• 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

THE:MILKMAN

Member
I still think the performance difference will lean more towards the CUs number more the th TF number deference

I agree the 18% difference is ultimately what we'll get/see. That's it. No magic required.

I find it really amusing how Sony and Microsoft's roles have reversed in what is important with specs looking back at articles at the beginning of this gen.
 

FALCON_KICK

Member
Maybe she simply prefers fixed platforms. There are more things to take into account if you want to develop for Android for example. She's still developing for Windows though, so... I don't know. We can ask her lol.


It can be a lot for a college student. Depending on where she's from, that might be 3 months or more of her whole monthly income.
She is actually a he according to SenjutsuSage SenjutsuSage
 

CobraXT

Banned
You typically reserve space on the drive for this, or a dedicated partition. If you did this to your whole drive, well then you would probably not have many games there :)

You can map an entire 64bit address space to a disk. That's a pretty large disk.

ok .. but isn't this approach will cause ssd to fail much faster ? because you have to write gpu usable data in a portion in ssd every time you open a game ? i am sure ssd isnt not as durable as ddr/gddr ram for constant data writing
 
ok .. but isn't this approach will cause ssd to fail much faster ? because you have to write gpu usable data in a portion in ssd every time you open a game ? i am sure ssd isnt not as durable as ddr/gddr ram for constant data writing
They are much better nowadays.
This approach is used in Databases everywhere.
 

oldergamer

Member
Well this is the patent that Ronaldo from b3d was referring to. https://patents.google.com/patent/US10388058B2/en

I haven't read through it yet, but If you look at the dates, this was first filed in 2017 and granted in Aug 2019. The patent is owned by microsoft, so whatever is in there, i'm assuming isn't something that older hardware was capable of, but I don't know for certain yet. I'd suggest anyone read through to see what can be surmised.

"These improvements include hardware residency map features and texture sample operations referred to herein as “residency samples,” among other improvements."
 
Last edited:

Handy Fake

Member
I still think the performance difference will lean more towards the CUs number more the th TF number deference

I'm probably wrong here...

But I don't think it affects speed in the traditional sense as it does on say a PC graphics card?
I think the entire IO architecture is built around increased frequency so rather than just seeing a bump in clock speed on the GPU, it's an exponential bump throughout the entire system.
 
You also have to keep the CUs busy too, it’s more difficult to use more CUs than a single fast CU. I think the Series X will flex it’s muscles here anyway as more and more engine developers build better job systems, like Id, or Epic
 

oldergamer

Member
P Panajev2001a just for this feature it does look like MS added specific hardware as part of SFS. This diagram shows the addtion of a residency sampler that in turn creates a residency map:

FIG. 2 illustrates graphics processing system 200 in an implementation. System 200 includes graphics processor 220 which further includes hardware cache 221. Graphics processor 220 can be an example of graphics processor 120, although variations are possible. Graphics processor 220 can provide graphics processing and rendering services for a system processor that indicates data and instructions over link 250, such as system processor 130 in FIG. 1. Cache 221 comprises a memory or storage area incorporated into graphics processor 220 for storing usage-based data and components for graphics processor 220. Cache 221 can store cached tiles 240 and residency map 241, among other items. In some examples, cache 221 comprises a hardware-based cache memory in the hardware components that from graphics processor 220, such as on-die cache memory.

As mentioned above, graphics processor 220 can include one or more stages or internal components that can process content/data into rendered images for display. These elements can include those shown in FIG. 1 for graphics processing unit (GPU) elements 260, although it should be understood that other configurations are possible. Elements 260 include shader 261, texture memory 262, texture unit 263, tile map 264, residency map 241, and residency sampler 266.

Texture memory 262 can include texture data that is provided by a content source for use by graphics processor 220. Texture memory 262 can serve as a tile cache that stores tiles for associated partially resident textures that are employed in rendering. Tiles can be any discrete unit/region of a texture map. Cached tiles 240 are included in FIG. 2 representative of one or more resident portions of texture data within a cache portion of graphics processor 220. Shader 261 determines color information during the rendering process, and in some examples comprises pixel shading features. Shader 261 can communicate with texture unit 263 to determine colors/shading for regions of textures, which can be referred to as texels. Shader 261 can indicate level of detail (LOD) thresholds to texture unit 263, which is used for associated tile handling processes. The LOD thresholds can include a detail level needed for a particular tile, which can be based on various factors, including 3D-viewability factors. LOD will be discussed in more detail below.

Tile map 264 specifies what tiles and detail levels are stored in texture memory 262, and can comprise one or more tables or other data structures to track the information in texture memory 262. Values in tile map 264 indicate integers corresponding to levels of detail that are resident in texture memory 262. Tile map 264 may optionally be implemented as a cache that is filled with contents from an area of texture memory 262.

Residency map 241 comprises a buffer in memory that exists alongside tiled resources referred to as PRT (partially-resident textures). Residency map 241 indicates a combined “A.B” representation of a mipmap level of detail resident in texture memory 262 (“A”) along with a fractional value employed in a filtering clamp (“B”). The residency map is provided to graphics processor 220 along with the PRT in the same “descriptor,” with the PRT being the primary data source and the residency map being the secondary “metadata” data source. The residency map is a relatively compact data structure that represents one data sample per (u,v) rectangular region of the PRT, and can be held in entirely hardware—such as cache 221. Each data sample in the residency map can represent predetermined region of the PRT, such as a rectangular region of the PRT.

Residency sampler 266 comprises logic that performs a sample from residency map 241. Residency sampler 266 generates memory addresses that reach the page table hardware in graphics processor 220, but do not continue on to become full memory requests. Instead, the residency of the PRT at those addresses is checked and recorded in residency log 267. Residency log 267 can be read by a designated entity so that unused pages in texture memory 262 can be discarded and missing pages in texture memory 262 are filled.



US10388058-20190820-D00002.png


Stupid image is rotated for some dumb reason.
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
P Panajev2001a just for this feature it does look like MS added specific hardware as part of SFS. This diagram shows the addtion of a residency sampler that in turn creates a residency map:

FIG. 2 illustrates graphics processing system 200 in an implementation. System 200 includes graphics processor 220 which further includes hardware cache 221. Graphics processor 220 can be an example of graphics processor 120, although variations are possible. Graphics processor 220 can provide graphics processing and rendering services for a system processor that indicates data and instructions over link 250, such as system processor 130 in FIG. 1. Cache 221 comprises a memory or storage area incorporated into graphics processor 220 for storing usage-based data and components for graphics processor 220. Cache 221 can store cached tiles 240 and residency map 241, among other items. In some examples, cache 221 comprises a hardware-based cache memory in the hardware components that from graphics processor 220, such as on-die cache memory.

As mentioned above, graphics processor 220 can include one or more stages or internal components that can process content/data into rendered images for display. These elements can include those shown in FIG. 1 for graphics processing unit (GPU) elements 260, although it should be understood that other configurations are possible. Elements 260 include shader 261, texture memory 262, texture unit 263, tile map 264, residency map 241, and residency sampler 266.

Texture memory 262 can include texture data that is provided by a content source for use by graphics processor 220. Texture memory 262 can serve as a tile cache that stores tiles for associated partially resident textures that are employed in rendering. Tiles can be any discrete unit/region of a texture map. Cached tiles 240 are included in FIG. 2 representative of one or more resident portions of texture data within a cache portion of graphics processor 220. Shader 261 determines color information during the rendering process, and in some examples comprises pixel shading features. Shader 261 can communicate with texture unit 263 to determine colors/shading for regions of textures, which can be referred to as texels. Shader 261 can indicate level of detail (LOD) thresholds to texture unit 263, which is used for associated tile handling processes. The LOD thresholds can include a detail level needed for a particular tile, which can be based on various factors, including 3D-viewability factors. LOD will be discussed in more detail below.

Tile map 264 specifies what tiles and detail levels are stored in texture memory 262, and can comprise one or more tables or other data structures to track the information in texture memory 262. Values in tile map 264 indicate integers corresponding to levels of detail that are resident in texture memory 262. Tile map 264 may optionally be implemented as a cache that is filled with contents from an area of texture memory 262.

Residency map 241 comprises a buffer in memory that exists alongside tiled resources referred to as PRT (partially-resident textures). Residency map 241 indicates a combined “A.B” representation of a mipmap level of detail resident in texture memory 262 (“A”) along with a fractional value employed in a filtering clamp (“B”). The residency map is provided to graphics processor 220 along with the PRT in the same “descriptor,” with the PRT being the primary data source and the residency map being the secondary “metadata” data source. The residency map is a relatively compact data structure that represents one data sample per (u,v) rectangular region of the PRT, and can be held in entirely hardware—such as cache 221. Each data sample in the residency map can represent predetermined region of the PRT, such as a rectangular region of the PRT.

Residency sampler 266 comprises logic that performs a sample from residency map 241. Residency sampler 266 generates memory addresses that reach the page table hardware in graphics processor 220, but do not continue on to become full memory requests. Instead, the residency of the PRT at those addresses is checked and recorded in residency log 267. Residency log 267 can be read by a designated entity so that unused pages in texture memory 262 can be discarded and missing pages in texture memory 262 are filled.



US10388058-20190820-D00002.png


Stupid image is rotated for some dumb reason.

My God, you are really digging into this :messenger_clapping: lovely some parents reading ;).
 

oldergamer

Member
P Panajev2001a just for this feature it does look like MS added specific hardware as part of SFS. This diagram shows the addtion of a residency sampler that in turn creates a residency map:

FIG. 2 illustrates graphics processing system 200 in an implementation. System 200 includes graphics processor 220 which further includes hardware cache 221. Graphics processor 220 can be an example of graphics processor 120, although variations are possible. Graphics processor 220 can provide graphics processing and rendering services for a system processor that indicates data and instructions over link 250, such as system processor 130 in FIG. 1. Cache 221 comprises a memory or storage area incorporated into graphics processor 220 for storing usage-based data and components for graphics processor 220. Cache 221 can store cached tiles 240 and residency map 241, among other items. In some examples, cache 221 comprises a hardware-based cache memory in the hardware components that from graphics processor 220, such as on-die cache memory.

As mentioned above, graphics processor 220 can include one or more stages or internal components that can process content/data into rendered images for display. These elements can include those shown in FIG. 1 for graphics processing unit (GPU) elements 260, although it should be understood that other configurations are possible. Elements 260 include shader 261, texture memory 262, texture unit 263, tile map 264, residency map 241, and residency sampler 266.

Texture memory 262 can include texture data that is provided by a content source for use by graphics processor 220. Texture memory 262 can serve as a tile cache that stores tiles for associated partially resident textures that are employed in rendering. Tiles can be any discrete unit/region of a texture map. Cached tiles 240 are included in FIG. 2 representative of one or more resident portions of texture data within a cache portion of graphics processor 220. Shader 261 determines color information during the rendering process, and in some examples comprises pixel shading features. Shader 261 can communicate with texture unit 263 to determine colors/shading for regions of textures, which can be referred to as texels. Shader 261 can indicate level of detail (LOD) thresholds to texture unit 263, which is used for associated tile handling processes. The LOD thresholds can include a detail level needed for a particular tile, which can be based on various factors, including 3D-viewability factors. LOD will be discussed in more detail below.

Tile map 264 specifies what tiles and detail levels are stored in texture memory 262, and can comprise one or more tables or other data structures to track the information in texture memory 262. Values in tile map 264 indicate integers corresponding to levels of detail that are resident in texture memory 262. Tile map 264 may optionally be implemented as a cache that is filled with contents from an area of texture memory 262.

Residency map 241 comprises a buffer in memory that exists alongside tiled resources referred to as PRT (partially-resident textures). Residency map 241 indicates a combined “A.B” representation of a mipmap level of detail resident in texture memory 262 (“A”) along with a fractional value employed in a filtering clamp (“B”). The residency map is provided to graphics processor 220 along with the PRT in the same “descriptor,” with the PRT being the primary data source and the residency map being the secondary “metadata” data source. The residency map is a relatively compact data structure that represents one data sample per (u,v) rectangular region of the PRT, and can be held in entirely hardware—such as cache 221. Each data sample in the residency map can represent predetermined region of the PRT, such as a rectangular region of the PRT.

Residency sampler 266 comprises logic that performs a sample from residency map 241. Residency sampler 266 generates memory addresses that reach the page table hardware in graphics processor 220, but do not continue on to become full memory requests. Instead, the residency of the PRT at those addresses is checked and recorded in residency log 267. Residency log 267 can be read by a designated entity so that unused pages in texture memory 262 can be discarded and missing pages in texture memory 262 are filled.



US10388058-20190820-D00002.png
Stupid image is rotated for some dumb reason. anyway, I'd be willing to bet this does not exist on any other hardware as of yet.
 
Stupid image is rotated for some dumb reason. anyway, I'd be willing to bet this does not exist on any other hardware as of yet.
In a way it's just another expression of this.


HW PRT moved page tables and filtering from software to hardware. A residency map is essentially the feedback buffer mentioned there. It seems to just be an additional hw buffer.
 

Ar¢tos

Member
Hehe maybe thats what he meant.... I meant his actual maths working out the percentage and his factor comparison with PS4.
What he meant is something else, only the digits are important, it's a psychological play.
Ps4 is 40% stronger that the X1, but the difference between ps5 and xsx is 18%, which is a non appealing number, all other differences are under 40%, so the difference in CU count, in percentage, completely ignoring the clocks, is now the important one because it's above 40, and it's the number themselves that count.
It's sad, but it is what fanboys cling to.
 

longdi

Banned
I think he is comparing apples to oranges, Sony’s approach seems roughly similar to the approach MS took with XVA with likely more logic spent to accelerate SSD I/O and higher throughout (GB/s).
XVA is a marketing term for the solution Sony have no name to, maybe they should have called it Lighting Data Transfer Architecture or Infinite RAM Architecture ;).

The block Cerny described in his presentation seems not to be trivial designed, but something they spent quite a bit of transistors and R&D time on, resources they did not add to the GPU hence the TFLOPS gap... they are probably happy as they were able to drive a bigger gap in I/O than the one they lost to XSX in the TFLOPS war).

Tbf, Sony lost more than tflops war this round. No? Series X has opened gaps in all areas except the 16 channel SSD lanes.
MS designed the SeX to win, to sell at a loss, to win over core gamers.
Sony designed PS5 like PS4, at least not to sell at a loss.
 

oldergamer

Member
Tbf, Sony lost more than tflops war this round. No? Series X has opened gaps in all areas except the 16 channel SSD lanes.
MS designed the SeX to win, to sell at a loss, to win over core gamers.
Sony designed PS5 like PS4, at least not to sell at a loss.
well... I don't know know if it will win over core gamers. I hope it does like the xbox 360 did. They have a much better focus time out of the gate (games), and leveraged a lot of the research they do in graphics hardware and PC technology. Funny enough, the time they were most successful in selling consoles was when they went an entirely custom hardware route.

I actually don't know if they will sell it at a loss, or if its more expensive then what sony has designed. Until the xbox one x arrived, ms went a cost effective route and lost out due to hardcore gamer mind share and lower performance. To me this feels more like a rehash of the xbox 360 where they came out of the gate the right way. They didn't execute correctly the last three years (kinect was neat, but the focus was too much on that and TV towards the end of the 360.).That mistake i don't see them repeating. Also they have far more value in first party then they previously did.
 

Panajev2001a

GAF's Pleasant Genius
MS designed the SeX to win

So did Sony, designing it like PS4 seems like a winning formula. Call me grumpy, but not too much of a fan of these kind of slogans / marketing platitudes...

Tbf, Sony lost more than tflops war this round. No? Series X has opened gaps in all areas except the 16 channel SSD lanes.
MS designed the SeX to win, to sell at a loss, to win over core gamers.
Sony designed PS5 like PS4, at least not to sell at a loss.

Not sure XSX will sell at as much a loss as you are predicting, but beyond the TFLOPS war (and the RAM bandwidth to feed it) by about 18% or so, the CPU side is likely even closer. It does not help to analyse and enjoy the breakthroughs MS has delivered if there is this constant marketing / console warrior spin...
It is annoying. Took one small section in the post as seemingly an excuse to start beating your chest about the XBX...
 
Last edited:
So, Sony pushed a GPU with variable 2.23 Ghz in 3 months just like that since MS confirmed 12 TF in December last year. Yeah, no! It doesn't make any sense what you said at all. Btw. they've pantented their cooling solution back in August last year. Doesn't mean if Github didn't have any data for 2.23 GHz, Sony pushed GPU to 2.23 GHz after that. It doesn't make any sense.
I'm not buying the "panicked Sony" theory either.

If the PS5 already had a cooling system that could handle those clocks then I don't see why they would go with lower clocks in the first place.

I apologize if I ever made it sound like I inferred Sony panicked in reaction to MS and increased PS5's GPU clock. That's not what I was saying, or trying to imply. My point was that they planned for a 36 CU system at high clocks, that's it. However, "high clocks" has taken on a whole new meaning these days. The 2 GHz clock we saw on Ariel and Oberon were already very high for a GPU, they just look tame now compared to PS5's actual retail GPU clock.

That said, I honestly do think the 2.23 GHz clock was a result of them shifting to a different power strategy. It has never been concluded if they intended on variable frequency; in fact Cerny basically confirms in a way they were looking at pushing high clocks on a fixed frequency setup but had troubles in doing so, which led to them adapting variable frequency). That may not have been the specific clock they were shooting for.

As for what may've motivated the push to 2.23 GHz? I think it was just Sony internally discussing the best means to squeeze out more performance. Higher GPU clock means you need that much more efficient cooling, but that better cooling costs money. So do faster GDDR6 modules. We already know Oberon can support faster modules but Sony went with 14 Gbps ones. I'm under the belief there was a discussion between the Japanese and American sides on which route to take, and ultimately they decided to up the GPU clock a bit further by employing a variable frequency power strategy, in lieu of 16 Gbps GDDR6 chips (the DRAM price increases would make going that route even less attractive).

So they always had the means of flexibility to push the clocks higher and switch to variable frequency, but may've been considering alternatives until certain market realities pushed them to take the route they've taken. None of this needs any influence from MS or XSX to have occurred, either.

excellent wall of text

That was a good post to come around when lurking x3; basically says what should be assumed. MS and Sony have taken very different means of resolving the same bottlenecks, with each approach having some advantages and disadvantages compared to the other, but both ultimately being very valid methods and serving to wider corporate strengths and interests between the platform holders and their technology partners (including each other, in fact).

It's unfortunate that immature console warring mentality convinces people to pit the options against each other to prove one as absolutely, monstrously superior to the other, when the reality is nowhere NEAR as simple as that. It's even funnier when some of those types cling to the paper specs in doing this, yet when it comes to the paper specs of other system components, we are told to "see the full picture" and be mindful actual real-world performance may make the paper specs less relevant...which goes towards paper specs of ANY system components, mind, but it's an element of irony that pops up time and again.

Because clocks mean nothing naturally, the more CUs the better, even if they were clocked at 800mhz.

I taste facetiousness here ;). No one says higher clocks don't have benefits to performance. However, they are HARDLY the only thing that matters when it comes to a GPU's performance capabilities. Clocks mainly influence pixel fillrate and the speed the caches operate at, but many equally vital aspects of GPU performance are not strongly benefited by higher clockrates.

You also have to keep the CUs busy too, it’s more difficult to use more CUs than a single fast CU. I think the Series X will flex it’s muscles here anyway as more and more engine developers build better job systems, like Id, or Epic

It's not that much more difficult to use more CUs vs. a single faster CU. I mean, these are GPUs, they are inherently designed to perform massive operations in parallel. If AMD's frontend improvements on RDNA2 are as big as efficiency improvements (and there's no reason to doubt that being the case at this time), that combined with general leaps of skill gains by developers in targeted GPU programming and parallelism techniques should make targeting a wider net of CUs much easier compared to past platforms.

Also I kind of feel it goes without saying if the actual GPU manufacturers are pushing for bigger GPUs with even higher rates of parallelism, they probably feel efficient targeting of a larger array of CUs (or equivalents for non-AMD GPU architectures) is plenty capable, otherwise they would not stress a push for larger GPUs in the first place (and I know there are other reason they go for larger chips, such as the ability to sell them at higher prices for larger profit margins, but let's isolate the financial case from the technological case in this instance as it really doesn't have too much bearing here).

My God, you are really digging into this :messenger_clapping: lovely some parents reading ;).

So you ask for someone to provide evidence, they provide evidence, then you say they've provided too much evidence?

I mean c'mon xD.
 
I apologize if I ever made it sound like I inferred Sony panicked in reaction to MS and increased PS5's GPU clock.

That's ok.

I really don't think that it's a panicked reaction because I don't believe Sony wants to run into severe issues with their product like Microsoft did with the 360.

Some individuals are using that theory to suggest that the PS5 will have severe hardware issues at launch. While I do believe hardware issues can happen I don't believe Sony will increase the clocks if it's going to cause a RROD type of situation.
 
Last edited:

Bernkastel

Ask me about my fanboy energy!
I seriously dont know whats going on for the last 18 pages.
Title updated by request to the mods because 4.8 GB/s is just BCPack general compressed data speed. Also I have added more details to the OP. Any new details I need to add ?
 
Last edited:

Ascend

Member
I seriously dont know whats going on for the last 18 pages.
Title updated by request to the mods because 4.8 GB/s is just BCPack general compressed data speed. Also I have added more details to the OP. Any new details I need to add ?
Basically we're trying to figure out what XVA does. Some people claim we already know everything about it, more than the PS5 in fact, but I do not share that opinion. So I'm probing the internet for any additional info. But I think there isn't really anything else left to find at this point.

She is actually a he according to SenjutsuSage SenjutsuSage
Ok. Quite possible. The info that person is providing is interesting regardless of his/her gender, and I can't see any other possibility of what the statement of instantly accessible 100GB could mean.
 
Last edited:

Bernkastel

Ask me about my fanboy energy!
Ok. Quite possible. The info that person is providing is interesting regardless of his/her gender, and I can't see any other possibility of what the statement of instantly accessible 100GB could mean.
Louise Kirby is a dude
WNtSS0u.png

He basically continued the conversation after that confirming what I said.
 
Last edited:
Ascend Ascend I think you have to let go of that person as any sort of credible source. The ten minutes I looked into them, I found empty github repos of some "engine" they are creating (which showed evidence of this being the first time they used a VCS), some dubious "roadmap" of features that are just pasted from various presentations floating around, a possible vertex shaded cube demo (i.e: a tutorial found online opened and ran), and some pretty crappy vb.net code. I suspect this person is learning to programme in college, as that that's what they teach. People like to bluff a bit on the net, which is fairly common of younger people who are enthusiastic about tech. Their thoughts and opinions are like some weird scrapbook of understanding that ultimately isn't really reflective of actual game dev experience, and more crawling around the internet too much as a layman. I mean really, ask yourself. How much could they really know about this stuff while being in college, while producing no output to backup anything they have said? This isn't some genius insight, it's like reading an autocomplete sentence. I think in the search for information you probably latched onto what sounded good, but you didn't really understand from a technical perspective, which is fine. I've been programming for a couple of decades now, and game dev is hard!

They just don't make much sense.
 
Last edited:
I've been programming for a couple of decades now, and game dev is hard!

Alot of people are using Kirby as a source of information but the truth is that seasoned developers opinions are alot more valid than his.

I'm not saying that people can't learn anything from Kirby but this individual hasn't even worked as a real developer yet and still learning the career field.

On top of that he's still pretty young and is a bit immature given the individuals that this person associates with.

It's similar as relying on the developer of Pirates of the First Star as a source of information.

pirates-of-first-star-xbox-one.jpeg
 
Last edited:

Bernkastel

Ask me about my fanboy energy!
Alot of people are using Kirby as a source of information but the truth is that seasoned developers opinions are alot more valid than his.

I'm not saying that people can't learn anything from Kirby but this individual hasn't even worked as a real developer yet and still learning the career field.

On top of that he's still pretty young and is a bit immature given the individuals that this person associates with.

It's similar as relying on the developer of Pirates of the First Star as a source of information.

pirates-of-first-star-xbox-one.jpeg
pirates-of-first-star-xbox-one.jpeg
You are complaining about Dynamic Voltage Games. You know whats shitty ? Fucking this


This video is based off a ResetEra post(he does not mention ResetEra posts just says "forum post") that has been reposted in this very thread. He even mentioned that Crytek guy who only worked on Crytek for a year and before that was in a Iranian Robotics company and browser game dev.
Nope, majority are 3rd party devs. Do some research instead of spewing crap. So, cut the crap
So, uh, about that "Sony Dev React" thing that is still going after 26 pages. From what I've gathered on this thread :

Kurt Margeneau (Naughty Dog) ---> Sony dev
Anthony Newman (Naughty Dog) --> Sony dev
James Cooper (no studio currently) --> Not Sony dev anymore
Andrew Maximov (Promethean AI) --> Not Sony dev anymore
Randy Pitchford (Gearbox) --> Not Sony dev
Matt Philips (Big Evil Corp) --> Not Sony dev
Mike Evans (ex AMD) --> Not Sony dev
Robert Boyd (Zeboyd Games) --> Not Sony dev
Andrea Pessino (Ready at Dawn) --> Not Sony dev
Dale North (composer) --> Not Sony dev
Olivier JT (VR audio) --> Not Sony dev
Jason Nuyens (Breakfall) --> Not Sony dev
Also, DICE developer is missing
This list includes Matt Philips, who makes homebrew games for Sega Megadrive and many smaller indie devs like Dynamic Voltage Games. I made a bigger list and that does not include Louis Kirby and Dynamic Voltage Games. Unlike the ResetEra list, a lot of them have Xbox Series X dev kits and are making games for Xbox Series X.
So, far what I have gathered on what devs think of Xbox Series X
Daniele Galante and David Garcia, Ninja Theory Audio Team, Ninja Theory
Series X’s dedicated audio chip will mean they no longer have to sacrifice sound in order to facilitate other parts of their projects.
Erwan LeCun, co-founder of Ape Tribe Games
Joel Baker, Technical Director at Hinterland Games
Ray-Tracing is one of the most potent features, with big benefits for weather effects and immersion on the horizon. NVMe SSD tech and the custom hardware surrounding it to help get data into memory faster.
Gennadiy Korol, Director of Technology at Moon Studios
Playing games at a high refresh rate, at those crazy resolutions, is the next big thing in gaming.
Kevin Floyer-Lea, Chief Technology Officer at Rebellion
Excited to be able to hit 4K 60 FPS as standard. Hardware-accelerated ray tracing is also a very welcome feature. SSD storage takes beautiful realistic environments and make them load in a flash
Mike Rayner, Studio Technical Director at The Coalition
SSD speeds improve load times without any code changes. DirectStorage APIs and hardware decompression on the Xbox Series X are also key factors in further improving loading speeds. Sampler Feedback tech, which is part of the Xbox Velocity Architecture, loads textures more efficiently based on what the scene needs, further reducing the load on memory.
Alexandre Sabourin, Team Lead at Snowed In Studios
Hardware-accelerated ray-tracing, improved processing power, and SSD boosts will help the Xbox Series X.
Grant Kot, working on a Real-Time Particle Physics based game for Xbox Series X and high end PC's
Xbox Series X is more powerful than his 8700K RTX 2080 PC.
Richard Geldreich, worked at Space X, Valve and Ensemble
Xbox Series X's BCPack Compression technique is better than PS5's Kraken
Chris Grannel, ex-Sony first party dev who worked on Formula One Series, Wipeout HD:Fury and Killzone 2 for Playstation.
Xbox Series X caught Sony off-guard. Microsoft has been working a little bit closer with AMD on some of the technologies that they have been working on. Xbox Series X can do better real time Ray Tracing because of the higher 52 CU count compared to Playstation 5's 36. Microsoft has got some pluses like Xbox Series X's 320 bus compared to Playstation 5's 256 bus. As a developer he would rather work on fixed performance deltas.
Johannes Kuhlmann, head of core technology at FishLabs
It's easier to develop for Xbox Series X than any other console.
Ljubomir Peklar, designer at Ebb Software
Biggest change between this generation and the next is on the pure power afforded by the CPU at the heart of Series X.
Tor Frick and Arcade Berg, co-founders of Neon Giant
They are doing 4K 60 fps which is blazingly fast. Faster load speeds – it just comes out, magically, you just click, and there you go, it loads way faster. They have more detail than you can actually see at lower resolutions. Draw distances will be as long as they can make them. When the camera changes position following a vicious gunfight, you should be able to count the bullet-holes in the walls. Everything is ludicrously detailed, so when you bring down the camera, it just works.
Microsoft is very open and approachable, they know what it takes to be able to make a game for their machines, so they’re prepared – the tools are there. Series X represents a leap in video game development, it’s becoming easier and easier to develop. So it’s easier and easier to actually use all that power, and to spend more time on the creative bits, because ideally, you just want everything to work, to realise your creations. An 11-person indie studio could not have done what they are doing 20 years ago.
Edge Magazine, July 2020 Issue 346, Page 54-65

This will get Off-Topic so I put a spoiler tag.
 
Last edited:
I seriously dont know whats going on for the last 18 pages.
Title updated by request to the mods because 4.8 GB/s is just BCPack general compressed data speed. Also I have added more details to the OP. Any new details I need to add ?

So here is a quote from James Stanard this morning when asked why the decompressor block was rated at 6+GB/s when the system was only rated for 4.8GBps:

"It comes down to what is typical vs what is attainable. 4.8 is conservative. Ideal circumstances can yield higher speeds, but we don’t want to set unrealistic expectations."
 
Basically we're trying to figure out what XVA does. Some people claim we already know everything about it, more than the PS5 in fact, but I do not share that opinion. So I'm probing the internet for any additional info. But I think there isn't really anything else left to find at this point.


Ok. Quite possible. The info that person is providing is interesting regardless of his/her gender, and I can't see any other possibility of what the statement of instantly accessible 100GB could mean.

Whats interesting is that person provides more insight (that can be backed up) then most of the people criticizing him...
 
You are complaining about Dynamic Voltage Games. You know whats shitty ? Fucking this


This video is based off a ResetEra post(he does not mention ResetEra posts just says "forum post") that has been reposted in this very thread. He even mentioned that Crytek guy who only worked on Crytek for a year and before that was in a Iranian Robotics company and browser game dev.

This list includes Matt Philips, who makes homebrew games for Sega Megadrive and many smaller indie devs like Dynamic Voltage Games. I made a bigger list and that does not include Louis Kirby and Dynamic Voltage Games. Unlike the ResetEra list, a lot of them have Xbox Series X dev kits and are making games for Xbox Series X.

This will get Off-Topic so I put a spoiler tag.


Most of the sources are solid but no idea why you put trust in Chris Grannel after the fallout.
 
I meant his past works.

But he isn't a game developer now is he?

Plus being out of the game for so long probably means his knowledge hasn't been updated with today's technology.

I could probably get good answers of I asked him about the PS3 but the PS5 is a different story.

In reality most of what he says about these systems is second hand information something that we can't ever prove. Which is similar to Jason when tells us about things that developers told him.
 

Bernkastel

Ask me about my fanboy energy!
But he isn't a game developer now is he?

Plus being out of the game for so long probably means his knowledge hasn't been updated with today's technology.

I could probably get good answers of I asked him about the PS3 but the PS5 is a different story.

In reality most of what he says about these systems is second hand information something that we can't ever prove. Which is similar to Jason when tells us about things that developers told him.
The only reason I made that list was because somebody posted that ResetEra list in this thread and I believe I did a better job. I dont think we can completely remove his opinion, it still counts. Also, its getting off-topic.
 
Last edited:

truth411

Member
No he was right .. % increase = 36 PS5 CUs - 52 XSX CUs / 36 = 16 / 36 = 0.44 %
1. If im reading his post correctly, is he saying the PS4 die is bigger than the Xbox one die because of the number of CUs. If so then thats false because the ESRAM takes up alot of Die space.

Edit: when comparing the original Xbox one with the Original PS4.

2. His post literally ignores clock speeds as if it doesn't matter.
 
Last edited:

Fafalada

Fafracer forever
Series X has opened gaps in all areas
I realize ppl love lists on internet but... what "areas" are we talking here? Faster graphics subsystem encompasses memory and the related checkboxes, its not a set of 'areas' - if you removed something like extra bandwidth, the gap would effectively vanish.

The .1 ghz faster cpu is... another 'thing' I guess?
 

Bernkastel

Ask me about my fanboy energy!
I realize ppl love lists on internet but... what "areas" are we talking here? Faster graphics subsystem encompasses memory and the related checkboxes, its not a set of 'areas' - if you removed something like extra bandwidth, the gap would effectively vanish.

The .1 ghz faster cpu is... another 'thing' I guess?
I didnt know PS5 has a 3.7 GHz CPU.
 
Last edited:
I realize ppl love lists on internet but... what "areas" are we talking here? Faster graphics subsystem encompasses memory and the related checkboxes, its not a set of 'areas' - if you removed something like extra bandwidth, the gap would effectively vanish.

The .1 ghz faster cpu is... another 'thing' I guess?

We need more insight into the PS5 but for now things in additon to whats being discussed in this thread

112 GB VRAM bandwidth advantage

Mesh shaders

Automatic Auto HDR without impacting CPU or GPU

13 TF of 16 bit float for DirectML also can leverage 97 tops of 4bit integer concurrently with the 12TF shader array.

Lots more to learn such as the ROP count, how 100GB Vram works etc etc.

I know you used to be a Sony dev Faf from what I remember of your talks on B3d.. Are you an Xbox dev now as well? (Maybe you always were and I didn't know.)
 
Last edited:
Top Bottom