DLSS for VR, does it make a difference?

What difference will DLSS make for VR in MSFS?

I currently have a GTX1080ti that I’m happy with. I’m wondering whether to upgrade to a RTX2070 Super which supports DLSS.

DLSS dramatically improved the VR performance of No Man’s Sky when it was introduced, but as it’s a DX12 feature and DX12 performance in MSFS is not great, we’ll just have to wait and see.

1 Like

DLSS does work on DX11 just fine, it’s not just DX12, various games support it under DX11.

In answer the OP, it will vastly improve the performance (assuming they do a good job implementing it)

One this to note, DLSS also sticks a fairly hefty load on the CPU, so it’s one thing to bare in mind.

@Raj360 personally, a 1080ti to a 2070 is a downgrade, 1080ti outperforms a 2070 in most cases, if you were looking for an upgrade you should think about a 2080ti or a 3070.

3 Likes

Based on the latest road map further improvements to the DX12 implementation are scheduled for SU10+, so I wouldn’t be surprised if we get a DLSS ‘preview’ in SU10 rather than a finished version.

1 Like

Thanks.

I have an older CPU, i7 6700 so not sure if it makes sense to pair it with anything more than a 2080. When prices calm down I’ll have to upgrade the whole system.

I’ll be waiting to see the first reviews with great interest.

Although DLSS is slated to come to MSFS in the next sim update, AMD’s new FSR 2.0 is also being worked on by Asobo. Being open source, it doesn’t need AMD hardware to work, so should be an option for both nVidia and AMD Radeon GPUs. It looks pretty good so far, so am looking forward to using it on my Radeon 6800XT. Comparison here in this video.

Impressive! AMD FSR 2.0 vs Nvidia DLSS.

2 Likes

I’m more excited for FSR 2.0 than I am DLSS. I use Mbucchia’s OpenXR toolkit and FSR is better than NIS with my 3090. First reviews of FSR 2.0 on Youtube have shown it beats DLSS in some instances. I have more faith in Mbucchia’s implementation of FSR 2.0 than I do Asobo’s DLSS.

Sorry to disappoint, but just like we could not implement DLSS in OpenXR toolkit, we cannot implement FSR 2.0. They have the same requirements for depth buffer, motion vectors, and subpixel jitter.

AFAIK Asobo announced they will add support for FSR 2.0, so I’m likely not even going to try myself.

That’s disappointing, although I am already exceptionally happy with the performance I’m getting right now anyway. I’ve started using the world scale feature too, increasing the size by 10 percent and it feels so much more real now!

So does FSR 2 use machine learning like DLSS does? Because FSR 1 doesn’t at all, right? Just curious why they are even being compared when they seem to be quite different things. Or are they?

No these 2 things are not to be confused: requirements can be the same but still different underlying technology. I think AMD made it clear there is no ML component in FSR 2.0.

Motion vectors and subpixel jitter are commonly used for a wide variety of effects. Depth re-adds the 3rd coordinate (z) when looking at a 2D image.

AFAIK DLSS and FSR 2 are only comparable because they fall into the temporal algorithm category (meaning relying on historical image data, rather that single frame).