from my eyeing framerates. I'm do a quick AvP bench and post numbers to give you a feel, think I'll play this a bit latter.
From other posters it would seem that tri-sli 2d surround sees *zero* gains over dual-sli.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Not true. Here's the thing. What I have noticed is that there is a 142FPS maximum frame rate cap in 2D Surround with 2 and 3 cards so if your game is running hitting that maximum frame rate, which even two cards can do in a lot of games the third doesn't boost average frame rates. Now if your game is running below that rate then the 3rd card seems to scale well. Let me play a little AvP and I'll demostrate. Be right back.
Nuby1Canuby
Gawd
- Joined
- Mar 4, 2008
- Messages
- 754
Anyone with CrossFireX 5970s care to post some benchies at 5760X1200? Running the core at 850 and the memory at 4800 would be appreciated, since that's the stock speed for the 5870 chips it contains.
Its not passive, if it was it wouldnt work at all.
ALL DP-VGA adapters are active, as its converting the signal. It just doesnt need the power from the USB cable because its an analog signal that doesnt need as much.
Sounds more like a CPU bottle neck than a frame cap. Those are pretty insane numbers in either case.
If it's always 142 FPS regardless of the game, then more likely it is a bandwidth bottleneck between the cards. PCI-E and the SLI bridge only have so much bandwidth, after all
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
If it's always 142 FPS regardless of the game, then more likely it is a bandwidth bottleneck between the cards. PCI-E and the SLI bridge only have so much bandwidth, after all
Oh well, I lied. Took a look at couple of things and with 3x SLI I was getting the same numbers as Kyle, no scaling on the third card. But in 3D still in Batman, 100%.
If it's a bottleneck or bandwidth issue it shouldn't scale at at. Need to dig into this when I have time.
You should go back to the review and look at the settings Kyle used. He went for the highest possible settings it was playable at. The GTX 480 is beating the 5870 while using more complex graphics settings. Even more amazing, consider those minimum frame rates, and imagine them with equal settings.
"If it's always 142 FPS regardless of the game, then more likely it is a bandwidth bottleneck between the cards. PCI-E and the SLI bridge only have so much bandwidth, after all "
I started mentioning this the first day the surround drivers were released. I am pretty sure there is a bandwidth limitation, thankfully it is rather high. I noticed it would hit a wall anywhere from 110 to 140 or so fps. I think it fluctuates in line with how much information the card needs to pull in, thus eating up that cross card bandwidth, and only leaving so much for rendered frames.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
But as I said I see scaling in 3D at least in Batman AA at 100% so a hardware limitation doesn't make sense if even one game especially in 3D since that's FAR more demanding than 2D.
I was wondering about that too, and figured it was rendering extra frames but unable to find bandwidth to pass them.
I know for a fact on COD4 when the gpus are maxed I should be getting higher FPS. There is a bottleneck somewhere.
If it's 142 in any game, then it sounds like a bandwidth issue. Not really an important one because even if you've running 120hz monitor you're still beyond the refresh rates.
Action_Parsnip
Limp Gawd
- Joined
- Jul 22, 2010
- Messages
- 169
Fascinating article.
To Kyle & the team: Perhaps rather than CF having poor scaling, is the case rather that the 5870s have serious issues with AA at triple monitor resolutions?
Taking a look at the Crysis Warhead and Metro 2033 apples-to-apples results, the numbers seem to reflect the delta between single card 5870/GTX 480 performance.
Games where having AA enabled is NOT a no-go from the start, i.e. all other titles tested, seem to show ok/decent performance without AA and a humungous crash in fps or input lag/lagging with AA turned on. In your opinion is there a limit with the software? or are the cards just not up to AA, with details up at very high resolutions? Did you get any insights from ATI as a result of this article? I ask because quite often they seem keen to give feedback to you, which you find out from trawling the comment threads.
I did think when reading through this "sideport, put your boots on mate!" but instead I would think that theres issues compositing very large frames in the displaying card, or by dint of the hardware design there's alot of data syncing between the two which the nVidia cards do not do/still do but handle it better.
You had the hardware on test, what do you think about it.
( huge props for the site btw, been visitng more or less everyday since the 3870 review )
To Kyle & the team: Perhaps rather than CF having poor scaling, is the case rather that the 5870s have serious issues with AA at triple monitor resolutions?
Taking a look at the Crysis Warhead and Metro 2033 apples-to-apples results, the numbers seem to reflect the delta between single card 5870/GTX 480 performance.
Games where having AA enabled is NOT a no-go from the start, i.e. all other titles tested, seem to show ok/decent performance without AA and a humungous crash in fps or input lag/lagging with AA turned on. In your opinion is there a limit with the software? or are the cards just not up to AA, with details up at very high resolutions? Did you get any insights from ATI as a result of this article? I ask because quite often they seem keen to give feedback to you, which you find out from trawling the comment threads.
I did think when reading through this "sideport, put your boots on mate!" but instead I would think that theres issues compositing very large frames in the displaying card, or by dint of the hardware design there's alot of data syncing between the two which the nVidia cards do not do/still do but handle it better.
You had the hardware on test, what do you think about it.
( huge props for the site btw, been visitng more or less everyday since the 3870 review
)The problem is that it dips below 120 on my 5760x1200 setup. I bet on 1920x1080 it is perfect for 120hz gaming. Not a big problem for me though, if it ever is I'll just pick up 3 1080p monitors and sell my sammies.
Fascinating article.
To Kyle & the team: Perhaps rather than CF having poor scaling, is the case rather that the 5870s have serious issues with AA at triple monitor resolutions?
Taking a look at the Crysis Warhead and Metro 2033 apples-to-apples results, the numbers seem to reflect the delta between single card 5870/GTX 480 performance.
Games where having AA enabled is NOT a no-go from the start, i.e. all other titles tested, seem to show ok/decent performance without AA and a humungous crash in fps or input lag/lagging with AA turned on. In your opinion is there a limit with the software? or are the cards just not up to AA, with details up at very high resolutions? Did you get any insights from ATI as a result of this article? I ask because quite often they seem keen to give feedback to you, which you find out from trawling the comment threads.
I'd love to see a company build a monitor with 2 24" 1080p monitors as one single panel. It would be almost like eyefinity, and give you that extra bit of peripheral vision you want, all while being able to be run off any graphics card and use the regular sli/crossfire profiles, which would mean great scaling.
I did think when reading through this "sideport, put your boots on mate!" but instead I would think that theres issues compositing very large frames in the displaying card, or by dint of the hardware design there's alot of data syncing between the two which the nVidia cards do not do/still do but handle it better.
You had the hardware on test, what do you think about it.
( huge props for the site btw, been visitng more or less everyday since the 3870 review
The problem isn't AA. The problem is that CF/EF drivers are hit and miss. They obviously use a vastly different crossfire profile to handle passing the information between gpu's, and they simply haven't finished the drivers.
I waited and waited on them to do that, bought an i7 setup to get rid of my old sli board so that I could go eyefinity. Yet I knew I was going to need multi-gpu's to get the graphics quality I wanted, so I waited and watched as driver after driver was released with only minimal work done on EF/CF. That's when I decided to wait for the 480's. Mutli monitor gaming is driver heavy, and Nvidia has the lead in drivers there still, though ATI caught up in single monitor gaming.
You do realize that on some games people report their second card at 0 % usage? That means the drivers are just ignoring the extra card. That means without a doubt much of it is drivers. There are bandwidth limits they have to deal with, the big problem is complete lack of support though.
Face it. Crossfire Eyefinity is a small part of their customer base, they are putting most of their driver programmers to work on single monitor setups. Which is how they finally caught up with SLI in single monitor, well got close enough anyways.
Also, while they have to make these drivers for the first time, Nvidia already has Mosaic which gave them a head start, regardless of the release date to consumer graphics.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Whatever the issue is AMD has some work to do. It could be that AMD might come out with a single GPU with a good lead over the 480 and still be behind or neck and neck with the 480 in multi-monitor multi-GPU setups.
http://hardocp.com/article/2010/07/20/nvidia_surround_technology_performance_review/9
You're joking right? I mean as HUGE a fan Kyle is of Eyefinity, or maybe was might be more correct now, for him to say Surround "completely owned" CF Eyefinity, what else do you want, cake?
EVERY site that's done this review has come up with CF EF getting its ass literally handed to it. It's the biggest blowout in between two first gen products (400s and 5000s are first gen DX 11) parts in years.
Try controlling the hot steam coming from your arse, the only "blow-out" there is the Aliens vs. Predator, the ati solution dropped 100% when enabling 2x AA, which I believe it is more about bugs than anything, all other "apples to apples" are within the margin like how the single cards perform, which is about 20%.
If this article shows anyting new, it shows how unreliable the "min fps" is, just look at the bottom of the Dirt 2 page, a dip to low fps in 2 seconds does not affect the overall smoothness of the gameplay.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
http://hardocp.com/article/2010/07/20/nvidia_surround_technology_performance_review/9
Dude, take it up with Kyle.
Out here is Australia
And yes I did look at the numbers, for the most part they were running the same settings but with AA or HBAO enabled, except on Metro 2033 and Crysis. The last 3 games tested only had the 480s 14-24% faster, which is about how much faster a single 480 is over a 5870 anyway. AVP looks more like a glitch, and it would have been interesting to see the apples-to-apples of the 1st two games going down to HBAO off/normal quality settings rather than forcing the 5870s to work above what they are happy to do (not that I'm suggesting the review isn't thorough enough, just saying it doesn't paint the whole picture).Obviously the 480s are faster, a single 480 is faster too, with how much people praise nVidia SLI scaling I was expecting it to better than simply being able to enable a small amount of AA and HBAO.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Check out other review sites like PC Perspective that do the more traditional apples to apples testing and you get pretty much the same story. The difference isn't a little, it's a lot.
AA is also something that makes a big difference in IQ, especially comparing no AA to any level of AA. And while HBAO isn't my favorite visual feature, it's definitely a resource hog, so it's nothing small if one card can run it well where another card can't.
IMO it depends on the game. Some games look fine without AA at all, some look better with a bit of AA, some dont look good until you have massive amounts of AA. I even have a friend who doesn't like AA at all because it takes away the "sharp" look to edges.
Action_Parsnip
Limp Gawd
- Joined
- Jul 22, 2010
- Messages
- 169
I have to run Battlefield 2 without any AA due to a punkbuster black screen issue, at least for my favourite server. Can't say its detracted from my playing. Mind you 8xAA can smooth things out a little too much. I find it hard to fault 2xAA really for pretty much anything.
Are you kidding?
Metro 2033, nearly the same fps with the 480's on HIGH global and the 5870's on NORMAL. If you've played this game you know that's quite a difference.
BFBC2, 480's with HBAO ON are faster than 5870's with it OFF, and you can deny the importance of minimum frames all you want, but look at the chart for this game and the 5870's drop much lower for much more of the test.
AvP, destroyed all the way around...nuff said.
Splinter Cell. 480's faster with 4xAA than 5870's with none. That's a beatdown however you look at it. And this quote doesn't help, "It just simply felt smoother and better with GTX 480 SLI, and even with just 2X AA it was too jittery and sluggish with Eyefinity6 CrossFireX. "
Dirt 2. 480's running faster once again with 8xMSAA and TR MSAA versus just 2x AD MSAA...another beatdown.
Crysis Warhead, not even close. Much higher graphical settings on the 480 and it still pounds the 5870 setup.
Apples to apples means jack when one of the setups still has enough juice to increase settings, add AA, etc. Using those as your argument is a Pyrrhic victory at best. Run the 5870 setups at the same settings at the 480's, which would also be apples to apples, and it would be an even bigger blowout.
GoodBoy
2[H]4U
- Joined
- Nov 29, 2004
- Messages
- 3,143
You must not be very familiar with ATI and their hit and miss driver history, which I had the unfortunate opportunity to go thru myself in the 9700 Pro days with about 7 months worth of issues beginning at the time that card was released (August 2002) before it seemed to stabilize (by which point I was fed up anyway), and since 2006 or so thru today with radeon onboard video since this is what all the PCs contain at work.
I've got to be careful about which drivers I use on the PCs I support, I almost never use the monthly catalyst releases since for one, they are rarely WHQL, and two, they tend to lack stability. Mind you not every month. There might be 3 to 6 months in a row where they work fine, followed by 3 to 6 months of unstable crap. This isn't even when gaming either, this is desktop users working in access, excel, and web apps. Don't even get me started on Catalyst Control Center... I can load up 20 new identical PCs with identical software images, and there will be a couple that for no apparent reason have constant "Catalyst Control Center has encountered a problem and needs to be closed.." when booted. I just uninstall it. Usually I don't install it at all but after 6 months I keep giving it another try, hoping the issues will be fixed, only to see the same damn things happen yet again. We don't have to have CCC, but it is helpful because at least 600 of my 800 users are running dual LCD's, and another 100 of them run triple LCD's, and the CCC sometimes helps order them and set the primary screen better. On some ATI cards, using the normal windows screen to set the default screen doesn't hold, the ati driver over-rides it.. grr. So you HAVE to use CCC in those cases. (Some users want the middle to be primary, some want the far left LCD to be primary).
I think the march catalyst (10.3) has been working ok for a few months. Still using it.
Now, for gamers, using an older driver to hold onto stability is a no-win situation because often you need the newest driver to have the latest optimizations in 3d games. This is where ATI has really been a big let-down. They are very often late to the game with the driver optimizations that improve performance, makes CF work, etc in games. If it takes ATI 6 months (if ever) to finally improve performance in a game, its too damn late because I've (and likely 80% of gamers who will ever play it) already finished it 3 months ago. This is where nVIDIA has to be complemented and their "Way Its Meant To Be Played" program. They spend the time and money to work with game developers and ensure that the needed optimizations are in place when a game launches (or very soon after release) so we can get the best performance right away. ATI rarely does this. And the monthly driver releases are hit and miss on top of that. Not saying nVIDIA has never had a driver problem (there was a fan issue with the 196 in certain situations) but their driver issues have been far less common. ATI apparently spends far less on driver support than nVIDIA does.
My speculation: A game/engine being made in support of one GPU does not mean it could not also fully support the other, i.e. being in TBIMTBP does not exclude ATI from also doing the same thing for that game. The Game dev would support the best performance of a title whether it's played using nVIDIA GPU or an ATI GPU. Games can have multiple rendering pipelines. So I really doubt being in one program means it will not play well on the other companies' GPU. (Maybe someone who is a game dev and has dealt with these programs directly can clarify this) Pointing this out before some jackhole complains about TWIMTBP is why game X performs worse on ATI GPU's (like somehow nVIDIA has sabotaged ATI performance with their work on said game), when in reality it is ATI's own fault. Also, nNVIDIA often makes additional performance tweaks down the road, ATI less so (quoting from memory an article [H] did in the last few months regarding performance of Catalyst and the changes over an 18 month or so period).
Action_Parsnip
Limp Gawd
- Joined
- Jul 22, 2010
- Messages
- 169
You dont seem to be working from the assumption that one 480 is substantially better than one 5870
Action_Parsnip
Limp Gawd
- Joined
- Jul 22, 2010
- Messages
- 169
Ive only had one issue with this 4870 in my rig and that was crashing with fallout 3, almost gone with the next catalyst, gone for good with the one after that. That was on xp but im on win 7 64 and it hasnt put a foot wrong. Had the card since launch. Not taking anything away from your experience just saying some must get unlucky.
I'd argue the performance not coming along in time point, for one nvidias perf. increasing alot over time vs. ati's may also mean timid launch drivers from the N. If you like apples to apples benchies look at techgage.com's review of 4870 vs. gtx 260 performance cat 8.10 vs. 8.11. The [H] article on 5870 performance with old and new drivers argueably found that the card was running into the limits of having 1gb of ram, and for Crysis warhead its such an old title now with so much public profile it would have been optimised up the ass in the 4000 series drivers long before the cypress launch anyways.
Ive seen performance come forward with my 4870 from when games are launched for the next few driver releases, plus rendering errors getting pulled out. Dead Space and Fallout 3 come to mind and Left 4 Dead 2 was edging forward right up to Cat.10.4 in my eyes. Sure its not heavy on more or less any gpu but if you like to run 'light' games under an underclocked profile in say Afterburner like me I noticed I could take the core and mem down a little bit more seemingly each month. So now 1400x900 8xAA is very playable at 575mhz core 675mhz mem.