Sli in DCS for VR?

Ive been away about 6 years whats the word on sli in DCS especially for VR?

Thanks in advance

Not supported.

Ive been away about 6 years whats the word on sli in DCS especially for VR?

 

Thanks in advance

SLI isn't supported by any of the VR systems currently available. It's not needed anyway as a single modern card has enough power for VR on its own.

We have to wait to see what is Vulcan bringing on this field.

Is there a technical obstacle that prevents VR from working with multiple GPU's, beyond the fact that they simply aren't supported?

Would it be possible to have VR drivers written that take good advantage of multiple GPU's?

Is there a technical obstacle that prevents VR from working with multiple GPU's, beyond the fact that they simply aren't supported?

 

Would it be possible to have VR drivers written that take good advantage of multiple GPU's?

As far as I can tell both the GPU drivers (which is less of an obstacle honestly) and the software of the application you want to use must be "written" for VR SLI. It seems that SteamVR SLI is working and that's because it was made for SLI in VR from the start. Not so DCS as of now sadly.

... It's not needed anyway as a single modern card has enough power for VR on its own.

Mate, Pimax 5k+ and 8K are out, this isn't true at all anymore. I wish I could have the benefit of a SLI setup just for DCS VR with Pimax.

Yeah, it's really frustrating that you can't even throw money at the Pimax/DCS performance problem.

Is there a technical obstacle that prevents VR from working with multiple GPU's, beyond the fact that they simply aren't supported?

 

Would it be possible to have VR drivers written that take good advantage of multiple GPU's?

Good question. In the Linux world you can wire up as many GPUs as you want for compute applications, but for graphics sli and crossfire was the only game in town and I guess that's not a thing now. Thing is, it's not totally clear that a 2080ti is enough for the next gen headsets. AMD has proposed dual GPU setups for VR -- one for each eye -- and plans to work this in to vulkan. We will see how it works. VR is all about the render time and we don't have much time to work with, you fix one bottleneck and another comes up. But yeah I'm with you, I'd much rather hack together a GPU cluster than to start shopping for research/professional GPUs, at whatever cost, hoping it'll work

Yeah, it's really frustrating that you can't even throw money at the Pimax/DCS performance problem.

LOL... You just aren't throwing enough money at it. I'm sure a big pile of cash to ED would get things fixed quicker.

The main issue is that DCS won't support VR-SLI. since its an NVIDIA specific technology.

The main issue is that DCS won't support VR-SLI. since its an NVIDIA specific technology.

Not necessary. With switching to Vulcan engine this might be solved by default or with some tweaking in graphics engine. Point is Vulcan give you such possibility DX11 not.

It is pointless to work on this issue on old engine when switching to Vulcan is around the corner. That is the point.

Lol... Ill be happy if we see vulkan in 2 years.

Is there a technical obstacle that prevents VR from working with multiple GPU's, beyond the fact that they simply aren't supported?

 

Would it be possible to have VR drivers written that take good advantage of multiple GPU's?

My understanding is latency. There isn't enough bandwidth between the cards to sync fast enough with VR head movement.

Well if it is latency and if the GFX makers care they can fix that in the next gen.

Well if it is latency and if the GFX makers care they can fix that in the next gen.

That is likely the crux of the situation.

VR SLI is a huge niche in an already niche technology.

It's only a niche till it's not. I think it's obvious GPUs have hit a wall, just like CPUs have. It's getting exponentially harder and costlier to push ever faster clock speeds and IPC. The current top card (2080ti) costs well over a grand to get a fairly marginal improvement over the previous top card (1080ti). This isn't sustainable. Nobody is going to pay $2K for a 3080ti that's 25% faster than a 2080ti.

Hardware vendors are going to have to start going wider instead of faster, likely delivering multi-GPU single cards and forcing code writers to deal with multi-GPU rendering. It'll be painful, just like multi-threading has been painful for coding on the CPU. It's inevitable, though. Moore's law is just too much of a #$%& to keep up with.

My understanding is latency. There isn't enough bandwidth between the cards to sync fast enough with VR head movement.

Latency is only an issue with alternate frame rendering. Committing a GPU to each screen largely solves that issue.

 

 

Latency is only an issue with alternate frame rendering. Committing a GPU to each screen largely solves that issue.

Then you have the syncing issue though.

With no disrespect intended, this is not correct. Nvidia simply is not supporting multi GPU gaming at the moment and they would need to extend the API in order for ED to be able to patch DCS to work with SLI. (unless ED wanted to spend big bucks and many man months of trying to reverse engineer an SLI patch)

No disrespect taken my emperor :). Well, in a way that was basically what I was saying, is that ED would probably have to do the heavy lifting to get it done, hence throwing big money at them for a miniscule market slice. Also I thought NVIDA did support it via VR-SLI?