• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.
Sinthor

Sinthor

Gold Member
Shmunter said:
The difference may be in that the 3080 has 10gig at it's disposal. On XsX it may be that when the slow pool of data is continually accessed by any component...e.g. the cpu, the entire bandwidth gets downgraded during the transfer and affecting gpu bandwidth.

There may be ways to optimize around it with some data striping and clever access, but it may be a headache many don't bother to deal with.

I mean it's all speculation, but when looking for answers - this is the difference between the systems and sticks out. What else can we point to? Could be a reason it's never been attempted before (ps3 is different)
Click to expand...
That's why I tend to think memory configuration may be the answer as well. Like you say and, at least in my non-expert opinion, the biggest difference between the boxes (besides overall clock speed on the GPU) is the memory configuration. So, it tends to draw the eye. Will be interesting to see how this ends up.

That this speculation is out there and we haven't seen Microsoft stomping all over it and promising different results later on also gives it more credence in my eyes. Sure, maybe MS just doesn't feel the need to respond, but they are so fast in jumping on and pointing out any advantage even if it's just marketing...that their silence seems more telling to me. Meh.. We'll see!
 
SlimySnake

SlimySnake

Flashless at the Golden Globes
THEAP99 said:
gamerant.com

Sony AI Could Change The Way Games Balance Enemy Difficulties

Sony submits a patent detailing the use of AI learning to balance video game difficulty in real time, potentially changing video game fundamentals.
gamerant.com gamerant.com

Developers are becoming this lazy now? Seriously? Want want truly unique hand-crafted dynamic experiences in our games, not soulless machine generated crap.. this is sad
Click to expand...
umm some games already do this. MGSV A.I adapts to you headshotting them by wearing helmets, or wearing night vision goggles if you chose to do missions at night, or wearing gas masks if you kept using sleeping gas on the enemy.

I am guessing this is more of that, but the A.I doesnt need the developer to key in those things and adjusts itself on the fly. We should want games to become dynamic like that.
 
SlimySnake

SlimySnake

Flashless at the Golden Globes
Sinthor said:
That this speculation is out there and we haven't seen Microsoft stomping all over it and promising different results later on also gives it more credence in my eyes. Sure, maybe MS just doesn't feel the need to respond, but they are so fast in jumping on and pointing out any advantage even if it's just marketing...that their silence seems more telling to me. Meh.. We'll see!
Click to expand...
yeah the silence is definitely curious after over a year of chest beating. i would like to think that they have been humbled, but with mostly cross gen games on the horizon and little to no next gen xbox exclusives, I think they know the worst is yet to come.

Even Horizon which is a cross gen game looks a gen ahead of Halo which is their cross gen game running on a 12 tflops monster that ate the last monster for breakfast.

HmOedDT.gif


Q5GBjVF.gif


Just imagine what happens in case GoW turns out to be next gen only.
 
Last edited:
Three Jackdaws

Three Jackdaws

Banned
MistBreeze said:
Yeah act1 that is
It is one of the best games of all time easily

but act 2

oh man it is a shit show
Click to expand...
Agreed.

The highlight of the game for me was the gameplay mechanics and and style, especially with the additions of D-dog and Quiet ( ͡° ͜ʖ ͡°)

The weapons and armour customisations were also terrific.

I’m just trying to imagine what a next-gen Metal Gear game would look like.
 
Last edited:
MistBreeze

MistBreeze

Member
villentretenmerth said:
Agreed.

The highlight of the game for me was the gameplay mechanics and and style, especially with the additions of D-dog and Quiet

The weapons and armour customisations were also terrific.

I’m just trying to imagine what a next-gen Metal Gear game would look like.
Click to expand...
Yeah the gameplay is just spot on

why konami this is a crime that you did not let kojima do act2 properly
Oh well

Cause of that I believe in kojima productions

I think Kojima should do a stealth game in the veins of mgs ,,, same formula he used to

but a new lure ... new protagonist and characters
to avoid trademark and legal issues

hope sony gives him enough time and money
 
Red_Driver85

Red_Driver85

Member
IntentionalPun said:
I don't know; why did Cerny claim CU occupancy was an issue for wider GPUs? Unless I'm misremembering.
Click to expand...
There seem to be smart guys sitting here. It's time to remember laws like Amdahl's Law. Which states that the more processors there are, the lower their efficiency is when fully parallelized. There is a limitation on performance growth when computations are parallelized and there has always been. In the case when a task is divided into several parts or threads, the total execution time on a system with a large number of computational units can't be less than the execution time of the slowest element. That is, all performance is limited to the very minimum that the system is limited to. Acceleration of program code execution due to parallelization of its instructions on a set of computers is limited by the time required to execute its sequential instructions.
 
M1chl

M1chl

Currently Gif and Meme Champion
THEAP99 said:
gamerant.com

Sony AI Could Change The Way Games Balance Enemy Difficulties

Sony submits a patent detailing the use of AI learning to balance video game difficulty in real time, potentially changing video game fundamentals.
gamerant.com gamerant.com

Developers are becoming this lazy now? Seriously? Want want truly unique hand-crafted dynamic experiences in our games, not soulless machine generated crap.. this is sad
Click to expand...
I mean AI is in dire improvement for a long time, last time someone tried was Halo: Reach. Even Destiny Ai suck ass. So I believe in this new technologies. Even this require a lot of work, nothing to do with laziness.
 
S

Superayate

Member
Sinthor said:
That's why I tend to think memory configuration may be the answer as well. Like you say and, at least in my non-expert opinion, the biggest difference between the boxes (besides overall clock speed on the GPU) is the memory configuration. So, it tends to draw the eye. Will be interesting to see how this ends up.

That this speculation is out there and we haven't seen Microsoft stomping all over it and promising different results later on also gives it more credence in my eyes. Sure, maybe MS just doesn't feel the need to respond, but they are so fast in jumping on and pointing out any advantage even if it's just marketing...that their silence seems more telling to me. Meh.. We'll see!
Click to expand...

In fact, for me that's not the splitted memory that could explain that the Series X does not do better than PS5. I think that's simply that the PS5 is an overall more efficient hardware.
The splitted memory is not the real problem as many seem to think and will never tank the performances (clearly not at a level to reach the PS5 bandwidth level foe example). When we speak about slow pool and fast pool, that's not physically the case. The access to the memory is done exactly in the same way than in PS5, simply because that is a virtual split. It has drawbacks yes (less flexibility with fixed memory quantity allocated for GPU, CPU, etc... and when the CPU is accessing memory it has slighty more impact than PS5 etc...)
I think when I will have time, I'll try to find a way to explain that with an example based of one splitted memory configuration I have worked with in the past. In fact, but for a customer (it was not for PC and smartphone), we have decline a chip with many configuration and boards, playing mainly with the memory quantity to adapt the cost, performances and the usage.
I wanted to rise one other thing, don't forget that if this splitted memory configuration would have tanked so much the performances, MS instead of using four 1GB chip + six 2GB chip, could have simply used eight 2GB chip and not used all the memory controllers (and keep the capability to use the chip in 20GB configuration for server) for example.
 
P

Panajev2001a

GAF's Pleasant Genius
Superayate said:
In fact, for me that's not the splitted memory that could explain that the Series X does not do better than PS5. I think that's simply that the PS5 is an overall more efficient hardware.
The splitted memory is not the real problem as many seem to think and will never tank the performances (clearly not at a level to reach the PS5 bandwidth level foe example). When we speak about slow pool and fast pool, that's not physically the case. The access to the memory is done exactly in the same way than in PS5, simply because that is a virtual split. It has drawbacks yes (less flexibility with fixed memory quantity allocated for GPU, CPU, etc... and when the CPU is accessing memory it has slighty more impact than PS5 etc...)
I think when I will have time, I'll try to find a way to explain that with an example based of one splitted memory configuration I have worked with in the past. In fact, but for a customer (it was not for PC and smartphone), we have decline a chip with many configuration and boards, playing mainly with the memory quantity to adapt the cost, performances and the usage.
I wanted to rise one other thing, don't forget that if this splitted memory configuration would have tanked so much the performances, MS instead of using four 1GB chip + six 2GB chip, could have simply used eight 2GB chip and not used all the memory controllers (and keep the capability to use the chip in 20GB configuration for server) for example.
Click to expand...
That is fair, I do not think it is a world ending problem for them, but it also means that if the GPU needs fast access to memory beyond the 10 GB section you either need to move it from one region to the other or simply take a bandwidth hit when accessing it. The address space is still unified and you do not have to treat it differently, but you have different throughout depending on the memory region you address it so I would bet your code has to be aware of it (or can be made aware of it... not the first console doing it, not the last possibly).

The “slow” memory region is still quite fast, but it is another hoop to jump through which may mean less time spent iterating on other parts of your code. Still eagerly awaiting your next post :).
 
Dibils2k

Dibils2k

Member
SlimySnake said:
yeah the silence is definitely curious after over a year of chest beating. i would like to think that they have been humbled, but with mostly cross gen games on the horizon and little to no next gen xbox exclusives, I think they know the worst is yet to come.

Even Horizon which is a cross gen game looks a gen ahead of Halo which is their cross gen game running on a 12 tflops monster that ate the last monster for breakfast.

HmOedDT.gif


Q5GBjVF.gif


Just imagine what happens in case GoW turns out to be next gen only.
Click to expand...
lol you think top gif is ingame
 
Lysandros

Lysandros

Member
Shmunter said:
Something is preventing the GPU to breathe properly that's for sure. It is still a 12tf RDNA2 architecture. But MS screwed up somewhere if it cannot reach it's potential.
Click to expand...
I think the main reason is the GPU architecture itself with L1 caches having to feed 40% more CUs per shader array while being ~20% slower; together with ~20% lower rasterization and pixel fill rate.
 
Lunatic_Gamer

Lunatic_Gamer

Gold Member

Sony AI Could Change The Way Games Balance Enemy Difficulties​


A new patent from Sony details its potential plans for handling difficulty in video games, titled "Use of Machine Learning to Increase or Decrease Level of Difficulty in Beating Video Game Opponent." That name is a bit of a mouthful, but it's essentially a form of reactive difficulty. As players progress through a game, the enemies and bosses will get stronger or weaker to accommodate the player's level of skill.
Click to expand...

2HFP2p4.jpg


gamerant.com

Sony AI Could Change The Way Games Balance Enemy Difficulties

Sony submits a patent detailing the use of AI learning to balance video game difficulty in real time, potentially changing video game fundamentals.
gamerant.com gamerant.com
 
Last edited:
kyliethicc

kyliethicc

Member
Superayate said:
In fact, for me that's not the splitted memory that could explain that the Series X does not do better than PS5. I think that's simply that the PS5 is an overall more efficient hardware.
The splitted memory is not the real problem as many seem to think and will never tank the performances (clearly not at a level to reach the PS5 bandwidth level foe example). When we speak about slow pool and fast pool, that's not physically the case. The access to the memory is done exactly in the same way than in PS5, simply because that is a virtual split. It has drawbacks yes (less flexibility with fixed memory quantity allocated for GPU, CPU, etc... and when the CPU is accessing memory it has slighty more impact than PS5 etc...)
I think when I will have time, I'll try to find a way to explain that with an example based of one splitted memory configuration I have worked with in the past. In fact, but for a customer (it was not for PC and smartphone), we have decline a chip with many configuration and boards, playing mainly with the memory quantity to adapt the cost, performances and the usage.
I wanted to rise one other thing, don't forget that if this splitted memory configuration would have tanked so much the performances, MS instead of using four 1GB chip + six 2GB chip, could have simply used eight 2GB chip and not used all the memory controllers (and keep the capability to use the chip in 20GB configuration for server) for example.
Click to expand...
Nah I think there's a reason why every GPU Nvidia and AMD make has evenly interleaved memory chips (and PS4 PS5 X1 X1X).

They clearly just wanted more than 448 GB/s bw but didn't want to spend more money on faster 16 Gb/s G6 RAM, or to spend even more and use 10 chips for 20 GB total. They cheapd out and compromised. Cutting down bus to 256 wouldn't get them extra bw.

It seems like theoretically the split bandwidth setup might not be an issue for some games if coded right, but its gonna take some extra effort from the devs.

Cause 1 big unified pool of fast RAM is clearly what console devs prefer. No devs complained when the One X had 1 big unified pool of 326 GB/s bw RAM.
 
Last edited:
assurdum

assurdum

Banned
Shmunter said:
Something is preventing the GPU to breathe properly that's for sure. It is still a 12tf RDNA2 architecture. But MS screwed up somewhere if it cannot reach it's potential.
Click to expand...
It's a bit early to say if multiplat point out there is an "issue" in the series X hardware architecture, though I'm firmly convinced the virtual splitted configuration doesn't helps at all, especially with such high CUs number to feed. I suspect more, one of the main reason of the clamour around the multiplat perfomance, it's too many people had unrealistic expectations about the real gap between the 2 hardware, especially when notorious channel DF insists to spread such narrative.
 
Last edited:
L

Loope

Member
assurdum said:
It's a bit early to say if multiplat point out there is an "issue" in the series X hardware architecture, though I'm firmly convinced the virtual splitted configuration doesn't helps at all, especially with such high CUs number to feed. I suspect more, one of the main reason of the clamour around the multiplat perfomance, it's too many people had unrealistic expectations about the real gap between the 2 hardware, especially when notorious channel DF insists to spread such narrative.
Click to expand...
Point on the doll where DF touched you.
 
TheThreadsThatBindUs

TheThreadsThatBindUs

Member
akaFullmetal said:
I think that strategy would be better than Sony or Microsoft just out right buying up companies and consolidating the market.
Click to expand...
Buying shares in a publicly traded studio/publisher, unless it's a controlling interest (and in that case you basically own them outright), isn't going to secure you any exclusives.

They need to money hat for exclusives, as it's always been. Or buy the studio. Or buy the IP. Or contract the developer by completely funding development.
 
TheThreadsThatBindUs

TheThreadsThatBindUs

Member
Rea

Rea

Member
TheThreadsThatBindUs said:
Because with rasterization, triangles/primitives are mapped to the pixel grid before pixel shading occurs. The ALUs (Arithmetic and Logic Units) within the CUs work on these portions of the pixel grid for each polygon by breaking it further down into pixel fragments and passing the work to ALU to work on it on a pixel-by-pixel basis.

To ensure the maximum level of parallelism, in order to keep you ALU doing useful work, you want triangles that span a large number of pixels. So that when shading, in batches of pixels sized to correspond to the width of the ALUs, you're maximizing utilization of your ALU.

A crude example is as follows:

Let's say I have a 4x4 fragment passed to my 16-wide SIMD unit within the CU.

If the polygon spans 14 of the 16 pixels, then 14 out of 16 pixels get shaded in a single clock cycle (87.5% utilization)
If the polygon is smaller and spans only 4 of the 16 pixels, only 4 pixels get shaded per clock cycle (25% utilization) so the efficiency is significantly lower.

So in the above example scene with many 4x pixel triangles, adding more CUs to my GPU barely helps real world shading performance at all, because I'm getting only 25% utilization out of my ALUs. Increasing GPU clock speed however, will help overall performance more, because in a 30fps game with 33ms frame time, increasing my clock speed will mean I can shade more real 4x pixel polygons within my frame-time budget.

The above is kind of a gross oversimplification, but it's just to give you the gist.
Click to expand...
Thx for the explanation, i really appreciate it.
 
LucidFlux

LucidFlux

Member
Red_Driver85 said:
There seem to be smart guys sitting here. It's time to remember laws like Amdahl's Law. Which states that the more processors there are, the lower their efficiency is when fully parallelized. There is a limitation on performance growth when computations are parallelized and there has always been. In the case when a task is divided into several parts or threads, the total execution time on a system with a large number of computational units can't be less than the execution time of the slowest element. That is, all performance is limited to the very minimum that the system is limited to. Acceleration of program code execution due to parallelization of its instructions on a set of computers is limited by the time required to execute its sequential instructions.
Click to expand...

This limit to parallelism scaling performance also highlights that there will always be a bottleneck in every system.

I also see a lot of people make the mistake of stating that a system only has a singular bottleneck or that bottlenecks have been removed. Any component in a system's pipeline can become the bottleneck if tasked with a particular workload. So when I see people claim that a system removed all bottlenecks or that X is the bottleneck of this system, it fails to recognize that the specific work being done is what determines what component is the bottleneck.

Put it this way, if a system didn't have ANY bottlenecks then we'd have infinite resolution and fps, but there is always something in the chain that everything else is waiting on to finish its work.
 
IntentionalPun

IntentionalPun

Ask me about my wife's perfect butthole
Red_Driver85 said:
There seem to be smart guys sitting here. It's time to remember laws like Amdahl's Law. Which states that the more processors there are, the lower their efficiency is when fully parallelized. There is a limitation on performance growth when computations are parallelized and there has always been. In the case when a task is divided into several parts or threads, the total execution time on a system with a large number of computational units can't be less than the execution time of the slowest element. That is, all performance is limited to the very minimum that the system is limited to. Acceleration of program code execution due to parallelization of its instructions on a set of computers is limited by the time required to execute its sequential instructions.
Click to expand...
I do async programming all the time, this is a concept I understand, kick off X threads, wait for them all to complete, last one to finish finishes the set. I'm not a coder for games though, but my understanding is that GPUs are designed for parallelization. Faster GPUs in the nVidia/AMD world add more and more parallelization.

But what others said might be what you mean here; that XSX will be limited by split memory slowing down some of those computations.
 
Last edited:
Great Hair

Great Hair

Banned
LiquidRex said:


Tbh I'm not a big fan of these BC third party titles on PS5, they'll never represent the Consoles capabilities... Really you could say they are just lazy ports for the console. 🔥
Click to expand...

Ubisoft has loaded The Division 2 on PS5. It has lost the FOG, the reflections are bugged, the maximum resolution it reaches is the one reached by PS4 Pro, it has a higher pop-in ... Of course, patching for backward compatibility instead of making a port takes its toll on PS5.
Click to expand...
 
Status
Not open for further replies.

Similar threads

AMD RDNA 5 To Be A Completely New GPU Architecture From The Ground Up, RDNA 4 Mostly Fixes RDNA 3 Issues & Improves Ray Tracing
News Rumor Hardware 2 3
128
7K
SoloCamo replied
Was 8 CPU cores enough for 9th gen consoles? Strix Halo, the prefect console.
Opinion Analysis 2
62
4K
Tertullian replied
Tom Warren - Sony’s PS5 Pro is real and Sony is asking developers to get games ready over the summer, with a push for ray-tracing support
News Hardware Hype 2 3 4
186
15K
DanielG165 replied
Report: Upcoming 'Turin' Zen 5 CPUs will support 4TB of RAM at 6000 MT/s
Rumor Hardware
43
3K
OverHeat replied
Leak - Zen 5 is now in mass production, release in late 2024
Hardware 2
61
6K
CuteFaceJay replied
Top Bottom