nvidia surround?

NoQuarter over at Anand has a good hypothosis on what the delay/problem is. If this is the case, we may be looking at a hardware/bandwidth limitation that will prevent 3D Surround from ever performing at higher resolutions...

My guess was something to do with SLI bandwidth and sending the frame buffer alternating directions each frame. SLI normally functions in AFR by sending the frame from the slave card to the master card for the frame the slave card is responsible for.

Supposedly the SLI link is capable of 1GB/s. Nvidia already stated (according to EVGA forums) that its Surround implementation would be limited to 5760x1080 max (ie can't use 1920x1200 monitors). The reason this would be is because the bandwidth required to send that frame buffer is too much.

ATI already said the Crossfire link is barely enough to handle Eyefinity resolutions and it's likely some frames get sent via PCI-e bus when the Crossfire link is saturated causing the stuttering a few users have reported when running Crossfire Eyefinity. The Crossfire link is capable of 0.9GB/s but you can double them up between your 2 cards, providing 1.8GB/s.

If the SLI link is capable of 1GB/s and a 5760x1080 resolution is 6220800 pixels, and each pixel is 4 bytes of data, the frame buffer would be 24883200 bytes or 0.023174286 Gigabytes. If this is all true, the max frame rate transfer between cards would be 43 frames per second not including necessary overhead but assuming it has to send the full frame.

In Eyefinity's implementation the frame buffer only needs to be sent across the Crossfire link every other frame (when the slave card is on duty). In Surround's implementation the frame buffer or at least a portion of it needs to be sent across the SLI link every frame (from master to slave, then slave to master).

If the drivers are capable of it it could be possible to send 2/3 of the frame buffer from slave to master for one frame, then 1/3 of the frame buffer from master to slave the next frame, cutting the bandwidth required for frame copies in half (sort of).

The SLI bridge apparently was designed to be uni-directional also, which may cause some latency issues waiting for the bridge to clear up to send data the other way.

Also using SFR rendering method gets terribly complicated as both cards would be rendering parts of the frame the other card needs.
 
harmattan said:
NoQuarter over at Anand has a good hypothosis on what the delay/problem is. If this is the case, we may be looking at a hardware/bandwidth limitation that will prevent 3D Surround from ever performing at higher resolutions...

My guess was something to do with SLI bandwidth and sending the frame buffer alternating directions each frame. SLI normally functions in AFR by sending the frame from the slave card to the master card for the frame the slave card is responsible for.

Supposedly the SLI link is capable of 1GB/s. Nvidia already stated (according to EVGA forums) that its Surround implementation would be limited to 5760x1080 max (ie can't use 1920x1200 monitors). The reason this would be is because the bandwidth required to send that frame buffer is too much.

ATI already said the Crossfire link is barely enough to handle Eyefinity resolutions and it's likely some frames get sent via PCI-e bus when the Crossfire link is saturated causing the stuttering a few users have reported when running Crossfire Eyefinity. The Crossfire link is capable of 0.9GB/s but you can double them up between your 2 cards, providing 1.8GB/s.

If the SLI link is capable of 1GB/s and a 5760x1080 resolution is 6220800 pixels, and each pixel is 4 bytes of data, the frame buffer would be 24883200 bytes or 0.023174286 Gigabytes. If this is all true, the max frame rate transfer between cards would be 43 frames per second not including necessary overhead but assuming it has to send the full frame.

In Eyefinity's implementation the frame buffer only needs to be sent across the Crossfire link every other frame (when the slave card is on duty). In Surround's implementation the frame buffer or at least a portion of it needs to be sent across the SLI link every frame (from master to slave, then slave to master).

If the drivers are capable of it it could be possible to send 2/3 of the frame buffer from slave to master for one frame, then 1/3 of the frame buffer from master to slave the next frame, cutting the bandwidth required for frame copies in half (sort of).

The SLI bridge apparently was designed to be uni-directional also, which may cause some latency issues waiting for the bridge to clear up to send data the other way.

Also using SFR rendering method gets terribly complicated as both cards would be rendering parts of the frame the other card needs.
Click to expand...

Then why not released it as is for 5760x1080? That is the maximum most people are running. Or are they simply not getting decent FPS with that resolution over SLI in the first place? Perhaps they are hesistant to release surround because of a performance issue? If the max framerate transfer is 43 FPS + overhead, maybe they can't even get 30 FPS going which would make them look bad.
 
Spare-Flair said:
Then why not release it as is for 5760x1080? Or are they simply not getting decent FPS with that resolution over SLI in the first place? Perhaps they are hesistant to release surround because of a performance issue? If the max framerate transfer is 43 FPS + overhead, maybe they can't even get 30 FPS going which would make them look bad.
Click to expand...

Yes, that's what he's supposing: 5760x1080 is the absolute highest they can push the bandwidth with the SLI bridge limitation. Not sure what would happen if the resolution was set higher, but can only presume it would either be severly FPS limited or would just not work at all. Either way, people with 1920x1200 monitors or higher look to be screwed.

I just did a relook at all the nV demos of 3D Surround and, low and behold, they largest monitors they use in all of them are 1080p (or 5760x1080 for three). This would suggest they do not yet have a workable solution for over that res.
 
Spare-Flair said:
Then why not released it as is for 5760x1080? That is the maximum most people are running. Or are they simply not getting decent FPS with that resolution over SLI in the first place? Perhaps they are hesistant to release surround because of a performance issue? If the max framerate transfer is 43 FPS + overhead, maybe they can't even get 30 FPS going which would make them look bad.
Click to expand...

I think thats a typo. 5760 is 1920x3. I assume he means 1680x3 is the max at 5040. Why nvidia wouldnt just say "ok but you can only use 22"'s" I don't know. Unless you're specifically nvidia is saying no because they can only support standard 1080p? WTF would Nvidia care? 1080p is well standard
 
piscian18 said:
I think thats a typo. 5760 is 1920x3. I he means 1680x3 is the max at 5040. Why nvidia wouldnt just say "ok but you can only use 22"'s" I don't know. Unless you're specifically nvidia say no because they can only support standard 1080p? WTF would Nvidia care? 1080p is well standard
Click to expand...

5760x1080 = triple 1920x1080 monitors (1080p)

5040x1050 = triple 1680x1050 monitors

I don't think that it's a typo. Triple 1080p is a perfectly respectable thing to release Surround under. It's not a limitation that would severely turn off the majority of customers. 5040x1050, if it were true as the real limit would be problematic.
 
Last I checked the GTX cards have two SLI connectors. So unless there is a hardware reason why they cannot use both to double the bandwidth...
 
Spare-Flair said:
5760x1080 = triple 1920x1080 monitors (1080p)

5040x1050 = triple 1680x1050 monitors

I don't think that it's a typo. Triple 1080p is a perfectly respectable thing to release Surround under. It's not a limitation that would severely turn off the majority of customers. 5040x1050, if it were true as the real limit would be problematic.
Click to expand...

Yep, that math is correct:1080p is the supposed limitation.

1080p respectable for multi-surround gaming? Sure, if you want to use TN panel LCDs. A major turn-off however for enthusiasts and anyone who already has 1920x1200 monitors or above (myself included).
 
harmattan said:
Yep, that math is correct:1080p is the supposed limitation.

1080p respectable for multi-surround gaming? Sure, if you want to use TN panel LCDs. A major turn-off however for enthusiasts and anyone who already has 1920x1200 monitors or above (myself included).
Click to expand...

Well enthusiasm is one thing and yes I've got 3 HP LP2465's so that includes me but technically TN 1080p panels are standard gaming monitors. When I say standard I mean more TN's are sold than anything else. I think even the Eyefinity review monitors at HARDOCP are TN's. With Nvidias track record I'd imagine 1080p would be good enough for them to get out the door.

I think we're looking at driver bugs. Jumping from 197 to 256? big hmm.
 
There is a huge difference between a bug and a fundamental limitation in architecture. If they're crippled by bandwidth issues and unable to produce more than 43 FPS on three-monitor gaming, this thing is dead on arrival.
 
night_2004 said:
Last I checked the GTX cards have two SLI connectors. So unless there is a hardware reason why they cannot use both to double the bandwidth...
Click to expand...

From what I understand that is only for tri sli and you get no benefit from using two for two cards as only one is used I think.
 
necrophile said:
From what I understand that is only for tri sli and you get no benefit from using two for two cards as only one is used I think.
Click to expand...
currently that is the case

I suppose the question is whether the 2nd link can be used to increase the bandwidth between two cards. That would require somebody to start making all new SLI connectors though.

I suppose this lack-of-bandwidth theory could explain why only dual SLI is (in theory) going to be supported in the beginning. If they're struggling to find the bandwidth for dual sli, tri-sli or quad-sli is certainly going to be a problem.
 
piscian18 said:
http://hardocp.com/article/2010/03/24/ati_radeon_5870_2gb_3x2_eyefinity_gaming_experience

p2210H TN's 2:45min

Id have to go back and watch the reviews to see what those other monitors are.
Click to expand...

Those 22"s are only used for a six-screen setup -- which I suspect is only due to cost constraints.

All the triple-monitor tests were run on Dell 3007s. Unless I'm mistaken, the vaporware that 3D Surround is turning out to be only can support three monitors (due to the lack of DP and, likely, also a bandwidth limitation).
 
harmattan said:
Those 22"s are only used for a six-screen setup -- which I suspect is only due to cost constraints.

All the triple-monitor tests were run on Dell 3007s. Unless I'm mistaken, the vaporware that 3D Surround is turning out to be only can support three monitors (due to the lack of DP and, likely, also a bandwidth limitation).
Click to expand...

If the bandwidth limitation speculation is correct, there might even be more overhead, since in the final stage, the image needs to be synced with one output on second card (most likely over PCI-E) that can cause additional lag. With all the crosstalk between SLI and over PCI-E bus, while keeping everything synced without one screen lagging behind, it might provide some challenge for the driver team.

How on earth will they then manage 120hz over 3 screens and each screen needs to be synced with shutterglasses? :eek:
 
Tamlin_WSGF said:
How on earth will they then manage 120hz over 3 screens and each screen needs to be synced with shutterglasses? :eek:
Click to expand...

The simple answer is they probably can't. Not on PCI-E 2.1 or the current SLI bridge. I think this entire 3D Surround debacle was a major case of nvidia putting their cart before the horse.
 
Hamish Campbell said:
The simple answer is they probably can't. Not on PCI-E 2.1 or the current SLI bridge. I think this entire 3D Surround debacle was a major case of nvidia putting their cart before the horse.
Click to expand...

Its starting to sound more and more like vaporware. :mad:

At least I am happy that Nvidia chose to pretend they had some multi-mon solution, instead of the usual downplay. This way, the developers have incentive to make solutions for a potential crowd of Nvidia and ATI multi-mon capable users.

@Kyle Bennet:
Is Nvidia still ignoring you on this?
 
piscian18 said:
I think even the Eyefinity review monitors at HARDOCP are TN's.
Click to expand...

That's incorrect. Kyle uses 3x Dell U2410s which are 1920x1200 IPS panels with a 'gaming' mode for low input lag.

With Nvidias track record I'd imagine 1080p would be good enough for them to get out the door.

I think we're looking at driver bugs. Jumping from 197 to 256? big hmm.
Click to expand...

If Nvidia can get 1920x1080 3D Surround out the door, that's a good 'first effort' and a step in the right direction. But I don't think they can just afford to abandon 1920x1200 and 2560x1600 users, because those people are most likely to shell out the big bucks for 470 or 480 SLI configurations.

night_2004 said:
Last I checked the GTX cards have two SLI connectors. So unless there is a hardware reason why they cannot use both to double the bandwidth...
Click to expand...

If the SLI connection is indeed one way, rather than bidirectional, Perhaps they could do an implementation where the master card data went over one bus, and the slave card date went over the other bus. The problem is that would only work for two card SLI. Tri-SLI -quad-sli would not be able to use that type of implementation.
 
Last edited:
duhokay said:
He is talking about 3d surround which is almost double that of surround right?
Click to expand...

What do you mean exactly? If the physical limitation on paper is 43 FPS, with overhead and crosstalk, maybe even 30 FPS is proving difficult.

Now for stereo, you basically render two images on the same screen on alternating Hz. You end up with 15 FPS 3D Surround.
 
Spare-Flair said:
What do you mean exactly? If the physical limitation on paper is 43 FPS, with overhead and crosstalk, maybe even 30 FPS is proving difficult.

Now for stereo, you basically render two images on the same screen on alternating Hz. You end up with 15 FPS 3D Surround.
Click to expand...
so you're saying my current 60hz monitor can do 3d? Excellent! :)

(on a serious note, I've heard 60hz 3d looks like ass, so I'm totally being facetious above)
 
The extra frames rendered for 3ds should take up extra bandwidth right? If this is the case then I wonder what they were doing to get playable frames on 3ds pre release demos at this "max" reslolution?
 
duhokay said:
The extra frames rendered for 3ds should take up extra bandwidth right? If this is the case then I wonder what they were doing to get playable frames on 3ds pre release demos at this "max" reslolution?
Click to expand...

Wood screws, obviously :D

The only demos I've seen were running 3x 1080p (5760x3240) which is feasible, even with the supposed bandwidth problem.
 
Got a reply today from NVIDIA. Been mailing since 4/26.

"I’ll have BDR send you an update."
 
Both Brent and my primary Eyefinity setups use 3x1 Dell U2410 panels.

We used those 22" TN panels for 3x2, just due to size and expense.

We have used 3x1 30" 3007 units here for testing as well. It is almost too big vertically.
 
harmattan said:
Wood screws, obviously :D

The only demos I've seen were running 3x 1080p (5760x3240) which is feasible, even with the supposed bandwidth problem.
Click to expand...
3x 1080p screens would only be 5760x1080

Or, 3240x1920 in portrait mode
 
I like the idea of drywall screws a bit more, though 5760x1080 was the mentioned "max" res right? This would give very low 3ds frames.
 
For those that don't get the wood screws comments, this is a joke regarding Nvidia's fake Fermis that they put out when they had no real product to show.

http://pc.watch.impress.co.jp/docs/news/event/20091001_318660.html

The PCI plate is held on by woodscrews, the vents don't line up to vent anything, the DVI port connects to nothing, the backplate blocks the SLI connectors, the power connectors have no corresponding pins on the PCB, etc. It's basically random parts glued and screwed together.

Fermi_end_plate_cropped.jpg


Fermi_back_cropped.jpg


Fermi_power_cropped.jpg
 
Last edited:
sbuckler said:
How is this possible months ago then:
http://www.youtube.com/watch?v=XP7u3ZBF1OM&feature=player_embedded

It's 3d surround on sli 480 GTX's using the new acer 23.6 inch 120hz full HD monitors?

Obviously it works, they just need to release the drivers.
Click to expand...

Don't you mean last month? And yea it looks like it could work but pay close attention to the scenes you're being shown in that video. A sky dive scene from JC2 that is probably the least taxing part in terms of graphics in that game. Then we're shown a racing demo (that doesn't even look that good imo) on a track with no other cars. Something seems off to me.

I don't think it's possible to run JC2 on 2 GTX480's at any kind of settings at 5670x1080 and manage to stay locked on the 120 fps required for the 3d effect. Especially not in high action areas.
 
Hamish Campbell said:
I don't think it's possible to run JC2 on 2 GTX480's at any kind of settings at 5670x1080 and manage to stay locked on the 120 fps required for the 3d effect. Especially not in high action areas.
Click to expand...

Stereoscopic 3D does not require 120FPS to work correctly. It will work equally well at 60FPS (each frame displayed twice), or even 30FPS (each frame displayed four times).

What is required is a 120Hz monitor so each eyepiece in the glasses can operate at 60Hz, which is about the bare minimum required to get rid of the flickering induced by LCD Shutter Glasses. Framerate doesn't matter at all.
 
Unknown-One said:
Framerate doesn't matter at all.
Click to expand...

Maybe not for the technology to function, but it matters a great deal if you actually want to have a decent game play experience. For example, you'd have to limit some games at that resolution to something like 15 or 20 fps because of the aforementioned bandwidth limitations. 3d Surround or not, that seems like more frustration than fun.
 
sbuckler said:
How is this possible months ago then:
http://www.youtube.com/watch?v=XP7u3ZBF1OM&feature=player_embedded

It's 3d surround on sli 480 GTX's using the new acer 23.6 inch 120hz full HD monitors?

Obviously it works, they just need to release the drivers.
Click to expand...

softTH and a lower resolution, like this?:
http://forums.steampowered.com/forums/showthread.php?p=14623593

Just because Nvidia shows something, doesn't mean they have it ready:
http://www.youtube.com/watch?v=bJJyG67by0U
http://hardforum.com/showpost.php?p=1035703261&postcount=443

They should release the drivers and prove they have it. :p
 
Sure glad I bit the bullet and jumped on a 5970.

I used to be a huge green team fan but it doesn't seem like you can trust them on anything anymore. They really need to stop trying to over-hype and be ahead of ATI all the time. If they have the means to back it up then fine but don't make up BS that you can't deliver.

They could learn a bit from amd/ati. Lay low till they get their shit together, then come out swinging. Right now they are just having a flailing slap hand sissy fight.
 
Tamlin_WSGF said:
softTH and a lower resolution, like this?:
http://forums.steampowered.com/forums/showthread.php?p=14623593

Just because Nvidia shows something, doesn't mean they have it ready:
http://www.youtube.com/watch?v=bJJyG67by0U
http://hardforum.com/showpost.php?p=1035703261&postcount=443

They should release the drivers and prove they have it. :p
Click to expand...

If they had faked it and were using SoftTH, is SoftTH compatible with 3D Surround? Obviously they were in 3D mode in that demo.
 
You must log in or register to reply here.
Back
Top