rickybooby87
Member
You're probably right. But still, it gives me great joy to see the PS5 continually "punch above its weight".
Digital Foundry: A Frame-Rate Free Lunch? FSR3 Frame Generation on PS5/Xbox Series X|S Tested
- Thread starter adamsapple
- Start date
You're probably right. But still, it gives me great joy to see the PS5 continually "punch above its weight".
rickybooby87
Member
Agreed, we need NXGamer to do a breakdown as well, with much more thorough testing. I am taking DF at their word in my conclusion.
Well for one - it's a game-specific thing - not a device thing - and some games absolutely do.
It was pretty common to implement specific latency reduction measures for console VR titles as an example (in many cases, superior to what you'd found in PC versions of the same titles).
Eg. one of the titles I worked on, at 60fps, had a 'engine-native' motion-2-photon latency at 45-50ms, which we reduced to 30ms by launch. PC version of the same title achieved a similar latency - but only because it ran at 90fps.
The biggest gains can be achieved by rearranging rendering and simulation both (and more importantly - how they interact) - but 'viability' of some of those varies greatly from game-to-game - as sometimes the changes are prohibitively complex to implement.
Work NVidia's doing on profiling tools/hw is pretty interesting though as - for many developers, the idea of even measuring this is pretty foreign.
So to show how he was right you are using a video where he already knew about this game getting this in consoles?
Black_Stride
do not tempt fate do not contrain Wonder Woman's thighs do not do not
I code and I didnt know this.
What makes it easier to optimize for PS5 compared to XSX?
funking giblet
Member
Not all workloads are parallelizable. All workloads benefit from faster GPU.
Last edited:
Black_Stride
do not tempt fate do not contrain Wonder Woman's thighs do not do not
And the Xbox Series X CPU is faster........whats your point?
funking giblet
Member
Talking GPUs here brah (obviously I said CPU above.. I meant GPU and I've edited it as such), as that's where the whole "Fast vs Wide" argument with TF comes from...
Last edited:
Zathalus
Member
It was the very first time they tested the possibility. Previously they claimed FSR3 would not really work for making 30fps games 60fps, and that according to the information they got directly from AMD themselves the performance cost was maybe way to high for the console GPUs.
The very first time they tested it themselves they then realized that AMD was being extremely conservative about the performance requirements and made the video.
Black_Stride
do not tempt fate do not contrain Wonder Woman's thighs do not do not
You said CPU.
Nonetheless that's incredibly incredibly reductive.
I thought you had some insight I missed.
There's nothing you've said that backs up your claims it's easier to optimize for PS5.
funking giblet
Member
Physics? The same workload on a CU on PS5 will run faster than the Xbox. Or do you think that isn't so?
Black_Stride
do not tempt fate do not contrain Wonder Woman's thighs do not do not
That's assuming all else is equal.
Which it isn't.
And that also doesn't tell anyone why it's easier to optimize for PS5.
The 6750xt runs faster than the 6800 by quite a bit. Does that mean it's easier to optimize for the 6750xt?
As I said earlier you are being very reductive and making it seem like clock speed automatically means easier to optimize for.
Shane89
Member
DF, especially the beaver said a lot of BS
funking giblet
Member
No as I'm not just talking about clock speed am I? Narrow+Fast vs Wide+Slow, it's the fact that you won't be able to use all the CUs effectively, meaning you will not get access to the power of an Xbox, vs a PS5, because when you're dealing with sync locks, memory barriers etc across multiple CUs, that's harder. Having them complete faster, means you don't have to go to excessive trouble to optimize your shader code to to ensure you have coherent data. So you don't have to work so hard. So it's easier.
64gigabyteram
Reverse groomer.
There has to be a catch.
Buggy Loop
get in here
Buggy Loop
get in here
Buggy Loop
Gold Member
There has to be a catch.
Uh?
Didn't really follow the discussion in this thread but watched the DF video. The results are about what was expected no? Motion artifacts and the FSR upscaling being a pixel soup in motion amplifies the error when there's rapid movement. I don't see a catch. Not sure I would enable it if I was on console, 50-60 fps with VRR would already be good enough.
Holy shit at the base latency on the game though, LOL. 130 ms. Garbage for a shooter.
Last edited:
adamsapple
Or is it just one of Phil's balls in my throat?
In a way it only adding 8ms for a boost of 72% in performance for adding interpolated frames is kinda great, the original latency notwithstanding.
Last edited:
Buggy Loop
Gold Member
Sure. But peoples were throwing tomatoes at Nvidia for double performance with 10~12 ms adder and less artifacts
But yes, it can be good. Personally though, if I were to have a game already in 60 fps range, not sure I would enable frame gen for the tradeoff in artifacts. Latency is not really the problem, never has been.
omegasc
Gold Member
the new UE5 version optimizations made room for the Frame Generation code, I believe. The dev mentioned in a reddit post that the version the game shipped with was very poorly optimized. The overall performance is greater now, even without frame gen enabled.There has to be a catch.
