• 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.

kensama

Member
I'm still mightily confused as to how a GCN @ 13.8TF can be worse than an RDNA1 @ 9.75TF.

To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.

Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.

Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.

What else is going on to affect the performance?


Because RDNA1 (GCN) has not RT core built in hardware. RDNA2 is like Nvidia RTX with tensor core built in hardware.
So for 13.8 TF in GCN you must add that vast part of TF will be used to do what tensor core or RDNA2 core inside GPU cn not on GCN achitecture (which must be done by software and use many ressources from GPU).
 
Last edited:
I'm still mightily confused as to how a GCN @ 13.8TF can be worse than an RDNA1 @ 9.75TF.

To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.

Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.

Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.

What else is going on to affect the performance?

It is a terrible measurment. It's only then really saying something if you have 2 systems/architectures that are a 100% the same an only differ in TFLOPs.
Then of course the system with more TFLOPs will always have an advantage.

But for now every system ( Nintendo/Microsoft/Sony/Google/PC ) has its own specs, tricks, apis, hardware, bandwith, bottlenecks and so on so relying only on tflops isn't necessarily the right way to determine the best system.
 
I'm still mightily confused as to how a GCN @ 13.8TF can be worse than an RDNA1 @ 9.75TF.

To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.

Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.

Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.

What else is going on to affect the performance?

EDIT/2nd Post: I haven't actually replied to your question, so another take:
Stating how many TFLOPs a hardware can handle can be done so in many ways, I could state that my GPU can do 20 TFLOPs. However this number is only theoratical referring only to one part of the GPU.
However is someone else checks if this statement is true they would realize the GPU would only reach this number on one specific task that isn't even practical. Because the system has some bottlenecks and needs to be fed the perfect match of data to reach this performance.
Or maybe someone will realize it will only reach that tflops for 1 second out of every minute but the average tflop count would be 12.5.
So yeah all these numbers that are floating around from every Company are marketing numbers.
Real world performance mostly differs at the end.
 
bw3u281.jpg


JLnoLax.jpg


736ckbY.jpg


qaJVlms.jpg


Cool concept.
 

B_Boss

Member
I'm still mightily confused as to how a GCN @ 13.8TF can be worse than an RDNA1 @ 9.75TF.

To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.

Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.

Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.

What else is going on to affect the performance?

I imagine it would be the better device in terms of TF performance and not necessarily overall? It’s possible, between two consoles, for the console with higher TF’s could be technically superior in every way but of course that is not a given. Many factors are involved in making a specific device “better” overall.
 

Redlight

Member
I imagine it would be the better device in terms of TF performance and not necessarily overall? It’s possible, between two consoles, for the console with higher TF’s could be technically superior in every way but of course that is not a given. Many factors are involved in making a specific device “better” overall.

That's really nothing more than wishful thinking. The two devices are designed around the same basic technologies so advantages in power will be there for all to see. This is the equivalent of saying "The PS5's SSD is technically superior in every way but many factors are involved in making a specific device "better".

You can trust the specs far more than any spin or commentary that contradicts them.
 

GetSchwifty

Banned
From DF, PS4 Pro's 1620p is checkerboard.
I'm pretty certain it's checkerboarded on X1X too.

Even if it's not, it doesn't matter. X1X is a bomb on the market and it's not exactly hard to see why.

Sony/Cerny represented PS4 Pro accurately at introduction. Consumers got exactly what was promised.

X1X was inaccurately represented as a "no compromises" native 4K machine. It just seems underpowered for that purpose.

Cerny was probably correct with his "personal estimation" of 8TF GCN being necessary for a true 4K experience.

PS5 should easily be able to hit 4K/60 if the devs choose to do so, so TFLOPs don't matter nearly as much next gen.

Xbox should probably have chosen a slower GPU and put money back into RAM and an optical audio port.
 

Kusarigama

Member
kraken is proprietary from RAD Game Tools, Sony bought a licence to pack it inside the ps5 (a price you as the user will pay). And we don't really know in reality how much better it will be, as always there is a diminishing return after a treshold, so maybe the real world difference (not some synthetic bench) will be irrelevant. Until we see some games, this is all speculation.
It is from RAD game tools but i don't think they charge Sony for including a decompressor unit for it. I think kraken is already being used in PS4 in "oodle", Shadow of War uses it. Now that kraken decompression can be done natively on PS5, even more developers will use it.

Edit: Oodle kraken is indeed being used on current gen consoles.
 
Last edited:

GetSchwifty

Banned
Why would you want an optical audio port when higher quality audio can be send via HDMI? Source
It's for convenience and legacy. It's strange that MS support BC so heavily but decide that their userbase shouldn't use their optical-only audio systems any longer.

Earlier leaks showed XSX had optical... I think it's less to do with pushing technology forward, and more a cost-cutting measure. I would be annoyed if PS5 gets rid of it also, like they did on PS4 Slim.
 

CrysisFreak

Banned
Xbox should probably have chosen a slower GPU and put money back into RAM and an optical audio port.

While I do not care about optical audio (but the option is good) you're right about RAM I think. More RAM would've been a more significant advantage overall probably.
But at the same time it's good that they have the same amount of RAM tbh. We need to keep them on about the same level, then they have no other choice but to differentiate through games.
And now that's gonna be real good for our fun times.
 

Shmunter

Member
While I do not care about optical audio (but the option is good) you're right about RAM I think. More RAM would've been a more significant advantage overall probably.
But at the same time it's good that they have the same amount of RAM tbh. We need to keep them on about the same level, then they have no other choice but to differentiate through games.
And now that's gonna be real good for our fun times.
Equivalent ram, but the jury is still out on that oddball XsX setup. By all known measures it looks to be somewhat special, special olympics special.

I take no pleasure in deriving to such a conclusion.
 
D

Deleted member 775630

Unconfirmed Member
It's for convenience and legacy. It's strange that MS support BC so heavily but decide that their userbase shouldn't use their optical-only audio systems any longer.

Earlier leaks showed XSX had optical... I think it's less to do with pushing technology forward, and more a cost-cutting measure. I would be annoyed if PS5 gets rid of it also, like they did on PS4 Slim.
Yeah it is probably a cost cutting measure, but just last week I was looking for a proper soundbar, and decided not to buy one because it only had optical and not HDMI. So maybe it's due to my situation that I don't care as much. For others it's indeed annoying
 

draliko

Member
It is from RAD game tools but i don't think they charge Sony for including a decompressor unit for it. I think kraken is already being used in PS4 in "oodle", Shadow of War uses it. Now that kraken decompression can be done natively on PS5, even more developers will use it.
Sow and other games already pay rad game tools for using the tools and sdks. In no way the algorithm is free, rad game tools lives selling tools, they for sure have made a super price to Sony but still they're getting payed, other than the fact than now more or less they have assured the usage of their tools from a majority of studios willing to work for ps5. Or sony payed and the tool is included in their sdk or the studio has to pay to compress in kraken format, and it's not assured that everyone is willing to do that. On the other half of the world bcpack is made by Ms so will be free for sure. Sorry for errors, writing on the phone.
 

CrysisFreak

Banned
Sow and other games already pay rad game tools for using the tools and sdks. In no way the algorithm is free, rad game tools lives selling tools, they for sure have made a super price to Sony but still they're getting payed, other than the fact than now more or less they have assured the usage of their tools from a majority of studios willing to work for ps5. Or sony payed and the tool is included in their sdk or the studio has to pay to compress in kraken format, and it's not assured that everyone is willing to do that. On the other half of the world bcpack is made by Ms so will be free for sure. Sorry for errors, writing on the phone.
I hope this is a win-win situation for Sony and RAD.
As a tech enthusiast I LOVE algorithms implemented in hardware. There's something poetic about it. No compromises, dedicated to optimisation.
 

draliko

Member
I hope this is a win-win situation for Sony and RAD.
As a tech enthusiast I LOVE algorithms implemented in hardware. There's something poetic about it. No compromises, dedicated to optimisation.
Yes, the funny thing is that the majority of people is lost doing console wars insted of noticing how both the system have made great strides to be absolute wonderful GAMING machines, everything was chosen to bring equilibrium and eliminate obstacles (in both camps). We finally have manufacturers trying to deliver solid all around systems, not gimped in any major area and this is the most important thing, next gen have the potential to be glorious, and to be very expensive for the studios catering to AAA budget titles
 

Kusarigama

Member
Sow and other games already pay rad game tools for using the tools and sdks. In no way the algorithm is free, rad game tools lives selling tools, they for sure have made a super price to Sony but still they're getting payed, other than the fact than now more or less they have assured the usage of their tools from a majority of studios willing to work for ps5. Or sony payed and the tool is included in their sdk or the studio has to pay to compress in kraken format, and it's not assured that everyone is willing to do that. On the other half of the world bcpack is made by Ms so will be free for sure. Sorry for errors, writing on the phone.
What i am saying is sure RAD charges developers to use their compression technology but why would Sony pay RAD to have it running on their machine? It's like Epic charges developers to use their Unreal engine to make games but Sony doesn't pay Epic to have their engine run on PlayStation. Hope this clears the confusion.
Unless there is some evidence that says that RAD and Sony made an agreement to have Kraken be on PS5 as the defacto compression, it is all your speculation.
 
Last edited:

CrysisFreak

Banned
What i am saying is sure RAD charges developers to use their compression technology but why would Sony pay RAD to have it running on their machine? It's like Epic charges developers to use their Unreal engine to make games but Sony doesn't pay Epic to have their engine run on PlayStation. Hope this clears the confusion.
Unless there is some evidence that says that RAD and Sony made an agreement to have Kraken be on PS5 as the defacto compression, it is all your speculation.
Kraken is now used by all PS5 software, they literally implemented the algorithm in hardware. So yes, Kraken is the factual default compression on PS5. System wide.

Do you have a link to this statement?
Kraken is proprietary. Unless RAD is giving it to Sony for free (they're not), Sony IS paying for it. Even if they get just 5$ per console. That's 500 million for all PS5s (projected).
 

GetSchwifty

Banned
Kraken is now used by all PS5 software, they literally implemented the algorithm in hardware. So yes, Kraken is the factual default compression on PS5. System wide.


Kraken is proprietary. Unless RAD is giving it to Sony for free (they're not), Sony IS paying for it. Even if they get just 5$ per console. That's 500 million for all PS5s (projected).
I bet it's closer to $1 per console. Just a guess.

Edit: And it's likely even less. Any more and they would have been better off just purchasing the whole company out-right or roll their own solution.
 
Last edited:

Kusarigama

Member
Kraken is now used by all PS5 software, they literally implemented the algorithm in hardware. So yes, Kraken is the factual default compression on PS5. System wide.


Kraken is proprietary. Unless RAD is giving it to Sony for free (they're not), Sony IS paying for it. Even if they get just 5$ per console. That's 500 million for all PS5s (projected).
PS4 has dedicated zlib decompressor but still some games use kraken on PS4. Like wise dedicated Kraken decompressor is on PS5 but as usual devs are free to use their choice of decompression system.
Again unless there is report saying that RAD & Sony have an agreement to push Kraken mandatorily on all PS5 games, it's like Epic Unreal situation.
 

draliko

Member
I bet it's closer to $1 per console. Just a guess.

Edit: And it's likely even less. Any more and they would have been better off just purchasing the whole company out-right or roll their own solution.
I think a bit more, rad is no small company is ade facto industry leader for tools. And licenses cost a shit load of money specially for big studios, development tools costs, and a lot too.
 

GetSchwifty

Banned
PS4 has dedicated zlib decompressor but still some games use kraken on PS4. Like wise dedicated Kraken decompressor is on PS5 but as usual devs are free to use their choice of decompression system.
Again unless there is report saying that RAD & Sony have an agreement to push Kraken mandatorily on all PS5 games, it's like Epic Unreal situation.
Of course it's not mandatory, but why would a dev not use it if it's baked into the system?

Does it mean Kraken Is it the de facto default decompression tech in PS5? It would appear to be the case.
 

draliko

Member
PS4 has dedicated zlib decompressor but still some games use kraken on PS4. Like wise dedicated Kraken decompressor is on PS5 but as usual devs are free to use their choice of decompression system.
Again unless there is report saying that RAD & Sony have an agreement to push Kraken mandatorily on all PS5 games, it's like Epic Unreal situation.
zlib is free, kraken is not. So for now i don't believe RAD is going away with nothing in hands. But i don't think we'll ever see an official statement, those things are not at our level.
 

Kusarigama

Member
Of course it's not mandatory, but why would a dev not use it if it's baked into the system?

Does it mean Kraken Is it the de facto default decompression tech in PS5? It would appear to be the case.
There could be myriad reasons for their choice.
Remember here RAD charges to provide compression and decompression algorithms.
 
Last edited:

CrysisFreak

Banned
PS4 has dedicated zlib decompressor but still some games use kraken on PS4. Like wise dedicated Kraken decompressor is on PS5 but as usual devs are free to use their choice of decompression system.
Again unless there is report saying that RAD & Sony have an agreement to push Kraken mandatorily on all PS5 games, it's like Epic Unreal situation.
Interesting. I thought any data accessed on the SSD goes through an IO pipeline that has Kraken as a basic, always-on part of it. Hard to guess rn I think.
 

GetSchwifty

Banned
There could be myriad reasons for their choice.
Remember here RAD charges to provide compression solution. Not decompression.
Hmm... compression seems useless without decompression. It's an integrated, end-to-end system. If you have a proprietary compressor, you'd need a proprietary decompressor too.

The decompressor could be reverse-engineered, but they wouldn't be marketing Kraken then.
 

GetSchwifty

Banned
Timestamped link for what Cerny said about compression:



Edit: He does mention hustling to build a "custom decompressor into the IO". It could still be interpreted either way.
 
Last edited:

Kusarigama

Member
Interesting. I thought any data accessed on the SSD goes through an IO pipeline that has Kraken as a basic, always-on part of it. Hard to guess rn I think.
Indie games like SteamWorld heist, which are very small <1GB total size amd playable on all sorts of devices might not even use any kind of compression/decompression.
 

Kusarigama

Member
Hmm... compression seems useless without decompression. It's an integrated, end-to-end system. If you have a proprietary compressor, you'd need a proprietary decompressor too.

The decompressor could be reverse-engineered, but they wouldn't be marketing Kraken then.
Let me ask you this, Do you think Sony is paying RAD for PS4 as PS4 also seems to be decompressing kraken data without a dedicated unit?
 

draliko

Member
Let me ask you this, Do you think Sony is paying RAD for PS4 as PS4 also seems to be decompressing kraken data without a dedicated unit?
the game studio is paying rad to decompress kraken via software not using a custom hardware already present in the system. This is totally different dude.
 
Equivalent ram, but the jury is still out on that oddball XsX setup. By all known measures it looks to be somewhat special, special olympics special.

I take no pleasure in deriving to such a conclusion.

Both systems use a shared pool for CPU & GPU, in reality, this equalizes the non-issues being brought forward regarding this technique. The CPU has less bandwidth on the XSX side, though in most games the CPU doesn't even need half the bandwidth offered by DDR4 on PC.
 

GetSchwifty

Banned
Indie games like SteamWorld heist, which are very small <1GB total size amd playable on all sorts of devices might not even use any kind of compression/decompression.
I agree with this, that many games won't be using it at all... which is why I believe the per-unit cost of including the tech should be fairly low. Maybe <$1 per PS5 sold, $20 per devkit.

Edit: This would still be highly, highly lucrative for RAD tools and really solidifies the tech as an industry standard.
 
Last edited:

rnlval

Member
Data is spread across all chips evenly, for GPU to access data from the 16bit half it would limit bandwidth to 280GB/s and CPU to 168GB/s
This is an undesirable outcome

edit: and the net results the same 160GB used 80GB gained= 80GB/s used for average CPU access of 48GB/s

My simplified XSX work-in-progress interleaved memory model for the logical single 320-bit channel model.

XfbP57G.png


For 2G GDDR6 chips, I factored the dual 16bit channels and AMD GPU's "combined scatter" memory access patterns.




I'm following this example
5-Figure3-1.png





There are other details to work out. Without documentation, I don't know XSX's customizations done on the "logical view" to "physical view" resolve map.
 

SirTerry-T

Member
bw3u281.jpg


JLnoLax.jpg


736ckbY.jpg


qaJVlms.jpg


Cool concept.
Not practical at all, seems like a nightmare to clean 🙄

Why those concepts always looks like this? Those graphics artists are pure nerds 🤣
Yeah, it looks cool but way over designed. That housing would be a paradise for dust collecting.
Why do a lot of these concepts insist on resembling something that's dropped off a Gundam mech. Sometimes less is more :)
 
Status
Not open for further replies.
Top Bottom