'you'll never make the turn' pmsl
'you'll never make the turn' pmsl
This is cool. Also has a download link in the video description
Is it safe to assume that they will eventually get better multi-core CPU usage in UE5?
Or the bells and whistles will be limited to 30fps.Is it safe to assume that they will eventually get better multi-core CPU usage in UE5?
Developers aren't beholden to utilize Lumen and Nanite. UE5 without those features is still THE top tier engine out there, capable of producing INSANE visuals... We already know this. The way I look at it is this. There's absolutely nothing wrong with Lumen and Nanite being actual forward looking technologies.. I think when people start seeing developers REALLY utilizing the capabilities of these features... they aren't going to care about 60fps. You'll look at it and you'll think.. "You know what.. it makes sense that it only runs at 30fps" because it will look visually leagues above what most other engines are capable of... Developers will go crazy on geometric detail.. producing visuals never thought possible in real-time at "playable framerates"...
Just a few years ago it was "I can't believe we're ray-tracing even some effects in real time".. and now we're on the cusp of a completely new paradigm in rendering with ray-traced lighting on top of it... and 30fps isn't enough?
I basically think people need to just relax. These are simply new toolsets which can/will allow developers to design aspects of their games with unprecedented detail.. and I can't wait to see how developers begin to mix and match and work around the limitations of these new features.. like they always do.
There's issues to be sure.. but it's early days.. despite us knowing about UE5 for a while now and having early access. If I recall correctly.. UE4 didn't exactly start off with a bang either despite having some impressive tech demos... New paradigms take time for devs to wrap their heads around.. so it's way to early to jump any gun to say what it is or isn't.. or what it will be.
Same exact situation as PS4, which didn't even have SMT. Was originally just 6 cores available for games, then Sony opened '7th core mode', but really only half the core's resources were available. This matches what XB1 was doing from the start.I wonder how it that possible?
Maybe…
Game: 6 physical + 5 virtual
System: 2 physical + 3 virtual
So it counts as 6.5 cores? Or there is a way to say to a process to use only half of the GHz of a core?
One minor clarification from the video: I believe the VSM resolution bias is actually the same as "Epic" on the consoles (-1.5) in the Matrix demo, but VSMs additionally pick their resolution based on the (dynamically scaled) render resolution. Thus if you're rendering at 1/2 the (pre-TSR'd) resolution or something, you will also get roughly half the shadow resolution/data, as the goal is to match the sampling rates as closely as possible.Alex released his video.
Thanks for the great insight there Andrew, that is some interesting behaviour (that makes sense!).One minor clarification from the video: I believe the VSM resolution bias is actually the same as "Epic" on the consoles (-1.5) in the Matrix demo, but VSMs additionally pick their resolution based on the (dynamically scaled) render resolution. Thus if you're rendering at 1/2 the (pre-TSR'd) resolution or something, you will also get roughly half the shadow resolution/data, as the goal is to match the sampling rates as closely as possible.
This behavior is a bit different than other systems that target the "upscaled" resolution in terms of picking texture LOD, Nanite LOD (to some extent) and so on. While this means that trying to compare "TSR vs Native" is made more difficult due to the difference in shadow resolution (i.e. some differences you might see are due to shadow resolution, not TSR itself), there are as many reasons to pick the render resolution as otherwise here. In particular, shadows is a pretty big chunk of frame costs in a lot of cases, and thus it's something we do want to scale with dynamic res.
Games can of course adjust this to work however they want, but just wanted to note the defaults if people are comparing any of this. Alex didn't explicitly say anything to the contrary, but just wanted to clarify since one of the shots was comparing the VSM resolution between some of the different platforms.
Haven't seen any yet.Are there examples of how it runs with nanite + standard lighting, and with standard geometry + lumen ?
Thanks for the great insight there Andrew, that is some interesting behaviour (that makes sense!).
Would you have any more insight as to exactly how it works on console? As based on what I am seeing on PC, PS5 VSM quality still lines up below what I am finding there even when pumping the internal pre-TSR resolution down quite a bit. Either implying we are quite sub-1080p on PS5 (under 30% screen percentage would be very unlikely) or perhaps it is not using Shadow Quality 3 on console (more likely)?
Notice how the poll shadow is the same level of sharp on all shots. I think that is mainly because the shadow texel grid is aligned well the the shadows casting direction, so you cannot notice its res really.It will be interesting to see how Intel Arc GPU's run this game with their
Maybe a bug?
Looking at the shadow cast by the pole it looks as crisp as TSR 100%.
And check out that PS5 texture resolution compared to the others, isn't that linked to internal resolution?