Chairman Tickles
Banned
Not enough games yet to constitute a trend. Could just be really crappy and sloppy ports to the xboner.
I'm not sure two Japanese games and a beta constitutes a trend.
Decreasing resolution to increase framerate has been a thing in 3D graphics since day 0 though.
I think everyone here, no matter what their platform of choice, fanboys and girls included that jaguar cpu is awful at this day and age and a major handicap in today’s gaming.
PS4 Pro is 8.4TF fp16 that's a fact but I'm not trying to twist it as if PS4 Pro is more power than Xbox One X it's just better at fp16 operations & have more ROPS. but even in these cases it has to be optimized to make sure it's not held back by the memory bandwidth.
I think everyone here, no matter what their platform of choice, fanboys and girls included that jaguar cpu is awful at this day and age and a major handicap in today’s gaming.
Do you even know what FP16 means ? Because as i read it you sound like you don't.
FP16 is half precision of FP32, you can use it for SOME tasks but you can't use it for ALL tasks.
You just can't "8.4TF" fp16 be any valuable metric especially for gaming console.
Whatever you like it or not XboxX has faster CPU and way stronger GPU so the only logical conclusion here is that devs didn't just spend a lot of time with this version or they outsourced job to someone else who wasn't up to task. Budget alone could be limiting factor because they also have to plan for amount of coppies they can reasonably sell.
The other conclussion could be that porting was targeted at slim version not pro where in case of pro they just lazely upressed it and done nothing else.
While it is true it is worthwhile to understand this is THE BEST cpu they could get around that time. There was simply no alternative to it.
Also CPU itself isn't that bad, i had a10 apu for my younger brother (4 core + gpu) and while it was about half the speed of my i5-3570K@4,5ghz it wasn't that slow either. It is just bad cpu when you compare it to top desktop cpus.
I know exactly what fp16 is and that's why I make sure to mention that the 8.4TF is fp16 & not just say 8.4TF by it's self. the fact is that now that it's becoming more useful on PC & consoles devs will find use for it when they can to get more out of the GPU's that have higher fp16 performance.
I know exactly what fp16 is and that's why I make sure to mention that the 8.4TF is fp16 & not just say 8.4TF by it's self. the fact is that now that it's becoming more useful on PC & consoles devs will find use for it when they can to get more out of the GPU's that have higher fp16 performance.
There's a reason that everybody laughs about FP16. Its not some super secret mega ultra weapon Sony will flop out one day that will make all your dreams come true. Its quite easy to think about really, but lets dumb it all the way down. FP16 is less the precision of FP32, hence, you can fill more FP16 operations in the same span as a FP32 operation.
So what makes this miracle amazing? Nothing.
For some things, its great. Things that don't need precision, things that end up being "random" by it's very nature. That's great. Maybe you want to create a noise filter for ambient occlusion before blurring? Awesome. But stuff that actually matters, you know, on the rendering side of things that's a heavy hitter, it suddenly changes. You Can essentially do some things well with FP16, but once you get to precise details, it loses its edge. Image your FP16 textures being compressed to hell and all macro blocked due to compression? Imagine your per object blur suddenly creating stochastic ghosting patterns. Essentially you lose all precision, because it just doesn't have it. Now imagine how many things in your game are rendered with some sort of precision, versus how many are rendered with randomness.
And there you go. Essentially, you end up having very little benefit to FP16, and it actually will take more effort to do for those couple of random situations that it will come in handy for, because now you lose all portability with other consoles.
So lets let the dream die in the same dustbin as the cloud, yeah?
F16 has nothing to do with randomness. Not sure where you are getting that idea from.
Not sure why people are getting defensive over exploring this discussion. You’re not saying anything inaccurate or controversial.
The pro gpu can do 2x the operations if the data is half the precision. If a whole gfx engine was half precision it would be twice as performant. In reality no engines are half precision, and some may only have certain parts in fp16 and gain only some efficiency.
The question arises; is fp16 code becoming a bigger part of the landscape as more hardware supports it, and likely next gen will too. A valid talking point based on the thread topic.
Fp8+8 sauce?Obviously some third party developers not fully utilizing the extra performance on X, but some are like Rockstar or Even ubisoft with division or ghost recon are good examples.
Absolutely no reason why a game on pro should perform better and I swear if someone mentions fp16 sauce...I hope your ps4 dies.
Well that is the very definition of Fp16 but what makes you think it's best used for randomness? Care to elaborate on that?I never said it was. You can use FP16 to render whatever you want, its just not as precise. Meaning that its function in this age is limited, and instead is best used FOR randomness.
Well that is the very definition of Fp16 but what makes you think it's best used for randomness? Care to elaborate on that?
If by randomness you simply mean BRDF vector calculations how does that not offer some gains for every PBR based game out there if the developer puts in the effort?
Just bad optimization
You're not thinking along the lines of what looks best, which is 9 times out of 10 what developers will shoot for. FP16 will, of course, render everything you need in a game. However floating point operations at such a low level will only do so much, and thus the game will look worse of for it. In theory, FP16, when combined with FP32, is best used for things where precision isn't a factor, and in this case, random floating point operations are of course a better use of the idea.
This this as by no means a decent comparison, but its best I can find on a quick search. Using old art work wont show the effect off at all, really, but it does still show it. Look for the specular highlights at the bottom of the page:
https://www.hwupgrade.it/articoli/skvideo/1013/radeon-x800-e-il-momento-di-r420_15.html
It would be interesting if they could do it on a game by game basis.
Like when you put an “enhanced” game in a me u would pop saying
Do you like Graphics?
Frame rate?
Or Balanced?
And it remembered your settings on a game by game basis.
He's a fanboy. Don't get into any argument with a fanboy. You are bound to be taken down a path of idiocy. I love games so I don't get the console war at all.What’s the beef with the fp16?
I see nothing mentioned about randomness being a good candidate there. You say it isn't well suited for today but that article you link shows that there is a massive performance hit for fp32. FP16 isn't some magic bullet but there is a shift to support rapid packed math more than ever today. New PC graphics cards have only started supporting it in the last 2 years or so. You are saying that there is no place for it in games today other than randomness but I think that is completely wrong. It can be used everywhere. cloth and hair physics, shaders (shader model 6.2) , and AI. If you want some actual examples of use:
https://gpuopen.com/first-steps-implementing-fp16/
A 10% performance gain is nothing to scoff at.
X actually runs FP16 at same speed than FF32... it is 6TFs for both.Waiting for the graphs that show
PS4 Pro fp16 = 8.4 fp32= 4.2
X One Xfp16= N/A fp32= 6.0
Just to really top this one off.
But he is talking about western devs too.This is the only correct answer.
I literally just said I couldn’t find what I was after but the pictures were correct.
I’m giving up all hope for those who can’t read or even understand the basics of this tech. A 10% increase on things that you can’t actually get many gains from anyway isn’t exactly worth the overall cost factor it would take to create and test different operations for different machines. You would be looking at, in real world performance terms in a real world setting, a <1fps gain.
Source: Actually having used, tested and worked with this in a realistic setting.
But let’s keep jumping through hoops about theoreticals etc,
There's a reason that everybody laughs about FP16. Its not some super secret mega ultra weapon Sony will flop out one day that will make all your dreams come true. Its quite easy to think about really, but lets dumb it all the way down. FP16 is less the precision of FP32, hence, you can fill more FP16 operations in the same span as a FP32 operation.
So what makes this miracle amazing? Nothing.
For some things, its great. Things that don't need precision, things that end up being "random" by it's very nature. That's great. Maybe you want to create a noise filter for ambient occlusion before blurring? Awesome. But stuff that actually matters, you know, on the rendering side of things that's a heavy hitter, it suddenly changes. You Can essentially do some things well with FP16, but once you get to precise details, it loses its edge. Image your FP16 textures being compressed to hell and all macro blocked due to compression? Imagine your per object blur suddenly creating stochastic ghosting patterns. Essentially you lose all precision, because it just doesn't have it. Now imagine how many things in your game are rendered with some sort of precision, versus how many are rendered with randomness.
And there you go. Essentially, you end up having very little benefit to FP16, and it actually will take more effort to do for those couple of random situations that it will come in handy for, because now you lose all portability with other consoles.
So lets let the dream die in the same dustbin as the cloud, yeah?
LOL
No more!
wtf
No more!
Actually, it is a form of macroblocking, though the phrase isn’t right. Banding will occurr but also compression-like artifacts, due to the precision being lower.
Also why would I provide benchmarks? I’ve wasted my time enough here. What I’m saying mirrors what everybody else who has worked with them has said, including digital fo........ oh wait now I see the issue.
This is 100% false, by the way.- DF’s main purpose is to make thier parent company money. They get the same YouTube consultants as anyone else who tell their parent company that they need these click bait titles , to go for controversy, and play to their viewers. So of course they are going to play up the ps4 “ performance “ advantage on a few titles even though they have lower settings and they didn’t play up the Xbox when the base Xbox had a string of lower settings but better performance games.
So is this. Like, holy shit.The thing to remember is that the guys doing the videos are not the guys making the decisions of what is shown and how it shown.
I think everyone here, no matter what their platform of choice, fanboys and girls included that jaguar cpu is awful at this day and age and a major handicap in today’s gaming.
Wow what a thread !
- no fp16 isn’t magic and doesn’t replace any graphical fp32 uses. At best you could use it for something small like lod distance calculations but even that isn’t going to gain you anything in performance gains.
- DF’s main purpose is to make thier parent company money. They get the same YouTube consultants as anyone else who tell their parent company that they need these click bait titles , to go for controversy, and play to their viewers. So of course they are going to play up the ps4 “ performance “ advantage on a few titles even though they have lower settings and they didn’t play up the Xbox when the base Xbox had a string of lower settings but better performance games.
The thing to remember is that the guys doing the videos are not the guys making the decisions of what is shown and how it shown. It’s an array of higher ups , number crunchers , and shitty consultants.
You know the only reason RTX 2060, Vega 56 and Vega 64 performs better than GTX 1070TI in FarCry 5 is FP16, no? Every other game has at least GTX 2060 and Vega 56 below GTX 1070TI.Wow what a thread !
- no fp16 isn’t magic and doesn’t replace any graphical fp32 uses. At best you could use it for something small like lod distance calculations but even that isn’t going to gain you anything in performance gains.
Ya know what really sucks? Extremely fanboys thinking you hate their preferred platform. Xbox fans think we hate Xbox, PS4 fanatics (ie - thelastword) think we hate PS4 and Nintendo fans think we're too hard on Switch. Don't get me started on the shit we get from AMD fans. Ya'll are crazy and make things a lot less pleasant than they should be.
Also, I understand people get attached to companies and plastic boxes, I've been there in the past but it's worth trying to get past it and embrace everything. Learn to love them all. I certainly do now.
This is 100% false, by the way.
So is this. Like, holy shit.
Every video I make is made in isolation from my house. What's said is what I wrote - nobody is deciding what's in there but me.
If higher-ups ever tried to do this, I'd leave in a second. That I get to basically make whatever I want is one of the things I enjoy most. The situation you describe would be HORRIBLE.
I'm absolutely saying that, yes.So you are saying there are no consultants involved and that you come up with the YouTube titles?
So you are saying there are no consultants involved and that you come up with the YouTube titles?
You know the only reason RTX 2060, Vega 56 and Vega 64 performs better than GTX 1070TI in FarCry 5 is FP16, no? Every other game has at least GTX 2060 and Vega 56 below GTX 1070TI.
That alone make your claim false. Wait more games using FP16 to compare performance running in Vega or RTX.
Wow what a thread !
- no fp16 isn’t magic and doesn’t replace any graphical fp32 uses. At best you could use it for something small like lod distance calculations but even that isn’t going to gain you anything in performance gains.
- DF’s main purpose is to make thier parent company money. They get the same YouTube consultants as anyone else who tell their parent company that they need these click bait titles , to go for controversy, and play to their viewers. So of course they are going to play up the ps4 “ performance “ advantage on a few titles even though they have lower settings and they didn’t play up the Xbox when the base Xbox had a string of lower settings but better performance games.
The thing to remember is that the guys doing the videos are not the guys making the decisions of what is shown and how it shown. It’s an array of higher ups , number crunchers , and shitty consultants.
...but we've never talked to any consultants, that's the thing. We've never used them before and have no plans to.I agree with you on the Youtube content consultants, these people are killing the thing that made the platform interesting to begin with (thankfully some people don't follow their advice). But I don't think that they are deciding that goes in and does not go in the video in details it's too much and if you ever worked with consultants you'd know that they don't go in details like that. They give them directions for video titles, push them to stir controversy since shocked people are more engaged, etc. and the guys at DF follow their advice because they want to put bread on the table, while doing something they apparently really are passionate about.