DirgeExtinction
Member
The GPU in the ps4 is 1800flops? Or do you mean the CPU+GPU combo in the ps4 is 1800flops?
If so, shouldn't you be doing the same for the 750ti and whatever CPU its usually paired with?
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Digital Foundry Face-Off: The Witcher 3: Wild Hunt (PC/PS4/XB1)
- Thread starter ScepticMatt
- Start date
If so, shouldn't you be doing the same for the 750ti and whatever CPU its usually paired with?
mugurumakensei
Member
You also have to take into account that, for gaming purposes, Nvidia's flops are worth more than AMD's flops.
Yes but how these components come together is also interesting as the gpu depends on the cpu and you also have to consider how much the code depends on cpu usage. Just adding the numbers and saying system a is better than system b will give you a wrong image on the performance.
Yeah ... sure. Right.
Except CDProject themselves said that this game wouldn't be what it is if it wasn't for consoles. And comfy couch, the old insanely bad argument.
That's also true. FLOP COUNTS are not directly comparable between GPU generations let alone different HW vendors. Of course, CPU and other things make a difference in performance as well.
Besides the real point of gaming consoles, which is the games, and the convenience of a single box which plays those games. I don't like throw away hyperbolic posts such as these.
DirgeExtinction
Member
True.
I was just wondering if he was comparing the flops of a discreet GPU vs the total flops of a console.
StrongBlackVine
Banned
This game wouldn't exist(in it's current form) without these consoles. Kind of a ridiculous thing to say regardless.
Nizz
Member
Which is why I'm waiting to see if a patch improves the framerate for the PS4 version. From what I've seen from videos the game looks great. I don't have a problem with that.
What does bother me are the posts about the framerate not being solid. If this ran at 30fps solid or at least 99% of the time I would have picked it up in a heartbeat. So I continue to wait and see if the devs improve the game with more optimizations.
StrongBlackVine
Banned
I picked up because it only cost me $40, but I won't play because yet because of all the complaints about the framerate. The lost save bug is a major issue too.
I think its fair to say patches will imrpove game performance. We've seen that a lot this gen, games being pushed out on the asusmption they can just patch later.
Hell, the Borderlands collection gained 20 to 25 FPS in a single patch alone, going from an unstable and fluctuating 40+ FPS all the way up to stable 60 on both platforms to boot. These CPU's may be low tier, but they can do some decent work if the dev does some extra tweaking to the code
RoadHazard
Gold Member
The PS4 version generally performs well framerate-wise IMO. It's at a stable 30 fps more often than not. The stuttering is a separate issue. Like, if you see stuttering when turning the camera, and then instead try running in a circle so that the camera automatically pans around, you'll often see a perfect locked 30 fps. There are situations where the framerate itself goes to absolute shit though, and it seems to have to do with alpha effects. A dense forest full of detailed foliage can be perfectly smooth in good conditions, but add some rain and maybe a fire or two on top of that? Hello 20 fps. And it is just that, 20 fps. This game never runs at 25, it's either 30 or 20. Looks like a case of simple double buffered v-sync. Had they implemented triple buffering or adaptive v-sync I'm sure the overall experience would have been much smoother. So if they can make that happen, iron out the random stuttering and do something about the massive performance hit of alpha effects, I think this could become a very nice experience. It mostly is already, just not consistently so.
flipswitch
Member
I didn't have any issues with the frame rate in white orchard, it was mostly at 30fps. Now that I'm in Valen, I reached a section where it drops to a very noticeable 20fps, bit of a pain but it's only in a small section. Mind you I haven't gone that far in Valen or anywhere else besides white orchard. I hope they can improve it in future patches.
https://m.youtube.com/watch?v=ibxiGr0V5iA
NXgamer video. Same finding as digital foundry and little bit more details. It's a mobile link however
NXgamer video. Same finding as digital foundry and little bit more details. It's a mobile link however
I pretty much completely agree with this, I've done about 28 hours so far and there's only been one area which has had a sustained drop (a certain swampy area) which I think will be a prime candidate for improvement, seeing as how the forest there is not even as detailed as other areas, yet a drop is still happening. A lot of other stutters appear to be as a result of dodgy streaming. I thought that GTA5 would never their streaming problems but they did so we'll see if they can do it here.
Im there with you. I have copies of both versions and if this weeks patch doesn't smooth things out ill return the game and wait for the fixed gold edition. I have some games in my backlog to fill the hole.
The 103 patch seemed to help a fair bit for pc users so im hoping for the same on consoles. The lack of optimization seems to be this games largest issue so there should be some nice headroom to gain.
My PS4 version looks nothing like that! What settings do you have for your PS4? My PS4 video output settings is 1080p, RGB Range to Full, and Deep Color Output to Automatic.
Do you need RGB Range set to Full? As in, is your TV/monitor set to Full RGB Range too? Because if not you'll get crushed blacks.
Also, all the games are designed for the Limited RGB Range. If your TV/monitor supports Limited, you should have it set to that on the PS4 and your TV/monitor. Just because it's 'limited' doesn't mean it's worse.
Maastricht
Member
Eh, just always use full if your TV supports it and is set to that. Games don't work with 16-250 range or whatever Limited is again, they work with 0-255. GPU hardware can then translate that to Limited output if the TV expects Limited, but if you can do full, you should do full. And sometimes there's even this 12bit per pixel thing, though I don't think any games support that natively, so then it's still some kind of upscaling.
RoadHazard
Gold Member
If you're absolutely certain you're really getting the full range, sure. But if you don't 100% know what you're doing, it's better to leave it at limited. It won't look any different. In fact, if you DO notice any difference in contrast or whatever after switching to full, something is wrong. If everything is working as it should your TV should map the brightest and darkest colors on both settings to the exact same shade on your screen. Brightness, contrast, etc, should all look identical. The only difference, which you're not likely to notice (I'm not even sure it's possible), is that full range offers a slightly more gradual change in shade from the darkest to brightest possible. More "steps" in the color scale, if you will. But the endpoints of those scales should look exactly the same on both settings.
UnrealEck
Member
How so? Could have finished the version they had shown initially then ported it to consoles when it was completed. The money from console sales couldn't have funded it. It must have already been funded.
thelastword
Banned
Analysis
Features:
Originally Dx9 based but has adopted many Dx11 features
Tessellation is in (ground, rocks, bridges)
SSR on water
Weather effects
Global illumination emulation
Dof effects on camera due to weather
Game uses image based lighting
Also uses dynamic point lights which reflects on the environment's objects including geralts armor etc..
PC (970GTX) vs Consoles
Consoles run foliage on high
Ultra foliage on PC draws a greater distance but very minor from high in difference
Lighting and PP effects are at the highest setting on both consoles and PC (bloom, shafts of light, long shadows from objects etc..)
Weather system is equal across console and PC
AO and AF dialed back on consoles (some say due to a patch) in cutscenes
HBAO+ works well with foliage in W3 on PC, not worth it if you don't have an Nvidia card, generally costs more than it impresses.
Grass density medium quality on consoles
Terrain quality is on high with detail level being equal to PC
Particle effects are equal to the Pc's best.
Shadows never seem to change from any notable degree from medium to ultra.
Discrepancies (console side)
Consoles have lower shadow quality and AO than the lowest pc settings at times, but they're predominantly at medium shadow quality and makes use of ssao most of the time.
Textures are predominantly high on consoles with some mix of medium textures, ground textures are usually the culprit for medium quality.
Ultra textures are the same quality as high, but then ultra textures never go beyond 1.8GB of VRAM on PC. (which makes the use of medium textures at all very questionable)
Color depth look 16-bit rather than 32-bit on consoles
Texture loading is slower on consoles, even the integrated A10 R7 on chip gpu handles texture loading much better/faster with ultra textures in place.(smh, should not be)
Loading times are much faster on PC, at least 5 times faster. Sometimes consoles take over a minute to load in some instances. (not common)
Water effects are high on console which includes tessellated movement and interaction, but there's some shader work that's absent on consoles that even the lowest setting on PC has (smh at cdpr, clearly they needed more time with consoles)
Water also lacks the interaction with spells (like the force blast), which is apparent on PC at high settings, such effects are not apparent on consoles.
Note: He indicates that the consoles seem to have been behind in development due to some of the results we're seeing here, just like dying light was. He also indicated that CDPR is primarily a PC developer and PC was the lead platform for development.
He expects that consoles will see a nice boost in quality and performance in the coming patches since it was behind PC in dev for the most part. The texture issue may also be linked to a loading time issue which looks to be API related something which more time and experience can fix.
PS4 vs XB1
PS4 and XB1 has the same settings employed for the most part, however;
PS4 is 1080p vs the XB 900p (which is standard fare by now)
PS4 has 8xAF throughout whilst the XB1 uses a dynamic 4x-8xAF depending on the scene.
Water on XB1 seems slightly better, closer to PC but not quite there.
IQ on XB1 is worse due to 900p and the upscaled image issues with the custom fxaa and temporal aa implementation
Framerate is smoother on PS4 over the XB1. Xb1 framerate is unlocked to help support drops better due to the limited buffer space on XB1.
PS4 would normally drop the odd frame whilst loading new areas (loading issue) which exists on all versions, but since PS4 is capped you will see it dropping to 29fps whilst horseback riding here and there. (my opinion, when the xb1 version is capped it will drop more frames in these sessions, because the unlocked framerate would no longer keep it just barely at 30fps or above)
General impressions and Summary
Cutscenes vary wildly in performance
Hairworks is very unrefined at the moment just as tressfx was initially. Hairworks is not that impressive considering the 20fps performance penalty on average. Geralts hair looks better without it.
Tessellation should be set to at least 8x when hairworks is on (PC) to make it look good.
PhysX is on consoles (though emulated on the CPU)
Witcher 3 is NOT CPU limited, it is GPU limited however. It is NOT a CPU killer. All the testing PC's were almost always at 100% buffer usage (gpu wise). On the flipside, cpu usage even on the A10 hardly every taps a single core. All cores operate with lots and lots of room/cycles to spare. No core is ever maxed out, even paired with the entry level 750ti.
A dual core is enough for the game. A quad core is not entirely needed as it hardly uses over 4 cores.
750ti Shenanigans
The 750ti performs close to the consoles with similar settings, helped of course by its unlocked state in the comparison, however it does dip below 30fps more often than the PS4 and it performs just slightly better than the XB1 which is also unlocked.
750ti is also constantly below 30fps in the swamp area, other areas it holds up a bit better and is more comparable to the PS4. It's nasty in the swamp areas though.
Micellaneous
PS4 version is sharper than both the PC and XB1 version (IQ is better there), that's even when the pc version's blur effect is set to off and sharpening is set to on.
PS4 versions suffers more in cutscenes due to the cap/double buffer and graphics API. There's camera and movement stutter in all versions. Geralt's movement feels more binary than analog. It's weird that Geralt leaps into a running animation (almost abruptly) rather than a smoother transition.
W3 2013 vs Release
Renderer completely changed from 2013. There was a denser lighting system, better G.I implementation, much richer particle system, far denser topology model in the 2013 version of W3 .
Features:
Originally Dx9 based but has adopted many Dx11 features
Tessellation is in (ground, rocks, bridges)
SSR on water
Weather effects
Global illumination emulation
Dof effects on camera due to weather
Game uses image based lighting
Also uses dynamic point lights which reflects on the environment's objects including geralts armor etc..
PC (970GTX) vs Consoles
Consoles run foliage on high
Ultra foliage on PC draws a greater distance but very minor from high in difference
Lighting and PP effects are at the highest setting on both consoles and PC (bloom, shafts of light, long shadows from objects etc..)
Weather system is equal across console and PC
AO and AF dialed back on consoles (some say due to a patch) in cutscenes
HBAO+ works well with foliage in W3 on PC, not worth it if you don't have an Nvidia card, generally costs more than it impresses.
Grass density medium quality on consoles
Terrain quality is on high with detail level being equal to PC
Particle effects are equal to the Pc's best.
Shadows never seem to change from any notable degree from medium to ultra.
Discrepancies (console side)
Consoles have lower shadow quality and AO than the lowest pc settings at times, but they're predominantly at medium shadow quality and makes use of ssao most of the time.
Textures are predominantly high on consoles with some mix of medium textures, ground textures are usually the culprit for medium quality.
Ultra textures are the same quality as high, but then ultra textures never go beyond 1.8GB of VRAM on PC. (which makes the use of medium textures at all very questionable)
Color depth look 16-bit rather than 32-bit on consoles
Texture loading is slower on consoles, even the integrated A10 R7 on chip gpu handles texture loading much better/faster with ultra textures in place.(smh, should not be)
Loading times are much faster on PC, at least 5 times faster. Sometimes consoles take over a minute to load in some instances. (not common)
Water effects are high on console which includes tessellated movement and interaction, but there's some shader work that's absent on consoles that even the lowest setting on PC has (smh at cdpr, clearly they needed more time with consoles)
Water also lacks the interaction with spells (like the force blast), which is apparent on PC at high settings, such effects are not apparent on consoles.
Note: He indicates that the consoles seem to have been behind in development due to some of the results we're seeing here, just like dying light was. He also indicated that CDPR is primarily a PC developer and PC was the lead platform for development.
He expects that consoles will see a nice boost in quality and performance in the coming patches since it was behind PC in dev for the most part. The texture issue may also be linked to a loading time issue which looks to be API related something which more time and experience can fix.
PS4 vs XB1
PS4 and XB1 has the same settings employed for the most part, however;
PS4 is 1080p vs the XB 900p (which is standard fare by now)
PS4 has 8xAF throughout whilst the XB1 uses a dynamic 4x-8xAF depending on the scene.
Water on XB1 seems slightly better, closer to PC but not quite there.
IQ on XB1 is worse due to 900p and the upscaled image issues with the custom fxaa and temporal aa implementation
Framerate is smoother on PS4 over the XB1. Xb1 framerate is unlocked to help support drops better due to the limited buffer space on XB1.
PS4 would normally drop the odd frame whilst loading new areas (loading issue) which exists on all versions, but since PS4 is capped you will see it dropping to 29fps whilst horseback riding here and there. (my opinion, when the xb1 version is capped it will drop more frames in these sessions, because the unlocked framerate would no longer keep it just barely at 30fps or above)
General impressions and Summary
Cutscenes vary wildly in performance
Hairworks is very unrefined at the moment just as tressfx was initially. Hairworks is not that impressive considering the 20fps performance penalty on average. Geralts hair looks better without it.
Tessellation should be set to at least 8x when hairworks is on (PC) to make it look good.
PhysX is on consoles (though emulated on the CPU)
Witcher 3 is NOT CPU limited, it is GPU limited however. It is NOT a CPU killer. All the testing PC's were almost always at 100% buffer usage (gpu wise). On the flipside, cpu usage even on the A10 hardly every taps a single core. All cores operate with lots and lots of room/cycles to spare. No core is ever maxed out, even paired with the entry level 750ti.
A dual core is enough for the game. A quad core is not entirely needed as it hardly uses over 4 cores.
750ti Shenanigans
The 750ti performs close to the consoles with similar settings, helped of course by its unlocked state in the comparison, however it does dip below 30fps more often than the PS4 and it performs just slightly better than the XB1 which is also unlocked.
750ti is also constantly below 30fps in the swamp area, other areas it holds up a bit better and is more comparable to the PS4. It's nasty in the swamp areas though.
Micellaneous
PS4 version is sharper than both the PC and XB1 version (IQ is better there), that's even when the pc version's blur effect is set to off and sharpening is set to on.
PS4 versions suffers more in cutscenes due to the cap/double buffer and graphics API. There's camera and movement stutter in all versions. Geralt's movement feels more binary than analog. It's weird that Geralt leaps into a running animation (almost abruptly) rather than a smoother transition.
W3 2013 vs Release
Renderer completely changed from 2013. There was a denser lighting system, better G.I implementation, much richer particle system, far denser topology model in the 2013 version of W3 .
Foliage Distance on Ultra makes a pretty noticeable difference. It's a heavy hit on performance but it's absolutely great to have if you've got the room for it. Makes a big impact in more open areas and especially when you're at an elevation looking over an area.
Also, HBAO+ is good not just for foliage, but for nearly everything. Indoor scenes, for instance, make great use of it, too. It's a big improvement.
And if we're talking GTX970 specifically, it's worth noting that you can play at higher-than-PS4 settings with a solid 60fps. Alternatively, you can play with the same settings, at 30fps, but at 1800p, providing an incredible level of IQ and downright transforms the foliage, almost as if they were replaced with higher quality assets.
Also, HBAO+ is good not just for foliage, but for nearly everything. Indoor scenes, for instance, make great use of it, too. It's a big improvement.
And if we're talking GTX970 specifically, it's worth noting that you can play at higher-than-PS4 settings with a solid 60fps. Alternatively, you can play with the same settings, at 30fps, but at 1800p, providing an incredible level of IQ and downright transforms the foliage, almost as if they were replaced with higher quality assets.
Interesting piece! The only points I might disagree with are...
The main difference in 2013 was that each scene was hand built (outside of an open world environment) with lots of hand placed lights and an attention to detail that wasn't possible on a large scale.
The game never featured a full GI system (ie - it doesn't model light bouncing). It didn't in 2013 and it doesn't now.
The main difference in 2013 was that each scene was hand built (outside of an open world environment) with lots of hand placed lights and an attention to detail that wasn't possible on a large scale.
I believe the developers have stated that tessellation is reserved exclusively for water simulation (well, and hair works, I suppose). Not 100% on this, though.
I really really doubt this. With the amount of color used in the game, 16-bit color depth would have produced severe dithering artefacts. I mean, 16-bit color was one of the biggest issues with GameCube games back in the day versus, say, PS2 which often stuck with 24-bit color instead.
That's obviously a matter of opinion but HBAO+ is always worth it, I feel. It makes a tremendous difference. That said, I haven't actually seen how it performs on AMD hardware so it may not be worth using there.
Yeah, I watched the video a few days ago and I don't remember NXGamer saying this. Furthermore, if the consoles were really using 16-bit colors, I'd expect to see a massive difference. Also, I get the feeling not everything in TheLastWord's post is what NXGamer said.
GoonerTRON
Banned
Agree. ULTRA to Foliage range/grass density with Depth of Field, HBAO and AA ingame and though Nvidia Control Panel make it look like a god damn painting in the distance.
Head.spawn
Junior Member
HBAO+ is damn near mandatory IMO.
If performance is an issue, turn something else down to deal with the performance. Kill your Ultra shadows, or lower your foliage range, or even grass density to accommodate.
If performance is an issue, turn something else down to deal with the performance. Kill your Ultra shadows, or lower your foliage range, or even grass density to accommodate.
This post deserves it's own thread. Fantastic summary, good video.
Aceofspades
Banned
He also usually gives more details than DF. So I agree with you.
Wouldn't surprise me in the least.
OrbitalBeard
Member
Denton
Member
All good points. Only regarding the bolded - if you lower tesselation factor in drivers, the ground will look much flatter. So at least the terrain uses a lot of tesselation. You can see the difference on the second ingame picture here:
http://pctuning.tyden.cz/multimedia...ky-hon-rozbor-hry-a-nastaveni-detailu?start=9
Move your cursor on the pic to see the switch.
Pretty stark.
PHOENIXZERO
Member
I'm watching a bit of it now and he says it at around the 7:52 mark regarding the shot he was looking at.
https://youtu.be/ibxiGr0V5iA?t=7m49s
Dictator93
Member
So much wrong in his analysis in spite of its length. Sigh... Just because you say a word... doesn't make it the fact.
It is actually a rather terrible video in terms of accurate content. Sorry.
Backfoggen
Banned
NX Gamer seeing phantom tessellation again. Also 16-bit color lol.
Sigh.
I don't understand why some here can't understand that the pc is irrelevant as far as consoles are concerned, the only place it is relevant is on forums and us forum dwelling individuals are in a vast minority we have to accept.
The PS4 and Xbox provide a hugely superior experience over their predessors, that's all that matters.
Aceofspades
Banned
On consoles PS4 is having a good advantage to X1.
Better Resolution
Higher AF
Smoother fps.
Better Resolution
Higher AF
Smoother fps.
DemonCleaner
Member
how about you replace that general shitover with actual arguments. kinda poor style of discourse, you know...
This. He deserves his own threads.
This is such a hyperbolic post. Look, you might not like him, but he's not wrong in everything. He even found subtle stuff DF didn't. Hell, even Dark said it was an interesting piece. So, please, people can rely on different sources without turning this into a system war like in many DF threads.
Dictator93
Member
Ready for this?
Nearly every single point posted by thelastword and number of smaller things in the video even.
As a starting point? The particle system is the exact same as before. Still forward shaded and everything... just different art!
There is no system wars in saying he is wildly inaccurate on a number of points (so so many that writing them out would take a long time).
jim-jam bongs
Member
alexandros
Banned
Good God, I sincerely hope that this is just thelastword's "interpretation" of NXGamer's findings.
Keep in mind that some of the stuff in TheLastWord's post is not what NXGamer said. Then again, I don't see the point in convincing you. You were extremely hostile towards NXGamer from the start. So, please, do yourself and everyone else a favour and stop whining about a source you find inaccurate.
I hope so. I don't remember him saying this.
Caayn
Member
That's the point where you realize it's not a serious post. Is it?
For those of us who didn't watch the video, what points did thelastword said differently from NXGamer? You've said this multiple times now but still haven't named an example.
blastprocessor
The Amiga Brotherhood
It's a bit of surprise but it's amazing what a bit of optimising can do. Don't think we're seeing the best the PS4 can do with 3rd party titles though.
For instance, TheLastWord seemed to make HBAO+ sound like it had little improvements. I forgot what NXGamer said, but he never put the improvements of HBAO+ as lightly as TheLastWord did. Also, the performance of the 750 Ti is how TheLastWord worded it.
Dictator93
Member
What is wrong with "whining" about a source that is inaccurate? If anything, challenging postulations is a positive thing. In this case, NXGamer is so often hyperbolic with his inaccurate claims that it clouds any discussion regarding console ports or console games. It is practically counterfactual at times for reasons that I have trouble understanding.
Two discourses develop:
1. where DF is apparently some PC/xb1 centric review site that knowingly ignores facts and presents PS4 version in bad light...
2. where NXGamer shows how the PS4 is somehow comparable to the highest end PC presets (or better according to taste) whilst fattening up his claims with technical wording.
The fact that we cannot agree on the factual empirical base is the most problematic point. The fact that one of these sources readily just uses inaccurate wording to describe the same content makes discussion so impossible at times.
alexandros
Banned
That is just the nature of multiplatform development. No platform is doing the absolute best it can do, including PC.
NX is good with video too. I wish I had a nice accent and a good speaking voice.
If you want to hear me curse like a sailor and become enraged, though, you can always watch this thing I did while trying to play Skynet for the first time last year. 90 minutes of Skynet, baby.
https://www.youtube.com/watch?v=mf3q7PNPKmg
If you want to hear me curse like a sailor and become enraged, though, you can always watch this thing I did while trying to play Skynet for the first time last year. 90 minutes of Skynet, baby.
https://www.youtube.com/watch?v=mf3q7PNPKmg