DavidGraham
Veteran
It ran much better under DX11.
DLSS doesn't require DX12, it works very well under DX11, and was featured in many DX11 titles.
Current Generation Games Analysis Technical Discussion [2023] [XBSX|S, PS5, PC]
- Thread starter BRiT
- Start date
- Status
And I assume that's without Direct Storage. Very impressive!
Yea, no DirectStorage. All the other ones load extremely fast too.
It's going to be interesting once games start utilizing DirectStorage more and more. I'm looking forward to AMDs upcoming GDC talk/demo about DirectStorage.
This is (hopefully) exactly what I've been waiting for. Something to comprehensively demonstrate upfront load speeds, streaming improvements, and how it will improve player experience compared directly against standard loading. Hopefully their demonstration has examples from actual upcoming games and not just synthetic ones.
Yes that'll be very interesting. I wonder if they'll demonstrate their smart access storage as well.
Flappy Pannus
Veteran
Yeah that's also what I'm really curious about, bypassing main memory entirely and just pumping it directly into VRAM.
This is an expensive feature you can turn on or off which costs multiple ms to render. It easily fits within a 60fps budget on a high end gpu (runs fine on my 3080) -- it's not "unoptimized" just because it's expensive or doesn't scale the way you expect from other features in other games (I know it may appear that way after what feels like decades of ps4 ports, but not every feature is fragment shader bound or scales by screen size.)
Maybe it could be faster, maybe it couldn't, but you can just turn it off. PC gamer "unoptomized game" complaints have gotten really ridiculous.
Last edited:
Flappy Pannus
Veteran
207 fps dropping to 129fps for hair on one character alone is quite drastic though, that's a 38% fps drop. In my tests without RT at much lower framerates, it was more like a 20% drop.
So there could be a bottleneck with how this is calculated when we're getting into very high frame rates when it's holding up the scene rendering, in that sense this feature could be considered 'unoptimized' in that it doesn't scale that well...perhaps (possible a release day driver could alleviate this somewhat too). I don't think it's completely out of pocket to see a performance drop you regularly get with something like RT reflections for slightly more detailed hair and not say 'wtf'.
That being said, I agree though that considering 1) A simple toggle fixes the issue, and 2) One that seems to only manifest this severely on very high end hardware, is no way just cause to label this title 'unoptimized' as a whole. It's a quirk for a feature with minimal visual impact to begin with, a very slight blemish on a release (not even a release!) that in most other aspects, is exceptional in its optimization. Like I said earlier, if this is your line in the sand for what you consider an 'unoptimized' title, you should have binned your PC years ago.
The percentages here can paint a misleading picture, it's probably only 40% of the frame at super high end framerates -- most games are not really designed to run at ~200+fps, unless your computer is hyper fast (like, from a decade from now) there's just so much going in to scheduling and parallelizing work to get it to run an entire frame in ~5ms every time that it's impractical. In my opinion 144fps is the absolute maximum "reasonable" target framerate for a super high end PC, and its sitting right under that with everything turned on for you -- that's basically perfect performance.
So the worst case scenario (maybe caused when like, the rt and the regular raster rendering have to wait until some expensive compute shader workload is finished before they can start?) it's ~4ms, and in the best case it's ~2ms.
Of course it doesn't! The work is almost surely in doing the simulation and generating the geometry.
Also, wait, your after number is 156 fps? It's OK to play a game at 1080p if you want to ensure a locked 144hz.
PS5 uses a terrible CBR, even on XSS the grass is not pixellated as ps5. Don't know the hell going on here. But it would explain why runs better and at slightly higher resolution than XSX, cheaper reconstruction method.
A 4090 can render a 1080p path tracing frame in Portal RTX within 10ms. And yet it needs 3.6ms to render and simulate hair... Cant believe that there is someone who defends this unoptimized mess...
davis.anthony
Veteran
Two different work loads so your point is moot.
And you can't say something is unoptimised without having some that is optimised to compare it to, what other hair systems are there? How do they perform in relation to this?
Hair has always been very complex and processing intensive which is why many games don't try to push it.
davis.anthony
Veteran
Is it of the same quality? A given the camera angles in typical Fifa gameplay the LOD for the hair should be pretty low where as Leon's hair should always be at the highest LOD.
So are they comparable?
When TressFX debuted in 2013's Tomb Raider that could drop FPS by up to 20% on some GPU's.
No, not "two different workloads". Both are pure compute workloads with simulation. One is the most complex rendering technique on this planet, the other is just hair rendering. A 3.6ms workload on a 90 TFLOPs GPU is absurd. My 4090 is barely faster than a 6900XT in 1080p: https://gamegpu.com/rpg/ролевые/resident-evil-4-chainsaw-demo-test-gpu-cpu
Yet it is 3x faster in Cyberpunk with Raytracing in 1080p: https://www.techpowerup.com/review/gainward-geforce-rtx-4080-phoenix-gs/33.html
Yet it is 3x faster in Cyberpunk with Raytracing in 1080p: https://www.techpowerup.com/review/gainward-geforce-rtx-4080-phoenix-gs/33.html
arandomguy
Veteran
"Realistic" hair rendering (in this case I believe stand based) is complex which is the reason why characters in games for the most part don't have actual hair but basically wear hats. It faces a similar issue that ray tracing has in that we're kind of used the "good enough" fake method (especially with static shot comparison, again something similar to RT) with everyone wearing hard hats.
Also in this case I wonder how much of work load is being limited elsewhere? Eg. the CPU and memory subsystem hit on the implementation.
I've only curiously looked into this but the PS5 implementation supposedly heavily drops the resolution to compensate? Can we say it's a PC port issue at this point? Other stand hair solutions so far have been rather heavy.
Also in this case I wonder how much of work load is being limited elsewhere? Eg. the CPU and memory subsystem hit on the implementation.
I've only curiously looked into this but the PS5 implementation supposedly heavily drops the resolution to compensate? Can we say it's a PC port issue at this point? Other stand hair solutions so far have been rather heavy.
- Status
Similar threads
