Digital Foundry Article Technical Discussion Archive [2014]
- Thread starter DieH@rd
- Start date
-
- Tags
- digital foundry
- Status
djskribbles
Legend
4xMSAA + PP AA on PS4? Wonder how 1080p with SMAA would compare visually and performance wise.
Looking at the screens and the vid, the difference between the visuals is marginal. The two look far more similar than not. The difference that is more pronounced, that I think is actually a much bigger deal, is the load times. Why does it take Xbox One nearly twice as long to load a match? Is the Xbox One drive that much slower, or is there something about the OS that's causing slow load times? Memory bandwidth should not be an issue because drive speed would bottleneck first. CPUs are a wash. Think the difference is in GPU computation that happens before the match starts?
Mostly, though I've found the 4xMSAA to be rather beneficial for edge stability, although that's mostly pertaining to the arena edges in the peripheral vision.
The higher sampled motion blur is one of those nice subtle things too (although they do make heavy use of it).
---
But yeah, the skin shading & general lack of aliasing is pretty darn nice. The MSAA doesn't break very often either (not that I've seen at least).
Shader caching?
I think it may just be upscale shenanigans + other post processing (DOF/bloom etc) rather than post-AA. You can get all sorts of pixel gradations from those alone.
Its a demo so theres hope for the final game but 30fps @ 900p for a fighter game is terrible
theres not a lot of edges thus why 4xMSAA, thats costing a bit of performance (it looks like often 1/4+ of the screen is just the floor texture!)
1080p & best screen AA method
perhaps also use a shader for the cage
result = better performance & looks better
theres not a lot of edges thus why 4xMSAA, thats costing a bit of performance (it looks like often 1/4+ of the screen is just the floor texture!)
1080p & best screen AA method
perhaps also use a shader for the cage
result = better performance & looks better
Why are they analysing a demo?
If the retail version is unchanged from this then it pretty much confirms that the developers aimed for platform parity.
*Shudders*
Inuhanyou
Veteran
^ If one actually read the article, it is much more than just resolution, and framerate. The PS4 has essentially everything of a higher quality than the XB1 version, including a much more refined motion blur(very computationally expensive) on top of twice the AA
SMAA is the best choice. I keep wondering what they could have accomplished taking out the 4xMSAA and going with a customized SMAA x1 It obviously would have been 1080p, but it probably could have been improved in other ways too
SMAA is the best choice. I keep wondering what they could have accomplished taking out the 4xMSAA and going with a customized SMAA x1 It obviously would have been 1080p, but it probably could have been improved in other ways too
They're saying that the PS4 version is targeting double the geometric sampling, double the motion blur sampling, and has better shadowing to boot. The images demonstrate a visually cleaner presentation on PS4.
I'm not seeing the "parity."
Parity being resolution sample – I guess?
I’m guessing some wanted 1080p with a smarter choice of AA sampling. Don't know...
Anyhow, the game looks bland and uninspiring in the graphic department. Nothing that screams “wow” … even the animation looks typical.
The game comes out June 17, so that'd suggest not a lot of changes. Then again we've seen lots of post launch patches I guess if you're hoping.
They didn't aim for parity as the differences suggest...
I hope we see devs continue to actually optimize for X1 like this instead of just shunting over a poor port of the PS4 code.
What if the bottleneck was something besides bandwidth?
Just the AA and Motion blur right? Not "virtually everything". In fact the videos look very similar. Also the PS4 dips frames/tears at the top instead of perfect 30 like X1.
That's seems like a rather strange thing to say, unless you actually know how much resources they're expending on MSAA.
hesido
Regular
For a game that's main visual is two guys kicking eachother, I think MSAA comes almost free because they are targeting 30fps anyway, as in, making it no MSAA wouldn't probably make it run at 60. (I'm sure they are spending a good deal of CPU on muscle and skin rippling simulation, but I still think they should have gone for 60)
Your just being deliberately obtuse.
It's not even close to free, and dropping AA would not double the framerate.
hesido
Regular
That's what I'm saying, since they are going for 30, imagine if without MSAA it would run at 35-40fps without vsync, instead of 32-36 without vsync, therefore MSAA becomes "free"*. Totally making numbers up, as you can tell.
That's not to say I agree with their choice. They just went for a visual fidelity that can achieve 4xMSAA at the resolution they were targeting.
- Status
Similar threads
