• 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]PS5 uncovered

ethomaz

Banned
2fe138b98c283623fcd2b6f3094c3735.jpg
Are you not in the Xbox discord?
They left a fan out lol
 
Last edited:

joe_zazen

Member
Of course he's not, he's part of the reason people were on this 13 teraflop "PS5 is more powerful" hype train to begin with. He's part of the reason there's so much disappointment, he helped setup the unreal expectations which didn't land.

Matt is fake af, he's a bullshit "insider", a nobody with a history of totally inaccurate information. This variability trash takes the cake though.

dude, matt has always said he believed xsx was more powerful. He was one of the only insiders to say that consistently.
 
Last edited:

ethomaz

Banned
Well, just like everybody else then.

But seeing your crowd go absolutely crazy and desperate with the special sauce narrative and the damage control pushed to limits never known before, has been a beautiful shitshow to watch.
I see the complete opposite.

Xbox fans trying everything crazy to try to make PS5 bad with FUD.

PS fans are happy and excited with PS5.
 
Last edited:

Proelite

Member
Agnostic2020 and B BlackBuzzard

I am perma-banned from that thread.


No, I don't make stuff out of thin air, what's the point? I tried to warn everyone about Tommy Fisher and OsirisBlack.

 
Last edited:

martino

Member
Well, just like everybody else then.

But seeing your crowd go absolutely crazy and desperate with the special sauce narrative and the damage control pushed to limits never known before, has been a beautiful shitshow to watch.

Conspiracy theory is a new low for me but i wasn't there last time , was ms community doing same kind of crap ?
 

joe_zazen

Member
Well, just like everybody else then.

But seeing your crowd go absolutely crazy and desperate with the special sauce narrative and the damage control pushed to limits never known before, has been a beautiful shitshow to watch.

you have low standards, or maybe you wernt around for wiiu or x1. I am sorry i missed the ps3, that was probably a thing of beauty. This go round? Not much meltdown or crazy.
 

ethomaz

Banned
Agnostic2020 and B BlackBuzzard

I am perma-banned from that thread.


No, I don't make stuff out of thin air, what's the point? I tried to warn everyone about Tommy Fisher and OsirisBlack.

Is it related to what you made up on discord?


 

Proelite

Member
Last edited:

joe_zazen

Member
Conspiracy theory is a new low for me but i wasn't there last time , was ms community doing same kind of crap ?

dont know about the community thing, but there were several mass deletions of social media marketers in 2013 here. You could see their posts from a mile away. Maybe more sophisticated now? Shrug.
 

martino

Member
I'm sure you do, buddy.

I'm sure you do.



Well, in the early days of this gen the MS community were acting the same as the Sony community is acting right now.

Turns out both communities are very similar.
your answer is not really a surprise :D
 

Psykodad

Banned
Well, just like everybody else then.

But seeing your crowd go absolutely crazy and desperate with the special sauce narrative and the damage control pushed to limits never known before, has been a beautiful shitshow to watch.
My crowd?

What makes you think I'm that invested with Playstation? The stuff I say?
Most of it I don't even care all that much about.

The only thing I care about are the improvements over PS4 and going by the info on PS5, the jump is more than enough for me.
 

ethomaz

Banned
I'm sure you do, buddy.

I'm sure you do.



Well, in the early days of this gen the MS community were acting the same as the Sony community is acting right now.

Turns out both communities are very similar.
Not surprise to be fair from your way to see the things.
PS fans cares more about PlayStation so PS5 is really hyping us for next-gen.
Of course we get mad when some Xbox fans lies about PS5 things but that is normal.

Just look at that thread with people even cutting phases to match the bad narrative.
 
Last edited:

caffeware

Banned
- Variable clock was choose due how difficult was to cool PS4 so that allowed a easy and effective cooling system in PS5.... he says you will be surprised here.
- The clock is not defined by temperature but workload so you won’t have unexpected performance drops... the opposite the performance is pretty consistent.

:unsure::unsure::unsure:
 

LordOfChaos

Member
- Variable clock was choose due how difficult was to cool PS4 so that allowed a easy and effective cooling system in PS5.... he says you will be surprised here.
- The clock is not defined by temperature but workload so you won’t have unexpected performance drops... the opposite the performance is pretty consistent.

:unsure::unsure::unsure:


It's not too hard to understand the eli5 version: They're flipping the traditional model around, instead of constant frequency, it's constant power.

The cooling design knows that the chips will always output a constant power, which makes designing the cooler around it easier than the old way which had fixed clock speeds with variable power.

It's also unlike PC SmartShift because it's not temperature bottlenecked.

Can't wait to see the cooler, the PS4s was pretty uninteresting, lets see if they outdo the OG PS3 monstrosity.

ps3-fat-ceche01-fan-heatsink_1_fd56cac5a706aa3eb6d60ca193b7d838.jpg




They can still however say performance is consistent because it's deterministic, the same code run a thousand times will perform the same way each time, unlike PC temperature based boost, because again it's instead power based.
 
Last edited:
dude, matt has always said he believed xsx was more powerful. He was one of the only insiders to say that consistently.
Give me break with this revisionist history, I've been through it with people and this nonsense. Matt has been all over the board and he was in cohorts with that Kleefan guy.

The shtick is up, move on.
 

Jonsoncao

Banned
Quoting Mark Cerny

So instead of using the temperature of the die, we use an algorithm in which the frequency depends on CPU and GPU activity information. That keeps behaviour between PS5s consistent.

If it is algorithmic not hardware based (temperature), any system engineers might wanna elaborate how an SoC can avoid the Schrodinger's Frequency problem in modern hardware? (if you profile the performance algorithmically, it consumes resources significantly sometimes, which slows down the performance)

For example, when I program in PyTorch for deep learning, I normally turn on the profiler when do initial testing, after figuring out the bottlenecks and successfully tweaked them, I turn off the profiler and run again (which gets 20% performance boost per epoch, the boost will be less significant if I use less iterations per epoch).
 

ethomaz

Banned
Quoting Mark Cerny



If it is algorithmic not hardware based (temperature), any system engineers might wanna elaborate how an SoC can avoid the Schrodinger's Frequency problem in modern hardware? (if you profile the performance algorithmically, it consumes resources significantly sometimes, which slows down the performance)

For example, when I program in PyTorch for deep learning, I normally turn on the profiler when do initial testing, after figuring out the bottlenecks and successfully tweaked them, I turn off the profiler and run again (which gets 20% performance boost per epoch, the boost will be less significant if I use less iterations per epoch).
There is a unit inside the APU just to decide that he said.
 
The amount of FUD you guys will believe is incredible. That guy was called out, then deleted those tweets that Jason Schrier pointed out. Must be getting worried by how next gen is going to turn out, huh?
SSD is no longer the buzzword of the week, FUD has now taken over.

Keep up the good work on being original.
 

THE DUCK

voted poster of the decade by bots
This article just proves that there are issues with speed when the system is pressed and the system really is 9 Teraflops as originally shared. Only die hard Sony fans would run in here stating this puts anything to bed.
Still a powerful machine in it's own right that will have some amazing games, just can't measure up to the series X. I will buy both as soon as budget permits.
 
Last edited:

MarkMe2525

Member
In all seriousness, I'm surprised at the demonstration they ran with the different cu counts. It literally told the exact opposite story of what Cerny said would happen when going with higher frequencies compared to the same tf over more cu's. We were told that running 36 cu's at higher clocks would provide better performance than more cu's clocked lower at any given TF. The 40 cu lower frequency gpu had the advantage vs the 36 cu gpu clocked higher.
 
Saw this on NPC Era, thought it was worth a quote:


Soprano said:
Just a little notice. The concern and FUD you are seeing around PS5 news lately is being done on purpose. Some of it is planned in a certain discord.

This thread heading to the dumps is apart of it i presume. There's this sad attempt to make PS5 look like it's going to fail and experience another PS3 generation.

That is all. Bye.
I dont normally do these conspiracy theories but in this case there has been a noticeable influx of console warriors in PS5 relates threads ever since the reveal.
I mentioned this before in another PS5 thread not long after the reveal.
I know I'm not the only one to notice this trend.
 
This article just proves that there are issues with speed when the system is pressed and the system really is 9 Teraflops as originally shared. Only die hard Sony fans would run in here stating this puts anything to bed.
Still a powerful machine in it's own right that will have some amazing games, just can't measure up to the series X. I will buy both as soon as budget permits.

After everything we heard on the 18th and today and somehow you still ended up being wrong. SMH.
 
Last edited by a moderator:

Panajev2001a

GAF's Pleasant Genius
I dont normally do these conspiracy theories but in this case there has been a noticeable influx of console warriors in PS5 relates threads ever since the reveal.
I mentioned this before in another PS5 thread not long after the reveal.
I know I'm not the only one to notice this trend.
The Kenpachii, KageyK, The_Mike, DynamiteCop!, Rightisright, Gavon West, etc... pack seem to fly very very close together...
 
Last edited:

StreetsofBeige

Gold Member
In all seriousness, I'm surprised at the demonstration they ran with the different cu counts. It literally told the exact opposite story of what Cerny said would happen when going with higher frequencies compared to the same tf over more cu's. We were told that running 36 cu's at higher clocks would provide better performance than more cu's clocked lower at any given TF. The 40 cu lower frequency gpu had the advantage vs the 36 cu gpu clocked higher.
Of course Cerny is going to say blasting a gpu at 2.23 ghz is better vs. going with more CU at lower clocks.

He's supposedly the "lead architect" and he saddled PS5 with 36CU (the same amount as PS4 Pro in 2016).

What he should have done is do the MS route of more CU and less gpu ghz. Find the right combo and PS5 could be 12tf too along with the super fast SSD.
 
I must have watched a different video to alot of people here. The one I watched said numerous times that the CPU would clock lower to make sure that the GPU can maintain the 2.23 speed.
 

Gavon West

Spread's Cheeks for Intrusive Ads
I see the complete opposite.

Xbox fans trying everything crazy to try to make PS5 bad with FUD.

PS fans are happy and excited with PS5.
Agreed. Because - what else are you going to do? Not like anything can change the specs at this point. And besides, The PS5 will be a great machine too. Just not as good as the Series X.
 
So I tried getting caught up on the thread but I’m confused.

Half the people use the quote from a dev saying they have to throttle the CPU back to obtain max GPU clock

Other half use Cerny quote that both can run maxed. So which is it?
 

Journey

Banned
Sim points....

- Variable clock was choose due how difficult was to cool PS4 so that allowed a easy and effective cooling system in PS5.... he says you will be surprised here.
- The clock is not defined by temperature but workload so you won’t have unexpected performance drops... the opposite the performance is pretty consistent.
- Inside the APU there is a power control unit.
- All PS5 runs and delivery the same performance.
- The power budget is enough to run both CPU and GPU at max speeds.
- Both CPU and GPU runs most of time near to or at peak frequency.
- He tive an example that dropping the frequency in 10% give a 27% power draw decrease so they only need a few percent drop in frequency to 10% drop I’m power draw.

A lot of crow eating here about the variable clocks.



You left the part out where he compared 2 GPUs matched at the same Teraflop number, one with higher clocks and 36 CUs like the PS5, and another RDNA GPU clocked lower, but with 40 CUs, and it turns out that the higher CU showed better performance despite equalized TF figure



hIG0YbF.jpg


HOE
LEE
SHIT

Where's the crow eating regarding CU count?
 
Last edited:

ethomaz

Banned
You left the part out where he compared 2 GPUs matched at the same Teraflop number, one with higher clocks and 36 CUs like the PS5, and another RDNA GPU clocked lower, but with 40 CUs, and it turns out that the higher CU showed better performance despite equalized TF figure

HOE
LEE
SHIT

Where's the crow eating regarding CU count?
Source?

BTW RDNA tests where the clock didn't scale like RDNA 2 is not valid.
But I'm curious to see your test.

How taste is the crow?
 
Last edited:
Assuming all you just said is true... I have the following questions for you;

  • Where is what Cerny talked about, regarding the PS5 doing things differently? Because what you just described is how basically every phone, laptop and PC out there works, and he was specific that the PS5 works differently.
  • Where does the power limit come in? Cerny was quite clear about workloads and the power limit. You didn't mention that anywhere.
  • If the console could handle both the CPU and GPU at max load, why would developers have to choose a profile to throttle the CPU to ensure the GPU runs at 2.23 GHz?
  • If developers prefer non-variable clocks for optimization, why have variable clocks if the console can reach the max clocks at max workloads at all times anyway?

Good luck.

that is actually a quote you can ask digital foundry directly

  • Where is what Cerny talked about, regarding the PS5 doing things differently? Because what you just described is how basically every phone, laptop and PC out there works, and he was specific that the PS5 works differently.
when a CPU goes idle, it or parts of it that are idle and do nothing, they go to sleep until they receive an interrupt, what PS5 do is reducing its frequency when the workload doent have to go full capacity for the task its the difference between going slower and stopping completely
  • Where does the power limit come in? Cerny was quite clear about workloads and the power limit. You didn't mention that anywhere
if you want to discuss something feel free to quote and we can discuss
  • If the console could handle both the CPU and GPU at max load, why would developers have to choose a profile to throttle the CPU to ensure the GPU runs at 2.23 GHz?
  • If developers prefer non-variable clocks for optimization, why have variable clocks if the console can reach the max clocks at max workloads at all times anyway?

"Regarding locked profiles, we support those on our dev kits, it can be helpful not to have variable clocks when optimising. Released PS5 games always get boosted frequencies so that they can take advantage of the additional power,"


developers need clock unlocked in devkits to see performance gains from functions and algorithms they implement

from my personal experience back when tried to make my own engine I made a function to display non-monospace text in opengl and c++ it reads a texture and display a quad made by 2 triangles 6 vertex total I dont index vertices in them because considered it pointles for sprites to have the triangle buffer, each quad had the texture part corresponding to the letter and space them from other letters based on a precalculated table with the horizontal size of each character I experimented the framerate I can achieve making one render call per character and how much I gain making a 3d model of the entire text block and keeping it until there is a change in text and then recalculate the whole thing if the cpu changes its speed based in the load then I wont know how much performance I was loosing in the drawcall and that was because CPU, the GPU remain clocked to max in both cases
 
Last edited:
Source?

BTW RDNA tests where the clock didn't scale like RDNA 2 is not valid.
I already addressed this, you guys need to quit... while you're behind...

These systems are only a few months away, and all this is going to come back to haunt you.
I really appreciate Richard covering this in specific because there's a lot of delusional assertions about frequency identically scaling compute. It can approach identical performance but when pressed hard more physical hardware will always handle load better, this is even the case for like for like compute ceilings.

This isn't going to make up a nearly 2 Teraflop advantage coupled to 16 additional compute units. Not even close.

szaNC8Q.jpg


What I also appreciate is him showing how a memory bandwidth bottleneck can present itself even at a mere 9.67 Teraflops with 448GB/s of bandwidth available.

6ntRomv.jpg


There's some hard realizations ahead for certain people, and it's best you come to grips with these things now because when the actual hardware hits the above is going to scale dramatically higher when compute variance of such a lopsided degree enters the fray.
 

ethomaz

Banned
I already addressed this, you guys need to quit... while you're behind...

These systems are only a few months away, and all this is going to come back to haunt you.
Yeap RX 5700 doesn't sustain 2100Mhz that is why it didn't show better performance it is running below the target.
lol

For example that is a test with a overclock to 1850Mhz but the card never reached it... it run around 1785Mhz.

graph-1.png.webp


If you can show that bench you showed is constant at 2100Mhz then you will have a point but it is hardly possible because RDNA doesn't sustain constant 2100Mhz.
 
Last edited:
Did you guys not watch the DF video? Cover your ears when you didnt want to hear certain facts?

Richard said it very early on in the video that Mark Cerny did acknowledge that ps5 games will be built with focus on jaguar CPU capabilities in mind.

I'm driving right now using speech to text so cant watch video and go back and quote for you, just watch the video again.
I think that has to do with back compat. It's why MS has the option of 3.8ghz single threaded, because most games developed with Jaguar in mind dont use multi thread, so they can just port over and benefit from the higher clock. It was actually a pretty smart decision from MS that didnt really cost anything, but gave a 200mhz increase.
 
Last edited:

Handy Fake

Member
Am I right in understanding that the increased frequency not only pushed up the (blagh) teraflops, but increases overall system performance in a way that extra CUs don't? I thought that's what he meant by increased gains through overclocking rather than widening.
I figured that was the whole point, to run in tandem with the IO and system buses as a unified system rather than brute forcing it?

Or am I singing sweet lullabies from my aged sphincter?
 
Last edited:
Top Bottom