Nvidia Blackwell Architecture Speculation

  • Thread starter Deleted member 2197
  • Start date
How does framegen work with fps caps? If my raw framerate is 100fps and I framecap to 240 fps with 4x frame gen does it force the raw frame rate down to 60 fps to make room for the interpolated frames?

Yes. You wouldn't want to do that. You'd probably want to pick the 2x option to go from 100->200, but you could do 3x to go from 80->240. At least this is my understanding of how it works. Although if you're gpu limited, you rarely get a perfect scaling on frame gen, so you'd probably have to play around with each one to see if it actually scaled to your fps limit.
 
Frame pacing would look terrible if they wouldn't be doing it. But we don't know how it's handled in MFG.

Maybe the frame gen is smart enough that if you select 4x, but you only need 2x, it'll dynamically adjust, but I'm highly skeptical. If you were at 100 fps and then capped at 240 and turned on 4x frame gen, you'd be generating 120 frames to 100 fps. So do they drop 20 of them? But then all of the frame intervals would be different, which would look not great. The other option is to limit fps output to 240 which would mean render at 60, and then all of the frames would have smooth and equal frame times. They have that flip metering hardware now, which means they're analyzing the display times to control pacing. Seems like their intent is to make sure 240 fps is as close to 4.16666 ms display times as possible.
 
If you were at 100 fps and then capped at 240 and turned on 4x frame gen, you'd be generating 120 frames to 100 fps. So do they drop 20 of them? But then all of the frame intervals would be different, which would look not great.
We're looking at a Gsynced scenario with Reflex frame limiter in play as FG doesn't really work with vsync. This alone takes care of minor frametime variations so these aren't important.
What FG does is insert an interpolated frame between two frames which are already rendered. Which means that you know their frametime and you know how close they are to the limiter. So dropping the interpolated frame here if the rendered frametime is at the limiter is easy.
When these frame times are far from the limiter then the interpolated frame will be shown. If the resulting frametime hits the limiter than the GPU will likely be throttled so that the rendered frames do not exceed the needed frametime.
With MFG though we're looking at most people being limited by the frame limiter most of the time probably so it will be interesting to see how that would work.
Starting with 60 and on a 120 display you would be running with 2X just fine but 3X and 4X can be a problem if every interpolated frame gets shown as it will push the rendered framerate down.

There was a changelog somewhere where Nvidia said that FG will now drop interpolated frames if you're running at the frame limit but I can't seem to find it.
 
We're looking at a Gsynced scenario with Reflex frame limiter in play as FG doesn't really work with vsync. This alone takes care of minor frametime variations so these aren't important.
What FG does is insert an interpolated frame between two frames which are already rendered. Which means that you know their frametime and you know how close they are to the limiter. So dropping the interpolated frame here if the rendered frametime is at the limiter is easy.
When these frame times are far from the limiter then the interpolated frame will be shown. If the resulting frametime hits the limiter than the GPU will likely be throttled so that the rendered frames do not exceed the needed frametime.
With MFG though we're looking at most people being limited by the frame limiter most of the time probably so it will be interesting to see how that would work.
Starting with 60 and on a 120 display you would be running with 2X just fine but 3X and 4X can be a problem if every interpolated frame gets shown as it will push the rendered framerate down.

There was a changelog somewhere where Nvidia said that FG will now drop interpolated frames if you're running at the frame limit but I can't seem to find it.
What is the relationship between FG and vsync? I think vsync gets greyed out when turning on FG and I really don't know what's going on. I don't see any obvious tearing with FG enabled, but some games don't have much tearing even with vsync off. 🤔
 
What is the relationship between FG and vsync? I think vsync gets greyed out when turning on FG and I really don't know what's going on. I don't see any obvious tearing with FG enabled, but some games don't have much tearing even with vsync off. 🤔

I think they’re just dodging vsync latency but it may have something to do with handling frame gen if you miss a vsync window.
 
How does framegen work with fps caps? If my raw framerate is 100fps and I framecap to 240 fps with 4x frame gen does it force the raw frame rate down to 60 fps to make room for the interpolated frames?
The best way to use frame gen afaik is to have V-Sync on in Nvidia settings. This ensures on a 120hz display that framerate is capped to 116fps which includes generated frames so game engine is limited to 58 actual rendered frames per second.

No tearing, frame pacing is good in most games (had some issues recently with FS 2024 in heavily CPU spiky scenarios though).

I would imagine with MFG the game engine will automatically just be limited to slightly below 1/3 or 1/4 of the monitor's refresh rate if V-Sync is forced on in NVCP.
 
Last edited:
The best way to use frame gen afaik is to have V-Sync on in Nvidia settings. This ensures on a 120hz display that framerate is capped to 116fps which includes generated frames so game engine is limited to 58 actual rendered frames per second.

No tearing, frame pacing is good in most games (had some issues recently with FS 2024 in heavily CPU spiky scenarious though).
The best way is to use a driver override in the NVCP?
 
The best way to use frame gen afaik is to have V-Sync on in Nvidia settings. This ensures on a 120hz display that framerate is capped to 116fps which includes generated frames so game engine is limited to 58 actual rendered frames per second.

No tearing, frame pacing is good in most games (had some issues recently with FS 2024 in heavily CPU spiky scenarios though).

I would imagine with MFG the game engine will automatically just be limited to slightly below 1/3 or 1/4 of the monitor's refresh rate if V-Sync is forced on in NVCP.

This makes the most sense to me.
 
The best way to use frame gen afaik is to have V-Sync on in Nvidia settings. This ensures on a 120hz display that framerate is capped to 116fps which includes generated frames so game engine is limited to 58 actual rendered frames per second.

No tearing, frame pacing is good in most games (had some issues recently with FS 2024 in heavily CPU spiky scenarios though).

I would imagine with MFG the game engine will automatically just be limited to slightly below 1/3 or 1/4 of the monitor's refresh rate if V-Sync is forced on in NVCP.

I thought the 116 fps limit was from Reflex. vsync via NVCP should lock you to 120hz on a 120hz display as far as I know (recommended for Gsync). But Reflex will limit you to 116.

If you cap frame rate via some other means like RTSS then it will usually limit the base frame rate with frame gen activate as far as I'm aware, i.e. 60fps lock in RTSS with FG activated will limit your base frame rate to 30fps. Some games may handle it properly though with their inbuilt limiters.
 
I thought the 116 fps limit was from Reflex. vsync via NVCP should lock you to 120hz on a 120hz display as far as I know. But Reflex will limit you to 116.

At least in Cyberpunk with Framegen on (so Reflex is forced on) I still get uncapped framerate if V-Sync is off and I'm going above my monitor's refresh rate. (Just tested this to see if this is still the case). V-Sync on fixes that, but I guess it could use Reflex to do the limiting.
 
What is the relationship between FG and vsync?
DLSS FG basically doesn't work with vsync since vsync destroys its framepacing and adds its own lag. With gsync and forced vsync you're getting automatic Reflex/driver limiter at some figure below your monitor maximum refresh which ensures that FG basically never hits vsync limit and in this case you get proper frame pacing. Without Gsync driver forced vsync locks the game at 1/2 refresh and uses FG to go up from there to vsync limit.

I think vsync gets greyed out when turning on FG and I really don't know what's going on. I don't see any obvious tearing with FG enabled, but some games don't have much tearing even with vsync off.
Well it depends on what you framerate is and what your monitor refresh is and if Gsync is working. You won't get tearing inside Gsync range with FG but you may get some at its edge and will get some outside of it without vsync.
 
Back
Top