Support NeoGAF

[Matt]

They both punch at exactly their weight.

 

[Luckydog]

That's not punching above its weight, that's punching exactly its weight. The discussion was about comparative performance given application of a new technology. The numbers are adjusted for comparison with a different metric, which is why I said "effective power" and not actual. Note also that PS4 Pro is not the only system using RPM, so the same calculations would apply to other GPUs as well.

My results could easily be wrong--they're based on a range of estimates by other posters--but they don't indicate a misrepresentation of the hardware.

The mental gymnastics are magical....It's why I can't bring myself to block people as these are some of the best joke posts on gaf.

 

[borges]

This is an artifact of the discussions' nature, not a strategy on my part. It's simply much more common on GAF for Pro's technology base to be underestimated; CBR in particular is a magnet for misinformed criticism. A little contemplation will show this to be true. When was the last time you saw a post about how the Pro GPU is a marvel of punch-above-its-weight technology? Probably last year, before folks started seeing the results.

On the flip side, some posters do undersell what One X is capable of. However, this is almost never based in technical claims, just in projections about "no exclusive games" or "devs won't bother". I don't think those things are correct either, but my knowledge of the hardware isn't going to change their minds. (And there are plenty of folks willing to argue with them besides.) So I don't usually respond.

All that said, I have corrected people who lowballed One X's tech, and people who overestimated PS4 Pro's. It just doesn't happen as often. And many of my downward adjustments on Pro are in my enhancements thread, where they're part of lists and not lengthy discussion that everyone reads. For example, many sources on the web report The Division as being native 4K on Pro. As far as I know, I'm the only one who has analyzed it as only 1660 with stochastic reconstruction.

Man, nothing personal, but when so many non connected people tells you that your responses looks like they follow a particular pattern... its because, at the very least, looks like if they follow such pattern.
I cant talk about your attitude on Playstation related threads, because I rarely access to them, but yeah, on xbox threads, you are always correcting people in order to downplay the thing. Its not that you said lies or sth, but you always tend to correct those posts that make Xbox One X better, and systematically omits those were the X is irrationally shitted.
But again, you are free to do it, and you always talk to others very respectfully. Maybe its just a matter of being a bit more balanced on your answers.

 

[Liliana Eileen]

This news should quell the naysayers and "concerned" people. Good stuff all around. Wasn't planning on getting a 1X but now I may have to.

 

[Negotiator]

Yeah I get what you are saying but just like consoles hold back PC, I am talking more about geometry which is the basis of the game. The number of triangles, of the environment, the characters etc.

X's difference from ONE is really such a big one that it could be pushing far more geometry-rich worlds.

This isn't such a big issue anymore:

I mean, even the relatively underpowered Switch can render high-poly models in 1080p60 (Splatoon 2) these days. There's a reason no one advertises polygon numbers anymore (remember the glorious PS2/GeForce 2 era?). We're talking about billions of polygons in modern GPUs.

A generational leap is defined by 3 factors: new CPU microarchitecture, better/faster GPU, vastly bigger and faster RAM.

PS4 Pro has a better/faster GPU compared to OG PS4, CPU microarchitecture is the same (with a mild overclock) and RAM is basically the same quantity, only slightly faster.

X has a vastly better/faster GPU compared to OG XB1, CPU microarchitecture is the same (with a mild overclock as well), RAM quantity is 50% more and it's a lot faster compared to DDR3 (not that much if you factor in eSRAM for framebuffer usage).

Both of them do not qualify as a "generational leap". That's why they're called iterative consoles. We usually get 8-16x RAM and 10x GPU horsepower. Even the lowly Jaguar qualifies as a generational leap compared to Cell/Xenon (PPU).

In other words, when the true next gen consoles arrive (PS5/XB2), Pro/X will seem really weak.

Imagine if Sony/MS released PS3.5 and XBOX480 with 1GB of RAM, 4 GHz in-order PPC CPUs and faster GPUs. They'd be weak compared to OG PS4/XB1.

And last but not least, raising the production values immensely won't happen forever and ever. Don't take this for granted. AAA games are already expensive enough to make, even with the low OG PS4/XB1 baseline/dev target. Ask yourself why ND probably won't release more than 2 games this gen (3 if we're lucky) and why Rockstar hasn't made a next-gen GTA yet.

Can you imagine AAA games targeting 15-20 Teraflops of GPU power (more elaborate assets/textures will require more work) and 8-12 Ryzen cores with more elaborate AI/NPC programming? I'd love to be wrong, but this isn't sustainable. I'm afraid we're gonna see more delays and video game companies going bankrupt in the PS5/XB2 generation.

 

[Liabe Brave]

I don't know, I think most people would see calling a 4.2TF GPU a 5.6TF GPU as punching above its weight.

Here's your misprision, I did not say the Pro's GPU was 5.6TF. What I said was, if a technique can be deployed on a 4.2TF GPU that raises its performance by 35%, then it would effectively be able to achieve the same results as an otherwise identical 5.6TF GPU without that technique. There are complications to this story--see below--but that basic logic seems solid to me. Where do you think it breaks down?

Unless I'm mistaken, I thought you had people, who are in the know, tell you that you were wrong on your assumption.

My memory is that no one argued against the math. Several people did argue that FP16 was unlikely to see much use on Pro, either for technical or business reasons. And that the advantage of RPM is unlikely to matter in real-world scenarios, where other bottlenecks would restrict Pro before FP16 could be much help. I'm perfectly willing to believe those things. As I said, the estimates I gave could easily be wrong, for these reasons or others. The idea of giving such an estimate, though, is defensible, provided the appropriate caveats are included.

I think you're sincere when you strive for accuracy but I'm not sure you do it objectively. Nothing really wrong with that, we all have preferences.

I do not claim to be wholly objective, because I think that's an impossible task for the human brain. However, I strive to take my predilections into account when constructing ideas, and I appreciate you bringing these potential lapses to my attention.

They both punch at exactly their weight.

Precisely. "Punching above its weight" to me means a claim that particular hardware can achieve results not possible on other hardware of the same construction, usage type, and performance level. This fundamental self-contradiction is why we can definitively say it never happens.

Man, nothing personal, but when so many non connected people tells you that your responses looks like they follow a particular pattern... its because, at the very least, looks like if they follow such pattern.

Consensus is not an infallible guide to truth. That said, of course it gives me pause when there's such a chorus of criticism. Though I'll continue to ignore the idiots who think "scoring points" against people is the purpose of the forum, I'll attempt to revise my approach based on the more constructive responses from you and others. Thank you for being civil.

Its not that you said lies or sth, but you always tend to correct those posts that make Xbox One X better, and systematically omits those were the X is irrationally shitted.
But again, you are free to do it, and you always talk to others very respectfully. Maybe its just a matter of being a bit more balanced on your answers.

If there's something I'm heartened by, it's that the criticism has not been about wide swaths of unexamined errors in my posts. I'm sure I've made mistakes on GAF, because of course I have. But I'm glad to know that at least a concerted effort toward truth is evident in what I write.

In the past, I've ignored dumb criticisms of Xbox because they're usually not about the details that I find so fascinating. They're widely and vaguely stupid instead, and any number of other posters can effectively refute them. I'll take a different tack in the future.

I must say that I'm wary about general requirements for "equal time", since in some cases one side may be irrevocably mistaken. But I do understand that conversation is not pure logic, and community is not dry law. I mean to improve.

 

[thill1985]

Here's your misprision, I did not say the Pro's GPU was 5.6TF. What I said was, if a technique can be deployed on a 4.2TF GPU that raises its performance by 35%, then it would effectively be able to achieve the same results as an otherwise identical 5.6TF GPU without that technique. There are complications to this story--see below--but that basic logic seems solid to me. Where do you think it breaks down?

...the part where we ignore the fact that a 5.6TF machine actually can reliably perform at 5.6TF as opposed to a machine we pretend can tap into the application of a novel technique used incredibly sparingly in modern games and results in a 35% flops boost out of nowhere.

Precisely. "Punching above its weight" to me means a claim that particular hardware can achieve results not possible on other hardware of the same construction, usage type, and performance level. This fundamental self-contradiction is why we can definitively say it never happens.

You are choosing to define a term in a way that it can never be achieved (by design on your part). Every person here knows what the term means colloquially and you sure as shit realize that is the context it has been used in in the thread's discussion. Here you are just attempting to change the context so you get the comfort of avoiding admitting something that is inconvenient.

 

[Liabe Brave]

...the part where we ignore the fact that a 5.6TF machine actually can reliably perform at 5.6TF as opposed to a machine we pretend can tap into the application of a novel technique used incredibly sparingly in modern games and results in a 35% flops boost out of nowhere.

You are not following the conversation. The 35% boost would not occur "out of nowhere", it would be generated by a novel hardware feature. Also, the sparing use of a technique now does not forever limit its usage. Furthermore, the reliability of a "feature-less" 5.6TF GPU versus a feature-driven "effective 5.6TF" GPU isn't germane to the idea. It would merely adjust the metric, by more or less given whatever averaging timeframe you imposed and by constancy of the effect.

Finally, and more fundamentally, you seem to be laboring under the mistaken impression that I'm the one who estimated that 35% boost, and that I'm married to it. Neither thing is true. In fact I'm quite willing to believe that in reality, the boost will be 0%, for any number of reasons.

You are choosing to define a term in a way that it can never be achieved (by design on your part). Every person here knows what the term means colloquially and you sure as shit realize that is the context it has been used in in the thread's discussion. Here you are just attempting to change the context so you get the comfort of avoiding admitting something that is inconvenient.

You're partially right here. Constraining the definition this way is artificial, and doesn't represent how people use the phrase. It's counterproductive for me to jargonize a common saying. The actual usage is looser, more like "surprisingly powerful given only a surface appraisal". I apologize for the error.

However, you're wrong in your ridiculous armchair psychology, and in the way this more commensurate definition affects the conversation. Under the better definition, RPM may now qualify as a claim of "punching above its weight". But the changed definition also makes such claims possibly true instead of inevitably false. So there's still no fundamental issue with what I presented (though as stated above, in actuality no such benefit may accrue).

Of course, none of this addresses the actual point of argument, that factually wrong claims about PS4 Pro hardware are common. A single tentative claim by me, of indeterminate truth, is hardly germane.

 

[KageMaru]

Here's your misprision, I did not say the Pro's GPU was 5.6TF. What I said was, if a technique can be deployed on a 4.2TF GPU that raises its performance by 35%, then it would effectively be able to achieve the same results as an otherwise identical 5.6TF GPU without that technique. There are complications to this story--see below--but that basic logic seems solid to me. Where do you think it breaks down?


My memory is that no one argued against the math. Several people did argue that FP16 was unlikely to see much use on Pro, either for technical or business reasons. And that the advantage of RPM is unlikely to matter in real-world scenarios, where other bottlenecks would restrict Pro before FP16 could be much help. I'm perfectly willing to believe those things. As I said, the estimates I gave could easily be wrong, for these reasons or others. The idea of giving such an estimate, though, is defensible, provided the appropriate caveats are included.


I do not claim to be wholly objective, because I think that's an impossible task for the human brain. However, I strive to take my predilections into account when constructing ideas, and I appreciate you bringing these potential lapses to my attention.

I didn't misprision anything. You came up with those numbers and some latched onto it later in that thread. I also think you're misunderstanding where those stats even came from. IIRC the 35% performance improvements weren't for the entire GPU but for the specific function being processed, so your math still doesn't add up. I don't care to go through that whole thread but yes I'm pretty sure people did shoot down your claims for the potential performance. I'm also not talking about the people who are too eager to dismiss any benefits FP16 will bring. If I get the time I can comb over it, but that just seems rather petty at this point.

I think FP16 will certainly have benefits and I'm eager to see how developers utilize the feature in future games. However this conversation we're having in yet another Xbox thread about the Pros potential is silly.

 

[onQ123]

With all that different TF values flying around (FP32, FP16) it would be nice to have a more reliable benchmark standard for consoles to make them comparable on their production output. I know a dream never will come to fruition.


The intention of those numbers was to give Devs a feeling how titles with different kind of game engines will scale with the Xbox One X and what they can expect with their own games. Thats why it was shown to devs only. It is by far not a pattern how games will actually use the performance increase.

Life was so much better when we didn't know what the hell bits was but we just knew that we wanted more of them.

 

[Head.spawn]

It's a cool initiative although I would have bought the game sooner if it was available to buy separately on PC, since play-anywhere titles are slow to go on deep discount (unlike most PC games).

Gears 4 hit $30 in December last year. Forza Horizon 3 did as well IIRC. Not too shabby.

Recore was also half-off around that time.

They aren't afraid to do nice sales in the Fall.

 

[hesido]

The intention of those numbers was to give Devs a feeling how titles with different kind of game engines will scale with the Xbox One X and what they can expect with their own games. Thats why it was shown to devs only. It is by far not a pattern how games will actually use the performance increase.

Surely, but still, I wish 4K didn't suck all that power. A nicely anti aliased 1080p or 1440p image looks really good, upscaled to 4K right in the set too. But I think most devs will choose to not increase shader fidelity, it's the easy way out to spend the extra GPU cycles (moar pixels)

 

[Liabe Brave]

You came up with those numbers and some latched onto it later in that thread.

No, this is not correct. As I stated multiple times in that thread, the 35% figure came from a developer on Beyond3D (see below). The lower bound 10% is also not my estimate, though I can't recall where it came from any longer. Posters with more technical knowledge than me concurred that FP16 use was widespread, even with PS3, making the figures plausible. I couldn't find any of these tech-minded folks categorically denying the figures I repeated. This quote, for example, sounds like general agreement (though understandably they doubted the high end would ever be practical).

I've heard different figures, and 35% among them is on the higher side of possible gain. I bet that it would be less on pratice, around 15-20%.

I also think you're misunderstanding where those stats even came from. IIRC the 35% performance improvements weren't for the entire GPU but for the specific function being processed, so your math still doesn't add up.

You're correct. It was said later in the thread that 35% was only for pixel shaders, not the whole engine's performance. Do you know who posted that correction? Me. And to further prove that I wasn't forcefully pushing for the highest possible number, here's this quote:

More important to the impact is the question of how much improvement can a game see by using it--1%? 10%? 30%?--and how will that manifest?

I don't care to go through that whole thread but yes I'm pretty sure people did shoot down your claims for the potential performance. I'm also not talking about the people who are too eager to dismiss any benefits FP16 will bring. If I get the time I can comb over it, but that just seems rather petty at this point.

It's not petty, it's the difference between your criticism of me being true or false. Do people with real knowledge directly refute my logic? (It doesn't matter if they disagreed with specific numbers, because as I've shown above they're not my own estimates, and I wasn't promoting any particular value.) I've gone back and read a sizeable portion of the thread, and I didn't see anything I'd characterize that way. But of course I'm not the best judge there, so if you do note anything I'd like to see it. Unexamined error doesn't correct itself.

What I did see reading the thread was a lot of statements by me praising the One X and clearly stating its superiority.

Scorpio is still the most powerful console ever made, and should perform better than Pro almost all the time. It's well-engineered hardware, though it doesn't seem to be quite as specialized and unique as some thought.

Yeah, the DVR sounds very well designed and executed. This is one area where Scorpio will undoubtedly have a very big advantage over Pro.

But of course even this high end is lower than Scorpio. It seems impossible for the gap to truly close; Microsoft will always have an advantage.

From Info thus far Scorpio is a native 4k console with 900p and 1080p xbox one game engines the majority will be.. Ps4 pro the opposite majority will be Checkerboard and few will be native 4k but without 4k assets or ultra settings. FP16 won't change that we can agree on this right?

We haven't actually seen how many games will be native 4K on Scorpio, but Microsoft's claims don't seem implausible. So with all this you're probably right.

For something I supposedly don't do at all, I seem to do this a bit. I've also been accused of ignoring overweening claims for PS4 Pro, or undersold claims for Xbox One X, and not correcting those posters. Yet:

No FP16 hurts this [Xbox One X] console's ability to keep up long term.

"Keep up" is wrong, Scorpio will always be more powerful than PS4 Pro. Its lack of packed math, if anything, may hurt its ability to stay as far ahead.

(As it turns out, Collingwood meant something different, but clearly I didn't know that when I responded.)

I guess this is a PS3 vs Xbox360 scenario, where games coded by Sony could possibly be heavily optimized to use FP16 and in the end make Pro seem more powerful but with multiplats engines where FP32 is more common Scorpio will always turn out on top. No?

No, absolutely not. Nothing we know indicates that Pro could ever match the Scorpio in power, no matter how much devs utilize the special hardware it has. There's just too large a gap to overcome.

 

[Gestault]

I don't know, I think most people would see calling a 4.2TF GPU a 5.6TF GPU as punching above its weight. Unless I'm mistaken, I thought you had people, who are in the know, tell you that you were wrong on your assumption. I think you're sincere when you strive for accuracy but I'm not sure you do it objectively. Nothing really wrong with that, we all have preferences.

This is really politely put.

 

[ShadowkillXNA]

Life would be better if posters who present reasoned and objective arguements werent villified/ridiculed simply because some think those arguments dont kiss the ass of their favorite box of consumer electronics as much as they would like.
On topic: I'd like to see some one x battlefront 2 footage before the game launches but I think the marketing agreement will prevent that.

 

[onQ123]

Life would also be better if posters who present reasoned and objective arguements werent villified/ridiculed simply because some think those arguments dont kiss the ass of their favorite box of consumer electronics as much as they would like.
I'd like to see some one x battlefront 2 footage before the game launches but I think the marketing agreement will prevent that.

I'm not really sure what you're getting at but ok

 

[Syrus]

Why...the hell are we on this FP 16 and 32 crap again.

 

[onQ123]

Why...the hell are we on this FP 16 and 32 crap again.

I have no idea & they can't blame me this time & hopefully they don't find away to ban me for the 3rd time because someone else brought up the subject.

 

[Colbert]

I have no idea & they can't blame me this time & hopefully they don't find away to ban me for the 3rd time because someone else brought up the subject.

Indeed this time you are innocent. Keep it that way.

 

[onQ123]

Indeed this time you are innocent. Keep it that way.

I was innocent last time I was banned for replying to people who brought up the subject.

 

[Caayn]

I was innocent last time I was banned for replying to people who brought up the subject.

I know that you and I don't always seem eye to eye.

But sometimes it's best not to reply It's one of the reasons why I stay away from certain discussions and don't engage in the cesspool called the "next-gen racing graphics" thread. For me it increased my enjoyment of Neogaf.

 

[Liliana Eileen]

I was innocent last time I was banned for replying to people who brought up the subject.

To be fair you have derailed many a topic.

 

[onQ123]

To be fair you have derailed many a topic.

Yes only I can derail speculation threads by speculating

 

[McFadge]

This isn't such a big issue anymore:

I've always disliked that image.

There is something to be said for diminishing returns, but this represents a lazy use of 'increasing' polygon counts. It is far more likely the right-most model was made, then had its polygon count reduced, as opposed to utilising a higher polygon count to create a better result.

 

[FZW]

Yes only I can derail speculation threads by speculating

it seems like you are derailing this thread right now

 

[Liliana Eileen]

I've always disliked that image.

There is something to be said for diminishing returns, but this represents a lazy use of 'increasing' polygon counts. It is far more likely the right-most model was made, then had its polygon count reduced, as opposed to utilising a higher polygon count to create a better result.

Isn't it still saying the same thing, though?

 

[Gestault]

Isn't it still saying the same thing, though?

They're saying the previous example image is exagerrated because the right-most model is a mesh with more divisions, but not more model detail. The triangle example shows an exagerrated (but accurate) way that same idea could be taken to the extreme.

 

[hiphopcr]

I think the point is that better graphics are so much more than poly-count and pixels on screen. More power means higher res textures, more effects, better lighting, etc.

 

[Rockeeo]

Could GTA4 at 60 fps be possible on the Xbox One X?

When it was first released as BC, it ran with an unlocked framerate. It seems to be running at 40~ fps on my Xbox One S.

Thanks.

 

[Erimriv]

I'd love to see AC Unity patched to 4k and locked 30 fps. Currently playing it on XB1 and it's beautiful. Did this game get any improvements on PS4 PRO?

 

[onemanasylum]

To be fair you have derailed many a topic.

Pot/Kettle......

 

[Ehker]

I'd love to see AC Unity patched to 4k and locked 30 fps. Currently playing it on XB1 and it's beautiful. Did this game get any improvements on PS4 PRO?

They went back to patch Syndicate, but maybe felt Unity is too old. Hopefully with XB1X out they'll feel more motivated for doing it.

 

[MaLDo]

I've always disliked that image.

There is something to be said for diminishing returns, but this represents a lazy use of 'increasing' polygon counts. It is far more likely the right-most model was made, then had its polygon count reduced, as opposed to utilising a higher polygon count to create a better result.

Agree. Going from 6000 polys to 60000 polys you can improve the countour of a high poly nose on one character or increase the number of characters on screen x10. The former is an example of diminishing returns, the latter is an example of a generation jump.

 

[BAW]

I'd love to see AC Unity patched to 4k and locked 30 fps. Currently playing it on XB1 and it's beautiful. Did this game get any improvements on PS4 PRO?

Much smoother performance using PS4 Pro's Boost mode. No proper Pro patch though.

If indeed unpatched XBO games can use 3 teraflops in X1X, the performance in AC Unity should be even smoother, but probably still hardlocked to 900p and 30 fps.

AC Syndicate has a much higher chance of getting a X1X patch, since it got a proper Pro patch.