• 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.

Xbox Velocity Architecture - 100 GB is instantly accessible by the developer through a custom hardware decompression block

On the other hand we dont have any idea about PS5 SSD except for pure hype and speculation.

So . . . Actual specs are pure hype and speculation?

Sounds more like another secret sauce argument "Well the logo from the same company looks similar
so I assume this ~500 dollar box is going to have some technology from a high end server in it"

Again... sounds kind of like speculation to me. I'm just saying.
If Sony's specs are in hand, and Microsofts are too... what good is bringing up AMD's technology
if both systems are running the same AMD architecture? You might have a point if they were
not literally both possibly using the same thing.

And yeah.... it IS just a name basically. Whats wrong with me saying its a name?
If the NAMED TECHNOLOGY whatever it may be is saying "Look we're going to give you
access at 6gb compressed" it doesnt matter what its called, its not going to be faster or
as fast as something that can do 11gb compressed.
 
Biggest issue io solution of xsx has compared to ps5 is not speed. Its latency. Ps5 io complex is physically inside the apu
Ms io complex is physically not inside apu. This creates latency issues due to law of physics limitation.

We have to see how they remedy this latency issue. Latency could make hitting max speeds difficult to achieve. But I m sure they will try to remedy it

I thought “instant access” implies no latency (for the 100 gigs)
 

rntongo

Banned
Custom magic HW that nobody can name of course... See if you do not have exact details on PS5 it must be Sony hiding something that should cause concern, but Xbox is just hiding a magic 2-3x memory multiplier.


It would not be, if it were true (2-3x multiplier over virtual texturing and PRT’s).



By the same token PS5’s SSD could be rated 11-12 GB/s since we are making do with random multipliers based on virtual texturing benefits over provided specs.
[/QUOTE]
I think you're biased. They have upcoming events where more information will be revealed. You're making a false claim that they do not want to reveal this information eventually.
 

Thirty7ven

Banned
Its not the same token though. The customizations of XSX SSD are not available in the PS5 SSD. It is a custom hardware decompression blockm that can deliver 6 GB/s, not the general 4.8 GB/s compressed. And then SFS "is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance."(its in the OP).
I/O Throughput2.4 GB/s (Raw), 4.8 GB/s (Compressed, with custom hardware decompression block)

Dude you're adding numbers when they are already included in the 2.4 RAW and 4.8 Compressed. The later takes into account the custom decompression block.
It's very impressive how you interpret the numbers and how they add up. So you say it's 4.8 general + 6GB/S custom X 3 (SFS) = 32 GB/S. Holy fuck that's fast.

But I'll let you do you.
 

rntongo

Banned
So . . . Actual specs are pure hype and speculation?

Sounds more like another secret sauce argument "Well the logo from the same company looks similar
so I assume this ~500 dollar box is going to have some technology from a high end server in it"

Again... sounds kind of like speculation to me. I'm just saying.
If Sony's specs are in hand, and Microsofts are too... what good is bringing up AMD's technology
if both systems are running the same AMD architecture? You might have a point if they were
not literally both possibly using the same thing.

And yeah.... it IS just a name basically. Whats wrong with me saying its a name?
If the NAMED TECHNOLOGY whatever it may be is saying "Look we're going to give you
access at 6gb compressed" it doesnt matter what its called, its not going to be faster or
as fast as something that can do 11gb compressed.
To be honest besides this part about the PS5 SSD Specs, the rest of her argument is pretty solid.
 
To be honest besides this part about the PS5 SSD Specs, the rest of her argument is pretty solid.

How so?

Neither is out. The manufacturers have given their specs. My only original argument is
Microsofts speculative, promised specs are still half of Sonys promised specs (using speculative as she did, in this case)
and if some reference to AMD architecture is made.... Sonys system is also AMD so why even say so?
 

rntongo

Banned
Dude you're adding numbers when they are already included in the 2.4 RAW and 4.8 Compressed. The later takes into account the custom decompression block.
It's very impressive how you interpret the numbers and how they add up. So you say it's 4.8 general + 6GB/S custom X 3 (SFS) = 32 GB/S. Holy fuck that's fast.

But I'll let you do you.
Nooooooo! The 4.8GB/s is the actual throughput! SFS is not part of the 4.8GB/s. It's a memory multiplier so even at 4.8GB/s it could stream textures equivalent to a system with 2-3x this throughput!
 

M1chl

Currently Gif and Meme Champion
I know unreal engine 5.0 is being developed for XsX (and PC) but I am wondering if they are making any specific customizations that taps into the XsX SSD as well. 🤔
I think goal of the MS always would be, that it would be handle by Direct X somehow automatically : D
 
I think goal of the MS always would be, that it would be handle by Direct X somehow automatically : D

Thats what all Dev kit / APIs are meant to do.
The more thats done behind the scenes the better- which has been a big goal in all the newer hardware.
Its also nice when they allow you low level access but if the unreal engine is ported TO the SDK for the XSX
then yes... it would be utilizing its max throughput in any instance where the assets provided by the dev
are taking advantage of the compression acceleration. You would expect any competent dev would do that,
as if its packaged right it basically makes free bandwidth unless it turns out there is some kind of BUS bottleneck
or something else we're not seeing yet.
 
Im assuming the demo isn’t 100 gigs or larger so it could all be ran from this special 4.8GB/s pool, right?

Yah and taps into all of the XsX SSD custom configurations as described in this thread and by MSFT themselves. I mean you cant believe everything anyone says even if its coming from MSFT and Sony Engineers and programmers themselves without some hype factor especially when making outlandish claims. For example:

Krazy Ken Kutaragi: "The Cell processor's processing power is = 1% of the human brain" 🙄🙄🙄🙄🙄🙄🙄🙄🙄

But since he is an engineer an had collaboration with Toshiba, IBM to make that processor he knows what he is talking about and I must believe him because it came from the horses mouth!
 
Im assuming the demo isn’t 100 gigs or larger so it could all be ran from this special 4.8GB/s pool, right?

So here is my only contention. If this 100gb pool is a specific , lets call it a slice or a partition on the SSD-
Then if your game is NOT in there to begin with is it MOVED there at started? If not then it would be treated
as ram, so if the game was around 50gb, and you got (After other work being done behind the scenes during loading) lets
say best case 5gb a second, you're looking at 10 seconds loading it into that memory space, which is pretty good-
and then from there youre looking at moving it into memory, for actual access.
The pool of memory would be streamed into , and scrubbed out of physical cache by the "velocity" setup as what Sony
has mentioned.

This essentially works exactly like Sony has described its entire SSD working but I am not sure how / why this 100gb is special? It seems this 100gb cant be
used for storage but that maybe the OS resides there. It must be mapped as a page file but why its used as a swap/page and not direct from the disk files-
I assume they are going to go into later. It certainly would make sense as a sort of scratch space for assets.
 

Bernkastel

Ask me about my fanboy energy!
So . . . Actual specs are pure hype and speculation?

Sounds more like another secret sauce argument "Well the logo from the same company looks similar
so I assume this ~500 dollar box is going to have some technology from a high end server in it"

Again... sounds kind of like speculation to me. I'm just saying.
If Sony's specs are in hand, and Microsofts are too... what good is bringing up AMD's technology
if both systems are running the same AMD architecture? You might have a point if they were
not literally both possibly using the same thing.

And yeah.... it IS just a name basically. Whats wrong with me saying its a name?
If the NAMED TECHNOLOGY whatever it may be is saying "Look we're going to give you
access at 6gb compressed" it doesnt matter what its called, its not going to be faster or
as fast as something that can do 11gb compressed.
Its not in any AMD desktop CPU/GPU. Its in AMD Radeon Pro SSG, which is for data centers and has a 2 TB hardware decompression block. Microsoft does not use AMD's solution, they use their own thing in Azure, from which they derived the velocity architecture for XSX. And then SFS is a 2-3x multiplier of that.
 
Last edited:

rntongo

Banned
How so?

Neither is out. The manufacturers have given their specs. My only original argument is
Microsofts speculative, promised specs are still half of Sonys promised specs (using speculative as she did, in this case)
and if some reference to AMD architecture is made.... Sonys system is also AMD so why even say so?

Without writing too much, the PS5 has revealed much more about their SSD and I/O architecture. MSFT has not yet revealed all the details but it's clear the PS5 system and in particular the SSD is far superior. Especially if they soldered the SSD onto the board. However, the Velocity Architecture, if it achieves what MSFT is claiming, would be able to offer clever and innovative ways of diminishing or eliminating advantages the PS5 SSD has despite having half the throughput. For example, a higher decompression ratio and significantly more efficient texture streaming. That's the fundamental argument.
 

M1chl

Currently Gif and Meme Champion
Thats what all Dev kit / APIs are meant to do.
The more thats done behind the scenes the better- which has been a big goal in all the newer hardware.
Its also nice when they allow you low level access but if the unreal engine is ported TO the SDK for the XSX
then yes... it would be utilizing its max throughput in any instance where the assets provided by the dev
are taking advantage of the compression acceleration. You would expect any competent dev would do that,
as if its packaged right it basically makes free bandwidth unless it turns out there is some kind of BUS bottleneck
or something else we're not seeing yet.
I am not disagreeing with you on anything.
 
Its not the same token though. The customizations of XSX SSD are not available in the PS5 SSD. It is a custom hardware decompression blockm that can deliver 6 GB/s, not the general 4.8 GB/s compressed. And then SFS "is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance."(its in the OP).
I/O Throughput2.4 GB/s (Raw), 4.8 GB/s (Compressed, with custom hardware decompression block)

I feel like the title should have been "Xbox Velocity Architecture - 100 GB is instantly accessible by the developer with custom hardware decompression block of 6 GB/s."

It can.

Its not marketing jargon. Its a concept both Microsoft and AMD have used for data centers. The PS5 SSD jargon of replacing all specs with SSD sounds more like marketing mare to me. And it was introduced before "Road to PS5".
It is not. stop lying . Not a single megabyte over 4.8gb/s . End of discussion
 
Its not in any AMD desktop CPU/GPU. Its in AMD Radeon Pro SSG, which is for data centers and has a 2 TB hardware decompression block. Microsoft does not use AMD's solution, they use their own thing in Azure, from which they derived the velocity architecture for XSX.

So you say... And again derived.
Why did you bring up AMD at all then?

Since azure is a cloud service and not a hardware service per se, Can you point out where Microsoft created this customer hardware
and what it is called, what its specs are, etc? I am honestly just asking because this is a lot of nothing without a source.
 

Panajev2001a

GAF's Pleasant Genius
Without writing too much, the PS5 has revealed much more about their SSD and I/O architecture. MSFT has not yet revealed all the details but it's clear the PS5 system and in particular the SSD is far superior. Especially if they soldered the SSD onto the board. However, the Velocity Architecture, if it achieves what MSFT is claiming, would be able to offer clever and innovative ways of diminishing or eliminating advantages the PS5 SSD has despite having half the throughput. For example, a higher decompression ratio and significantly more efficient texture streaming. That's the fundamental argument.

You are making this “claim” up though, MS is not claiming a 2-3x or more multiplier over PRT and page prefetching done in software.

This is dGPU level of “we invented DirectX, no way the gap would be this wide”, “c’mon there cannot be a power greater than X, the gap cannot be negative for Xbox and still be wide”, etc... Admitting that would be admitting PS5 is not this inferior half step some want it to appear as, but a console that made different trade offs with their engineering R&D.
 
Without writing too much, the PS5 has revealed much more about their SSD and I/O architecture. MSFT has not yet revealed all the details but it's clear the PS5 system and in particular the SSD is far superior. Especially if they soldered the SSD onto the board. However, the Velocity Architecture, if it achieves what MSFT is claiming, would be able to offer clever and innovative ways of diminishing or eliminating advantages the PS5 SSD has despite having half the throughput. For example, a higher decompression ratio and significantly more efficient texture streaming. That's the fundamental argument.

Everything you said except to note,

The number of lanes is whats important not if or not its soldered to the board in the case that its going through a controller chip anyway. This isnt the case where
youre trying to get more memory channels by putting individual chips on different channels rather than having slots. It could be in a slot, a socket ,or soldered. That doesnt change
performance at all- we already know its not an off the shelf SSD though and of course its going to be soldered. Or at least- it would surprise me if it werent, but
it would not change its performance.
 

rntongo

Banned
So here is my only contention. If this 100gb pool is a specific , lets call it a slice or a partition on the SSD-
Then if your game is NOT in there to begin with is it MOVED there at started? If not then it would be treated
as ram, so if the game was around 50gb, and you got (After other work being done behind the scenes during loading) lets
say best case 5gb a second, you're looking at 10 seconds loading it into that memory space, which is pretty good-
and then from there youre looking at moving it into memory, for actual access.
The pool of memory would be streamed into , and scrubbed out of physical cache by the "velocity" setup as what Sony
has mentioned.

This essentially works exactly like Sony has described its entire SSD working but I am not sure how / why this 100gb is special? It seems this 100gb cant be
used for storage but that maybe the OS resides there. It must be mapped as a page file but why its used as a swap/page and not direct from the disk files-
I assume they are going to go into later. It certainly would make sense as a sort of scratch space for assets.

1.) The 100GB refers to up to 100GB of an actual game install on the SSD, not a special partition.
2.) Theoretically the PS5 should be able to access data directly on the SSD like the XSX. It has a custom processor for mapping so it should be possible.
 

Bernkastel

Ask me about my fanboy energy!
So you say... And again derived.
Why did you bring up AMD at all then?

Since azure is a cloud service and not a hardware service per se, Can you point out where Microsoft created this customer hardware
and what it is called, what its specs are, etc? I am honestly just asking because this is a lot of nothing without a source.
I literally mentioned Project Denali in the previous page. I brought up AMD, because its not a fancy term but an actual technology that is used in data centres.
Since azure is a cloud service and not a hardware service per se,
Becaue Azure runs on data centres, which "stores data".
 

Entroyp

Member
Its not the same token though. The customizations of XSX SSD are not available in the PS5 SSD. It is a custom hardware decompression blockm that can deliver 6 GB/s, not the general 4.8 GB/s compressed. And then SFS "is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance."(its in the OP).
I/O Throughput2.4 GB/s (Raw), 4.8 GB/s (Compressed, with custom hardware decompression block)

I feel like the title should have been "Xbox Velocity Architecture - 100 GB is instantly accessible by the developer with custom hardware decompression block of 6 GB/s."

It can.


Its not marketing jargon. Its a concept both Microsoft and AMD have used for data centers. The PS5 SSD jargon of replacing all specs with SSD sounds more like marketing mare to me. And it was introduced before "Road to PS5".

Replacing all specs eh.. yea that sounds right...
 

rntongo

Banned
You are making this “claim” up though, MS is not claiming a 2-3x or more multiplier over PRT and page prefetching done in software.

This is dGPU level of “we invented DirectX, no way the gap would be this wide”, “c’mon there cannot be a power greater than X, the gap cannot be negative for Xbox and still be wide”, etc... Admitting that would be admitting PS5 is not this inferior half step some want it to appear as, but a console that made different trade offs with their engineering R&D.
You're making claims up. I said a 2-3x multiplier on RAM and SSD performance. We know PRTs do not work as well as whatever solution MSFT is going to use in the XSX because their data shows that on the Xbox One whole assets went into RAM that were never used. SFS is supposed to eliminate this.
 

Bernkastel

Ask me about my fanboy energy!
If the claim in red is accepted, then so is the 22 GB/s Kraken decompressor one... ;).
PbeGEnf.png
 
1.) The 100GB refers to up to 100GB of an actual game install on the SSD, not a special partition.
2.) Theoretically the PS5 should be able to access data directly on the SSD like the XSX. It has a custom processor for mapping so it should be possible.

If so why is it limited to 100GB?
 

DeepEnigma

Gold Member
PbeGEnf.png
If the claim in red is accepted, then so is the 22 GB/s Kraken decompressor one... ;).

And then it was so!
 
I literally mentioned Project Denali in the previous page. I brought up AMD, because its not a fancy term but an actual technology that is used in data centres.

Becaue Azure runs on data centres, which "stores data".


So literally you are saying its Microsofts Fancy term for an EXISTING actual technology that is not exclusive to Microsoft but
is used by the same company that makes the tech in.... BOTH consoles.

"Runs on data centers" As a technical person this aches to read.

A data center is a building or room in a building which is specialized to house
servers and storage systems. Azure runs on the servers. Those servers are made by
companies like HP, Dell, IBM, Quanta, Sun/Oracle, Fujitsu and others but those are most of the
big ones (not withstanding "custom" servers built off the shelf like supermicro type....

They run an OS like Linux and Azure runs as software on those servers.

And project denali is a standard.... Actually everything I read has been saying that the whole point of that
is in cloud solutions to move the overhead off the drive and controller to the HOST of the system (onto the CPU)
and this means in this case onto a storage controller. That is what Sony is doing too. Except twice as fast.
 
That 100GB referred to the game size as far as I recall, so it can as well be 40 or 150, depends on how much space the game takes. That's how I understand it.

So it doesn't actually mean anything then. There is nothing special about that size or space. Its just them saying "The game will be basically as fast as the SSD
to access and load into the faster memory on the system"? That sounds like a bunch of extra steps to describe the storage.
 

Panajev2001a

GAF's Pleasant Genius
You're making claims up. I said a 2-3x multiplier on RAM and SSD performance.

You are claiming this, you are failing to back it up. MS themselves said it clearly that SF is better but not a never seen before magic wand (2-3x multiplier would be gigantic) and beyond a custom texture filtering mode there is zero solid evidence for your claim that 2-3x performance is a speed up on the best PRT + software page prefetching can do on an RDNA1 GPU let alone these semi custom RDNA2 ones (despite what is or not officially part of DX specs).
2y8Jr63.png


Beyond SF you are in the secret dGPU land or “inventors of DirectX could not possibly allow a gap this wide in this metric”. You are not happy Xbox has faster overall memory, higher clocked CPU, and faster GPU shaders... nope they also must have a very small gap SSD wise.

Curiously, this effort on minimising the gap came after the past week or so, something happened to change the marching band tune of “oh come on... it will only get you 1-2s faster loading times”... I wonder what happened to that narrative, it is almost unreal ;).

We know PRTs do not work as well as whatever solution MSFT is going to use in the XSX because their data shows that on the Xbox One whole assets went into RAM that were never used. SFS is supposed to eliminate this.

No, they are implying that for understood marketing reasons... else they would be claiming ALL titles had at least a 2-3x multiplier. You know this though and at this point is sounding more and more disingenuous.
 

Bernkastel

Ask me about my fanboy energy!
So literally you are saying its Microsofts Fancy term for an EXISTING actual technology that is not exclusive to Microsoft but
is used by the same company that makes the tech in.... BOTH consoles.

"Runs on data centers" As a technical person this aches to read.

A data center is a building or room in a building which is specialized to house
servers and storage systems. Azure runs on the servers. Those servers are made by
companies like HP, Dell, IBM, Quanta, Sun/Oracle, Fujitsu and others but those are most of the
big ones (not withstanding "custom" servers built off the shelf like supermicro type....

They run an OS like Linux and Azure runs as software on those servers.

And project denali is a standard.... Actually everything I read has been saying that the whole point of that
is in cloud solutions to move the overhead off the drive and controller to the HOST of the system (onto the CPU)
and this means in this case onto a storage controller. That is what Sony is doing too. Except twice as fast.
Off-Shelf computers wont have a 2 TB hardware decompression block. Again, Microsoft is not using AMD's solution. Project Denali is way more than just custom hardware decompression block. AMD's solution and Project Denali are not meant for personal computers or home consoles. Xbox engineers used Project Denali research to derive something for home consoles.
Cerny never talked about a custom hardware decompression block in his deep dive even though a significant portion of it was dedicated to SSD. So, we can presume Sony didnt derive anything from AMD's data center tech.
 
Last edited:

rntongo

Banned
You are claiming this, you are failing to back it up. MS themselves said it clearly that SF is better but not a never seen before magic wand (2-3x multiplier would be gigantic) and beyond a custom texture filtering mode there is zero solid evidence for your claim that 2-3x performance is a speed up on the best PRT + software page prefetching can do on an RDNA1 GPU let alone these semi custom RDNA2 ones (despite what is or not officially part of DX specs).
2y8Jr63.png


Beyond SF you are in the secret dGPU land or “inventors of DirectX could not possibly allow a gap this wide in this metric”. You are not happy Xbox has faster overall memory, higher clocked CPU, and faster GPU shaders... nope they also must have a very small gap SSD wise.

Curiously, this effort on minimising the gap came after the past week or so, something happened to change the marching band tune of “oh come on... it will only get you 1-2s faster loading times”... I wonder what happened to that narrative, it is almost unreal ;).



No, they are implying that for understood marketing reasons... else they would be claiming ALL titles had at least a 2-3x multiplier. You know this though and at this point is sounding more and more disingenuous.
We're running in circles where you make false equivalencies. You are really good at making false equivalencies to be honest.
 
Off-Shelf computers wont have a 2 TB hardware decompression block. Again, Microsoft is not using AMD's solution. Project Denali is way more than just custom hardware decompression block. AMD's solution and Project Denali are not meant for personal computers or home consoles. Xbox engineers used Project Denali research to derive something for home consoles.

Off shelf servers can have whatever controller they want inside it that is available to buy. That includes whatever technology is on the controller.
Dont think server hardware is something weak or its like a PC. A data center doesnt use systems that are very much like a home PC at all.
Microsoft isnt using any server technology in the XSX. Those decompression blocks in a server environment are feeding a raid array made of up
multiple channels with multiple SSDs on each.

Once again I insist - I am not sure what point youre trying to make? The specs are from Microsoft. The device is half the speed of the specs of the
device Sony is using. Done. End. It doesnt matter what you reference, what its derived from or the fact that it has a cool name.

If Microsoft says it can do 5GB compressed and Sony says it does that much UNCOMPRESSED and it can also compress in hardware its not a contest.
Even if Sony had NO HARDWARE COMPRESSION/DECOMPRESSION the raw speed of Sonys SSD matches these specs.... which would be great for MS
to just force devs to use this compression to get a similar result. Well Sony has a hardware decompression chip too. Now what?

Better bring up the name, that'll fix it.
 

Bernkastel

Ask me about my fanboy energy!
Off shelf servers can have whatever controller they want inside it that is available to buy. That includes whatever technology is on the controller.
Dont think server hardware is something weak or its like a PC. A data center doesnt use systems that are very much like a home PC at all.
Microsoft isnt using any server technology in the XSX. Those decompression blocks in a server environment are feeding a raid array made of up
multiple channels with multiple SSDs on each.

Once again I insist - I am not sure what point youre trying to make? The specs are from Microsoft. The device is half the speed of the specs of the
device Sony is using. Done. End. It doesnt matter what you reference, what its derived from or the fact that it has a cool name.

If Microsoft says it can do 5GB compressed and Sony says it does that much UNCOMPRESSED and it can also compress in hardware its not a contest.
Even if Sony had NO HARDWARE COMPRESSION/DECOMPRESSION the raw speed of Sonys SSD matches these specs.... which would be great for MS
to just force devs to use this compression to get a similar result. Well Sony has a hardware decompression chip too. Now what?

Better bring up the name, that'll fix it.
This is literally what they said.
Our second component is a high-speed hardware decompression block that can deliver over 6GB/s.
 

rntongo

Banned
If so why is it limited to 100GB?
We're yet to find out. I read an article where either Phil or Jason Ronald clearly states that it's 100GB of a game install. When I find it I'll reply to you here. There is no advantage in allocated a part of the SSD to move over a game install for dynamically loading data.
 
This is literally what they said.

Yes. And? Sonys words are that theirs does 6gb raw. They also have a compression block.

""By the way, in terms of performance, that custom decompressor equates to nine of our Zen 2 cores, that's what it would take to decompress the Kraken stream with a conventional CPU," Cerny reveals. "

A powerful one by the sound of it.
 
We're yet to find out. I read an article where either Phil or Jason Ronald clearly states that it's 100GB of a game install. When I find it I'll reply to you here. There is no advantage in allocated a part of the SSD to move over a game install for dynamically loading data.

That was my question, why are we talking about a GB amount at all? It sounds like a lot of weird confusing jargon.

"You can load game data from the SSD without seek times, at X GBPS" is the sentence a normal person would put together.
but that doesnt make it sound special.

Thats what the PS5 is doing too... but faster in the case of the PS5.
 
Last edited:
If you want more answers best bet is probably the Hot Chips XSX System Architecture presentation August 17th.

well the Xbox velocity architecture is a group of tech(that each can exist separately), the memory controller for example is harware, SFS is an API that may be assisted partly by GPU, other parts are an API that is also coming to windows, velocity architecture may include unique tech for xbox(specially the controller) I am not saying it doesnt(quite the contrary) but SFS specifically is already in directX12(or at least something with the same name and description), rntongo says Xbox implementation its different from DX12(wich its curious as imply MS games on PC wont benefit from it even in RDNA2 GPUs in the brand new DX12) implying its much better to the point it close the gap to a SSD solution 2 times faster(even if by definition it only targets textures and there is no indication its gains are over already similar solutions), that is interesting but by your response I suppose it is a misunderstanding and as you say we will have to wait for a presentation to know about this
 

FranXico

Member
Yes. And? Sonys words are that theirs does 6gb raw. They also have a compression block.

""By the way, in terms of performance, that custom decompressor equates to nine of our Zen 2 cores, that's what it would take to decompress the Kraken stream with a conventional CPU," Cerny reveals. "

A powerful one by the sound of it.
It's 5.5 raw. And their decompressor block can output up to 22 GB/s. In practice, these theoretical maxima never happen.
 
It's 5.5 raw. And their decompressor block can output up to 22 GB/s. In practice, these theoretical maxima never happen.

Yes. But its the spec. If we're arguing spec for spec and considering likely similar overhead since we're going to be comparing apples to apples workloads,
then 5.5 raw and much more compressed... why even mention theoretical maximums? Microsoft is subject to the same real world application vs paper
differences.
 

rntongo

Banned
That was my question, why are we talking about a GB amount at all? It sounds like a lot of weird confusing jargon.

"You can load game data from the SSD without seek times, at X GBPS" is the sentence a normal person would put together.
but that doesnt make it sound special.

Thats what the PS5 is doing too... but faster in the case of the PS5.

Here they’re talking about giving the CPU direct access to data on the SSD up to 100GB. The PS5’s I/O architecture looked setup to stream everything to RAM but it’s possible the XSX has a DMA from the CPU to SSD. The reasons for why up to 100GB could be explained either by someone more knowledgable or later on when more information is released.
 
Last edited:
Top Bottom