Yep, as a joke. But Nvidia's doing it. It looks OK in a 15 second CyberPunk clip but that's got to be the most overfit game.
Yep, as a joke. But Nvidia's doing it. It looks OK in a 15 second CyberPunk clip but that's got to be the most overfit game.I remember that someone on this forum said when nvidia anounced the dlss3 framegen that the next generation they will just keep putting more interpolated frames to claim big performance increases versus the previous generation. Spot on.
I'm removing my old 4070Ti super from my case and will use it as a cat littler pooper scooper. Old, slow, stupid, single-frame-gen garbage.
aI frAmE gEn mAx prO - now we generate more fake frames to interpolate even more in between real frames, your fps can now go from 60 to 240. But we only limit to 5xxx series because 4xxx doesn't have enuf bw.
/s
I remember that someone on this forum said when nvidia anounced the dlss3 framegen that the next generation they will just keep putting more interpolated frames to claim big performance increases versus the previous generation. Spot on.
Yep, as a joke. But Nvidia's doing it. It looks OK in a 15 second CyberPunk clip but that's got to be the most overfit game.
I called it.... Only my naming was off. I even got the frame gen factor correct 😂.View attachment 114378
DLSS4 isn’t all locked to RTX 50 series, only the Multi Frame gen is
the raw data is the FC6 result which doesn't look all that good
next is multiframe x8 exclusive for RTX 6xxxonly the Multi Frame gen is
Keep in mind that human "moving quickly" is eons for a computer running at GHz speed. I remember programming my Atari 800 in Assembly and in the time it took the electron beam to move from one side of the screen and get to the other side to start drawing the next line, called the Vertical Blank Interupt, there was time to change color clocks and "fake" more colors than were actually available. And that was using a CPU running on the order of MHz, not GHz.If it's like FG v1 (which it probably will be in this sense), it will work really well for when you are standing around or moving relatively slowly, but once you start moving quickly or make a sudden move in a different direction, it becomes really messy. It will be interesting to see how quickly the GPU can drop the predicted frames and switch to rendered frames in these situations.
Called the RTX 5080 pricing but got the RTX 5090 pricing a bit off.MLD is likely wrong and someone leaked some fake pricing just to make the real price seem pretty bearable. That pricing is stupid high and would lead to backlash.
Nvidia will use the founder edition and other things to keep pricing good for atleast the launch.
I think something like 999 to 1099 for the rtx 5080 founder edition to get reasonable reviews. With partner cards 100 to 200 more than this. So real pricing will be 1099 to 1199.
The Rtx 5090 will probably be 1799 for founders to get good reviews with partner cards and thus the real pricing be 1999.
Nvidia is not stupid.
$999 and 1799 will atleast lead to reasonable reviews and not much outrage if we compared to this so called leaked pricing. Partners will have higher prices which is where the true pricing will lie.
Because it can't predict what you are going to do. It can only at most predict until your next input.I dont see why the predictive frames cant use your input, especially for simple things like camera translation or rotation
It would be jarring in an FPS if the frame rate dropped by 75% every instant you pulled the trigger, but i am very curious to see how it handles sudden screen changes like muzzle flash, or opening a door, or scrolling text or a million other things
So last gen's killer feature is killed...?We have also sped up the generation of the optical flow field by replacing hardware optical flow with a very efficient AI model
Well it was slow. Good they offed it.So last gen's killer feature is killed...?
NV whitepapers are pure nontent.I swear it seemed like a quarter of Ada's whitepaper.
Of course, they're for AI summarization.NV whitepapers are pure nontent.
I dont see why the predictive frames cant use your input, especially for simple things like camera translation or rotation
It would be jarring in an FPS if the frame rate dropped by 75% every instant you pulled the trigger, but i am very curious to see how it handles sudden screen changes like muzzle flash, or opening a door, or scrolling text or a million other things
Chungus vapor chamber with tons of heatpipe and moar airflow.I didn't look too closely. But it's 2 slot GPU cooling 575W? 🤔
I wonder how that works.
110-ish vs 240-ish, but with DLSS4 - so generated frames are 1+1 and 1+3?4090 vs 5090
Because it can't predict what you are going to do. It can only at most predict until your next input.
But yes, how do you predict a muzzle flash for 3 frames and they have it match the computation on the 4th frame without something not looking "right?"
It's not predictive. These are still in between frames. It's holding back a real frame, inject 3 fake ones, and then showing the real one, that it was holding back.
Nothing can change those three fake frames, because they are in the past, and it already has a real frame after them used to generate the steps in between. You can only influence the future after the held back real frame drops.
Example:
Roll a ball across floor.
No FG:
Real frame 1, ball 2 feet away,
Real frame 2, ball 8 feet away,
RTX 40 FG:
Real frame 1, ball 2 feet away,
Fake frame 1, figure out ball is 5 feet away, by interpolating between real frames.
Real frame 2, ball 8 feet away (this frame was held back and used to generate fake frame 1)
RTX 40 FG:
Real frame 1, ball 2 feet away,
Fake frame 1, figure out ball is 3.5 feet away by interpolating more steps between real frames.
Fake frame 2, figure out ball is 5 feet away by interpolating more steps between real frames.
Fake frame 3, figure out ball is 6.5 feet away by interpolating more steps between real frames.
Real frame 2, ball 8 feet away (this frame was held back and used to generate fake frame 1,2,3)
Not much they can do until 24Gb G7 ICs have ramped.Still think it's kinda cheap that NV only put 12GB vram on a 5070.
Either the CPU limit or something in the uarch has scaling oopsies.That seems to suggest no real substantial perf increases with non-DLSS in that CyberPunk, how is that possible with bigger chip, faster memory, clocks regression???
Yeah, it's a wait for the Super or whatever they call the 18GB refresh. But the requisite memory chips must be prioritized for laptops.Still think it's kinda cheap that NV only put 12GB vram on a 5070.
Yeah, that why i'm looking into 9070. I'm not interesting into multifake frames, never gonna enable themStill think it's kinda cheap that NV only put 12GB vram on a 5070.
It's lower than I expected. Only 1.77x membw...Wonder why did they not show any 8k results to showcase 2x membw bump on gb202.