MasterCornholio
Member
No I thought the laptop was running it at 1440P. It explains why the laptop with a weaker GPU was getting higher framerates than that PS5 demo.
-
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
- Thread starter Bernkastel
- Start date
- Platform
No I thought the laptop was running it at 1440P. It explains why the laptop with a weaker GPU was getting higher framerates than that PS5 demo.
Agnostic2020
Member
Xsx will have better compressed performance at 4.8 gb/s yes. So if the compression technique of sfs is good then xsx can theoretically out perform the Samsung evo 970
100 gig part is just to say instant access and zero seek time . Ssd by nature has instant access and zero seek time. Its just a marketing buzz. all ssd have instant access to the disk drive .
thicc_girls_are_teh_best
Member
Right, I'll take your word on this versus highly educated engineers at a trillion-dollar megaconglomerate...
You may work in the field, but you're a person like any of the rest of us. So you probably have your preferences when it comes to certain hardware features and whatnot. In any case, I'd wait until their Hot Chips presentation rather than making potentially wrong conclusions.
As for ECC, I wasn't implying it's a specific server market technology. But if MS have mentioned adding ECC features to their GDDR6 memory for a reason, then I'll take their word on it. I do the same with many features Sony specifies, as well. But out of the two, it's MS who probably have the desire for a server market-compliant system so "Nothing about the XSX makes it suitable as a server..." really doesn't have a lot of substance to back it up. At the very least, there's at least some aspects of it being designed with that market in mind, because why wouldn't it?
So about this from the other thread:
"It is easy. It is useless to have 12 boxes if they do not fit through the door all together.
You have 12 boxes to fill. So you can't pass all the boxes at once. You must decide which boxes will pass and which will not. That is handled by a coordinator. And the coordinator tells the delivery man which boxes to take.
Mrs. XSX wants to make the move as soon as possible, but it turns out that only 8 boxes can fit on the door at a time. The coordinator is fast, and also uses a box compressor so that 10 boxes can go through instead of 8, but there are several drawbacks. The compressor can only compress the red boxes, and the coordinator also has to coordinate many other things, street traffic, people passing through the door, the space in the room where the boxes are stored, the noise of neighbors who distract the delivery man, search and select what the boxes are filled with, etc. Also, the delivery man is not so fast and is very distracted filling and transporting boxes. So it passes the 10 boxes (not 12) at a certain speed "1x". The lady demands that the boxes arrive, but they do not arrive as quickly as the lady would like, since although she has many boxes, the system is not capable of managing all of them properly.
On the other hand we have Mrs. PS5. You only have 10 boxes to fill. But its door is twice as big, enough for all its boxes to enter at once and there is room for people to also enter and exit through the door. Furthermore, the coordinator has the ability to automatically discard unnecessary boxes, so he doesn't waste time checking boxes that are not going to be used. In addition, anyone in the environment can do the job of the coordinator or the delivery man (even at the same time). The compressor is not that new, but it can compress all boxes, whether they are red or blue. All. And the delivery man is more than twice as fast and manages to pass the boxes at the speed of "2.5x" in the worst case, and "5x" on many occasions. In addition, if someone is left free or without work, they can help to distribute boxes with the delivery man or coordinate work with the coordinator. All this makes this removal company the most efficient ever seen and that the number of boxes available is irrelevant. For that moving system, 12 boxes are not needed, with 10 you can do the same job (and more or better in some cases). Having more boxes would only make the price of the move more expensive without needing any of it.
Of course, having more boxes available always helps to advertise yourself as a top removal company compared to the competition, even if your removal company is normal and ordinary. But it is only that, a smokescreen.
That does not mean that XSX is bad, far from it, it is an extraordinary machine. But PS5 has an efficiency NEVER seen before."
For XsX for this anology from the developer, did he take into account:
The subset of customizations specific to XsX? I am assuming he did?! Does this correlate with the video @BrentonB posted?
I've seen a couple of this guy's vids; they're generally informative but they also jump the gun on a lot of their mentions. In one of their recent vids they made some mentions about XSX that were already proven to be the other way around, though TBF they addressed those in a pinned comment.
Still though, it would always be preferable to get that info correct before committing to video, because only a fraction of the viewers will actually read the pinned information. Also I kind of question how they structure their graphs; in the pic you screencapped you can see the difference in XSX raw SSD speed and 970's raw SSD speeds, which in reality we know is a 800 MB delta.
So that amount we can assume is 800 MB in visual representation. The difference between PS5's SSD raw speed and XSX's is 3.1 GB, or 800 MB x 3.875. However you can clearly see the visual representation of that is actually MUCH greater than it should be, if the scale were being kept equivalent to the visual representative delta between the XSX's raw speed and the 970's.
There's usually a psychological reason why people would manipulate visual representation of data in that way; an observer will perceive any differences more prominently if they are visually communicated, especially in an exaggerated manner. Visual perception is easier to reconcile with many people than numerical or textual ones, because people nowadays are tuned to emotional responses and visual data strongly favors emotional responses and thinking over logical responses/thinking.
Basically, that one single screen shot tells me a lot more about the actual intentions behind the poster's video and the message/agenda they have with regards it, than they probably thought it ever would. It might seem like I'm making a mountain out of a molehill but IMO it's very important to mention these kind of things when they come up because I generally REALLY dislike visual manipulation of data to subconsciously lean a person one way over another. Clean data and clean visual representation of that data should be more than enough, if you trust people's own sense of individual judgement.
Plus, unfortunately the outright manipulation of visual data in other areas by the MSM have kind of made me more alert to notice when other people (even outside of areas like politics) do it, too, because when it's done it's generally for the same reason: to emotionally sway impressionable people to buy into their own bias/preference messaging, which usually hints at an agenda. Again I'm not knocking this Youtube channel specifically, but I noticed immediately what was going on with that screencap and had to talk about it.
I think the 4.8 GB/s is for the compressed speed rate; we don't know (well, okay. I don't know
) what the 970's compression data sequential read rate is, because that would rely on compression/decompression hardware which many PCs don't have. In other words, they tend to do it through software, but maybe there's a few SSDs that handle some of that on the drives themselves.The 100 GB stuff is in reference to something else. IMO it might have to do with a direct access link of that memory space to the GPU via DMA, maybe the 100 GB being a SLC NAND cache block, basically mirroring a scaled down implementation of AMD's SSG cards. That's just speculation, though, nothing confirmed on that front.
...and we gotta wait almost three damn months to hopefully get confirmation too
Last edited:
I'd liken SFS in the XSX to the PS5's use of variable clock rates for the CPU/GPU. Finding a tricky innovative design to achieve better performance. I'm inclined to believe SFS will be able to achieve closer to it's goal due to the nature of the task and other parts of the system architecture aimed at supporting it. It's clear with the work on BCPack and having different memory bandwidth RAM(A cost decision as well) that will hold textures that MSFT sees textures as the biggest hoarder of RAM and if they cut it down they can achieve results similar to a higher throughput SSD such as the PS5.
This video is misleading by claiming the XSX won't reach 4.8GB/s throughput. It's highly misleading.So about this from the other thread:
"It is easy. It is useless to have 12 boxes if they do not fit through the door all together.
You have 12 boxes to fill. So you can't pass all the boxes at once. You must decide which boxes will pass and which will not. That is handled by a coordinator. And the coordinator tells the delivery man which boxes to take.
Mrs. XSX wants to make the move as soon as possible, but it turns out that only 8 boxes can fit on the door at a time. The coordinator is fast, and also uses a box compressor so that 10 boxes can go through instead of 8, but there are several drawbacks. The compressor can only compress the red boxes, and the coordinator also has to coordinate many other things, street traffic, people passing through the door, the space in the room where the boxes are stored, the noise of neighbors who distract the delivery man, search and select what the boxes are filled with, etc. Also, the delivery man is not so fast and is very distracted filling and transporting boxes. So it passes the 10 boxes (not 12) at a certain speed "1x". The lady demands that the boxes arrive, but they do not arrive as quickly as the lady would like, since although she has many boxes, the system is not capable of managing all of them properly.
On the other hand we have Mrs. PS5. You only have 10 boxes to fill. But its door is twice as big, enough for all its boxes to enter at once and there is room for people to also enter and exit through the door. Furthermore, the coordinator has the ability to automatically discard unnecessary boxes, so he doesn't waste time checking boxes that are not going to be used. In addition, anyone in the environment can do the job of the coordinator or the delivery man (even at the same time). The compressor is not that new, but it can compress all boxes, whether they are red or blue. All. And the delivery man is more than twice as fast and manages to pass the boxes at the speed of "2.5x" in the worst case, and "5x" on many occasions. In addition, if someone is left free or without work, they can help to distribute boxes with the delivery man or coordinate work with the coordinator. All this makes this removal company the most efficient ever seen and that the number of boxes available is irrelevant. For that moving system, 12 boxes are not needed, with 10 you can do the same job (and more or better in some cases). Having more boxes would only make the price of the move more expensive without needing any of it.
Of course, having more boxes available always helps to advertise yourself as a top removal company compared to the competition, even if your removal company is normal and ordinary. But it is only that, a smokescreen.
That does not mean that XSX is bad, far from it, it is an extraordinary machine. But PS5 has an efficiency NEVER seen before."
For XsX for this anology from the developer, did he take into account:
The subset of customizations specific to XsX? I am assuming he did?! Does this correlate with the video @BrentonB posted?
Kazekage1981
Member
yah something about it wont be fully saturated, and not being saturated due to the other anology of the developer stating that you cant get the boxes on time due to distraction or something.
In other words:
idontbelieveyou.gif (not you, but the analogy and the video)
Last edited:
thicc_girls_are_teh_best
Member
This is what's looking will be the case. Granted, none of this means XSX's SSD I/O will suddenly completely close the gap with PS5's. Sony's approach is just beefier overall, it'll always have an advantage since it has more physical/hardware "headroom" to do so. Very much like how no matter what optimizations PS5 has on the GPU side, XSX's will have the advantage because it simply has beefier/more hardware to it.
That said, I think a lot of people are under the impression PS5's approach to resolving I/O bottlenecks is the only approach, and that's where they get it wrong. There's ALWAYS been multiple viable solutions to any given problem in the tech world, some being hardware-focused and others software-based. Just for example, if there were ever only one clear-cut answer to the "problem" of computing architectures, why do literally hundreds of architectures exist? Yes, we're mostly familiar with x86, but you've got POWER, ARM, RISC, and plenty of others.
Or take polygons/3D. Before we ever got dedicated 3D hardware in systems like PS1 and Saturn (and tbh, it's a bit debatable if they were "real" 3D the way we know 3D today (they both lacked Z-buffering for example, though N64 had it), devs could already do 3D on systems like SNES through custom ASICs, or MegaDrive through the CPU. You don't NEED dedicated silicon for 3D visuals and we almost got a taste of that again with the PS3 (before they switched to using Nvidia); while in the past that usually meant poorer 3D I think if you a 3D programming language specifically targeted at high-end super-server CPUs, it would probably be pretty capable, not to mention utilize techniques not bound by dedicated hardware.
It's not really surprising Sony and MS have taken divergent approaches to improving storage I/O on next-generation architectures, but the implication one approach is immutably superior to the other (and by order of magnitudes more) is immature; we should HOPE the overall SSD I/O performance with the two systems is pretty close all told even if one still holds the edge, because that ultimately benefits the vast majority of game devs, 3rd-parties.
loktaiextatus
Member
See some replies in line. I dont mean to write a novel I just want to be clear, its nothing about personal preferences.
Its more like talking to an F1 driver and saying "Civics Si are made like race cars, the commercial implies it" And the race driver saying no.., an F1 car
and a civic have so little in common you cant compare them and the civic Si is just a juiced up commuter.
The ECC could be for anything - it could be for something that isnt the same reason servers use it. I am telling you outright an APU is absolutely
not suitable for a server. For many reasons. Fixed NON REDUNDANT memory, power and storage are not suitable for a server . I can describe a basic server to you.
2 processors, and they usually are not the same thing as what a desktop CPU has. They are never APUs. They are socketed for removal, or upgrade. They have larger than
normal cache and they each have their own memory slots. a server can have a lot of memory slots, 32 would not be weird. A server today on the low end will have 64 gigs
but many have 128 or more . The storage is a hardware accelerated, battery backed redundant RAID setup with multiple disks. Sometimes it a mirror because it only runs
the OS locally and isnt used for storing data directly, either it has another purpose or does so on a SAN through fiber or iSCSI. It has a good number of expansion slots
for such cards and comes with multiple NICs and a remote administration controller like iDRAC, iLO, etc.
A server also has the ability to fail acomponent like a memory dimm and keep running. Or a FAN, and keep running, depending on the system you can lose a couple fans
and the system wont shut down or overheat, it ramps the other fans up to compensate while it awaits repair.
It has multiple power supplies, a mid range 2U server has 2 generally and a bigger 4U systems or something else may have 4, I wont get into storage arrays with
multiple dual powered redundant shelf systems but lets stick with the enterprise servers....
Ok I am not close to naming all the things a server must have... but lets go down the list.
Does the Xbox have a remote management port?
No.
Does it have redundant power supplies?
No.
Redundant fans?
No
Redundant memory or removable memory?
No
A server type CPU(s) (Never an APU)
No.
Does it accept expansion PCIE cards ?
No
Does it support hardware RAID or contain a hardware RAID write cache?
No
Does it have a lot of memory ?
No.
No. Its much closer to gaming PC engineered to fit inside a small shell for the livingroom. It has nothing in common with a server.
Dont take this as me trying to be an ass but its just not server like in ANY aspect.
Kenpachii
Member
So about this from the other thread:
"It is easy. It is useless to have 12 boxes if they do not fit through the door all together.
You have 12 boxes to fill. So you can't pass all the boxes at once. You must decide which boxes will pass and which will not. That is handled by a coordinator. And the coordinator tells the delivery man which boxes to take.
Mrs. XSX wants to make the move as soon as possible, but it turns out that only 8 boxes can fit on the door at a time. The coordinator is fast, and also uses a box compressor so that 10 boxes can go through instead of 8, but there are several drawbacks. The compressor can only compress the red boxes, and the coordinator also has to coordinate many other things, street traffic, people passing through the door, the space in the room where the boxes are stored, the noise of neighbors who distract the delivery man, search and select what the boxes are filled with, etc. Also, the delivery man is not so fast and is very distracted filling and transporting boxes. So it passes the 10 boxes (not 12) at a certain speed "1x". The lady demands that the boxes arrive, but they do not arrive as quickly as the lady would like, since although she has many boxes, the system is not capable of managing all of them properly.
On the other hand we have Mrs. PS5. You only have 10 boxes to fill. But its door is twice as big, enough for all its boxes to enter at once and there is room for people to also enter and exit through the door. Furthermore, the coordinator has the ability to automatically discard unnecessary boxes, so he doesn't waste time checking boxes that are not going to be used. In addition, anyone in the environment can do the job of the coordinator or the delivery man (even at the same time). The compressor is not that new, but it can compress all boxes, whether they are red or blue. All. And the delivery man is more than twice as fast and manages to pass the boxes at the speed of "2.5x" in the worst case, and "5x" on many occasions. In addition, if someone is left free or without work, they can help to distribute boxes with the delivery man or coordinate work with the coordinator. All this makes this removal company the most efficient ever seen and that the number of boxes available is irrelevant. For that moving system, 12 boxes are not needed, with 10 you can do the same job (and more or better in some cases). Having more boxes would only make the price of the move more expensive without needing any of it.
Of course, having more boxes available always helps to advertise yourself as a top removal company compared to the competition, even if your removal company is normal and ordinary. But it is only that, a smokescreen.
That does not mean that XSX is bad, far from it, it is an extraordinary machine. But PS5 has an efficiency NEVER seen before."
For XsX for this anology from the developer, did he take into account:
The subset of customizations specific to XsX? I am assuming he did?! Does this correlate with the video @BrentonB posted?
RIP xbox series X, its confirmed guys it can't even hit 2Gbps. Also dat PC and its shitty 1,5Gbps samsung evo. What will the 2-3x sized v-ram pool at 2x+ the speed with 3x more gbps pool ( uncompressed ram vs PS5 compressed ) at 1000 times faster access speed ever do with that 1,5gbps samsung evo that can probably load in data all day long into the memory ready for the next 3 levels while the PS5 is trying to slam as much as it can in its tiny 8gb memory pool on the fly choking to death.
Also 3rd party developers are totally not going to target the lowest common dominator like they always do, totally not guys. Which would change that entire graph in a complete 180%
Also good luck storing (24) 8k textures on a 700GB ssd. can't wait on the download PSN file sizes and 5 blu-rays game boxes.
Last edited:
loktaiextatus
Member
Kind of funny if you think about it . . .
And dont get me wrong I am all for a good debate but the Microsoft camp is screaming at the top of their lungs that double the read
speed from the systems storage is not a big deal, that the developers will target the common denominator (Or dominator I guess if you're into that ....)
or that it will be impractical in some way to actually utilize...
But man.. let me tell you that 15 percent higher floating point operations per second on the GPU is going to be a total game changer. SONY IS FINISHED. OWNNNNEDDD
thicc_girls_are_teh_best
Member
Once again, ALL of these are your assumptions. I take it MS have invited you to get a look at the entire system's architecture, otherwise these are just tales.
The XSX APU is meant to serve as a basis for the server implementation, this should be more than obvious. What you're saying is analogous to, say, if SEGA revealed the MegaDrive specs ahead of System 16, and then someone said "there's zero way they're using MegaDrive in an arcade system because nothing about it is for the arcade market".
Which ignores that various aspects of the adaptation would be modified to serve that very same market. Who says the server version of XSX doesn't use larger RAM capacity modules? Who says it doesn't support hardware RAID? Who says the server version has no way of supporting expansion PCIe cards?
The truth is the server version could support those things, but since there'd be no need for the home version to do so (and since they have been sharing info on just the home version thus far), those things would not be present, nor need to be spoken up on.
You should read my other post that goes into the psychological aspect of the graph
Kazekage1981
screencapped and how it was visually manipulated to suspend logical reasoning and push an emotional response with its own narrative.You want people to not take paper specs of TFs for an end-all, be-all, you should expect them to not take paper specs of the SSD I/O as an end-all, be-all, either. It's only fair, and many of us are providing very solid grounds of constructive speculation while still acknowledging realities to it all.
loktaiextatus
Member
See part of the reply in-line above.Once again, ALL of these are your assumptions. I take it MS have invited you to get a look at the entire system's architecture, otherwise these are just tales.
The XSX APU is meant to serve as a basis for the server implementation, this should be more than obvious. What you're saying is analogous to, say, if SEGA revealed the MegaDrive specs ahead of System 16, and then someone said "there's zero way they're using MegaDrive in an arcade system because nothing about it is for the arcade market".
Which ignores that various aspects of the adaptation would be modified to serve that very same market. Who says the server version of XSX doesn't use larger RAM capacity modules? Who says it doesn't support hardware RAID? Who says the server version has no way of supporting expansion PCIe cards?
The truth is the server version could support those things, but since there'd be no need for the home version to do so (and since they have been sharing info on just the home version thus far), those things would not be present, nor need to be spoken up on.
Its not the XSX Apu. Its AMD's APU - and yes of course the system is going to have things in common - but the things that make the xbox series X the XBOX SERIES X arent useful
in a server. The APU itself is near useless, its the CPU side And yes of course there will be CPUs built in the same generation as the Xbox - that doesnt mean they will be derived from
the XBOX's chip. As another example the original xbox uses a coppermine 733mhz CPU if Im not mistaken. That CPU is from like late 1999 but the coppermine and Tualatin P3s
were used in servers as the Xeon line was born just after that. Previous to that there was the Pentium Pro and such but... the Pentium 3 was SO good you could use it as it was
especially because some of them had more cache on board. even just a few years ago I would come across 1st generation HP DL380 systems with those Pentium 3 CPUs
and old IBM 1U systems as well, and they would still be running. It doesnt mean the XBOX was like a server... it just had a CPU in common.
In this case they wont even most likely be using the APU it'll just be the Zen 2 based 7nm processors. No more XBOX like than PS5-like or Desktop-like.
It just sounds real nice to talk about it in those terms of course.
And this is exactly what I am talking about. "if SEGA revealed the MegaDrive specs ahead of System 16, and then someone said "there's zero way they're using MegaDrive in an arcade system because nothing about it is for the arcade market" But they didnt. the Megadrive came out YEARS after the system 16. Hell the Megadrive came out after the X68000.
Not a great analogy because the genesis 68000 CPU was in SO many things- but if you look at what makes the Genesis UNIQUE.... right.
The argument you're kind of standing behind is that the System 16 and the Genesis are somehow like a Macintosh desktop computer.
You should read my other post that goes into the psychological aspect of the graph
You want people to not take paper specs of TFs for an end-all, be-all, you should expect them to not take paper specs of the SSD I/O as an end-all, be-all, either. It's only fair, and many of us are providing very solid grounds of constructive speculation while still acknowledging realities to it all.
I never said anything about be-all end all. I said we have solid examples of a terraflop gap of say, 50 percent in the current gen Pro and X, and a bit less of a gap
in the PS4 and Xbox One. And then we have these coming out. I didnt say its no difference but 15-20 percent in the GPU I think will change the experience less than
moving data at double the speed. I could be wrong but its what I feel.
Last edited:
JareBear: Remastered
Banned
I always feel like X and Pro comparisons are false equivalencies. No game is built from the ground up for either in mind. The PS4 is the more powerful lead console platform this gen and that showed from the very beginning (BF4 being noticeably less sharp on XBO than PS4 at launch, cross-gen multi-plat title). If you think Spidey and God of War look great on Pro then you have to admit they would look even better on X, that is proven with multi-plats (sometime drastically so), but the base XBone being the base for first party MS games will never ever let the X truly shine. Hell, even the Pro doesn't get to truly shine, IMO. Neither mid-gen refresh was ever fully utilized imo.
I said this before, but I am not convinced that storage bandwidth should be weighted equally as CPU/GPU raw render ability in the overall rendering process. It is great, and is a marked advantage in the PS5 architecture, but I know where CPU/GPU power can translate directly into higher resolution. maybe higher settings, etc.
I have an idea now of what storage bandwidth can do but I am not convinced that twice the storage bandwidth will translate into areas of an image that are twice as detailed or geometry that is truly twice as complex throughout a game (when directly comparing XsX and PS5 versions of a game).
We will see. I believe the XsX's storage bandwidth is more than fast enough to do all sorts of cool new things for consoles next gen, and yes the PS5 will do it even better, but I don't know if the end results will look like a 2x advantage on your TV screen.
loktaiextatus
Member
I never argued nor will argue that the SSD will lead to higher fidelity but given this thread is about the "velocity architecture" and what it is capable of
thats sort of the topic of discussion . It does help in things like reusing assets vs loading new ones in, any jutter or lag time that may be involved or tricks that
would otherwise need to be used to stream or refresh assets, and in general its a nice thing to have.
That said when you mention things like the launch where the PS4 had,, well I guess lets do the math for equivalent.
1.84, you drop 28 percent to hit 1.31. So it was a 28 percent weaker GPU and what we saw was as you mentioned... Less sharpness .
They were running basically identically except for the sharpness if I am not wrong? Not that sharpness isnt important- it really is.
12.15 you drop 15 percent to hit 10.28. So the difference between them is about HALF what the difference between the PS4 and Xbox one was at launch.
The One X 6.0 down to the Pro 4.2 is also a 30 percent drop... an even bigger difference than the launch machines. So seeing a difference when
the systems are milked doesnt surprise me .
So basically imagine that the Xbox One launched at 1.55 Teraflops instead of 1.3, and its hard drive was twice the speed of PS4s,
and it had an advantage gpu clock (which may never matter but vertexes and some specific GPU math will be faster).
Im not saying it would tie them completely but it would make you say hmm" considering many people considered the xbox one a
decent proposition ANYWAY, imagine if it were much closer in power and had a massive I/O advantage... From that perspective the
PS5 is in a position the xbox one could have only dreamed of.
Obvious disclaimer at the end of the day they're all toys and they are for entertainment- but this is also part of that entertainment I think
for a lot of us... just playing devils advocate for a hobby computer.
Kazekage1981
Member
I think another way of understanding would be helpful: lets say that you implement Sony's fast SSD drive to PS3 (not PS4 and Xbone) would it increase graphics fidelity by:
-Level of detail
-Billions of triangles
at 1080p/30fps? Would games look better? For example would you be able to take a game like Uncharted 2:
And make it look like uncharted 4 because of: increase number of triangles and level of detail increasing graphics fidelity at 1080p/30fps:
God damn he looks sexy.
Because all I would understand is that: there would be no loading times, uncharted world would be bigger, no pop in, from what has already been established with the benefits of the SSD as a storage device.
-Level of detail
-Billions of triangles
at 1080p/30fps? Would games look better? For example would you be able to take a game like Uncharted 2:
And make it look like uncharted 4 because of: increase number of triangles and level of detail increasing graphics fidelity at 1080p/30fps:
God damn he looks sexy.
Because all I would understand is that: there would be no loading times, uncharted world would be bigger, no pop in, from what has already been established with the benefits of the SSD as a storage device.
Last edited:
loktaiextatus
Member
I think another way of understanding would be helpful: lets say that you implement Sony's fast SSD drive to PS3 (not PS4 and Xbone) would it increase graphics fidelity by:
-Level of detail
-Billions of triangles
at 1080p/30fps? Would games look better? For example would you be able to take a game like Uncharted 2:
And make it look like uncharted 4 because of: increase number of triangles and level of detail increasing graphics fidelity:
God damn he looks sexy.
Because all I would understand is that: there would be no loading times, uncharted world would be bigger, no pop in, from what has already been established with the benefits of the SSD as a storage device.
That isnt a perfect analogy. Youre talking about improving an existing games load times by making it use I/O the way it already does.
Now imagine Uncharted 2 without any of the limits of its RAM or GPU cache , imagine instead it can now stream assets so fast, by the time Drake turns around
there can be a whole new level loaded behind him. You know all those huge setpieces uncharted 2 and 3 have? That was done in HOW much memory?
Now imagine it could fill its memory, both system and GPU memory in a couple of seconds, and naughty dog knew this when developing it.
Kazekage1981
Member
Right, the SSD implementation was there from the get go system/launch of PS3
loktaiextatus
Member
Right. If it were there from inception they'd have developed games around it.
If they dont- its no different from smacking a faster drive into a PC and loading the game faster, it wont change much (unless you were already bottlenecked which isnt common due to how the PC loads and assumes slow storage).
If a developer knows its there, You can essentially load a lot more textures or models into gpu cache, or I should say, larger ones, without worring "this has used up all my GPU cache"
because now you can replace it, including the intelligent cache scrubbers to evict efficiently rather than flush the entirety.... and you can for example load a lot
nicer texture set into memory without worrying about leaving space for the textures in the next area, because they can be loaded so fast you wouldnt even see it assuming its been developed properly.
MasterCornholio
Member
Well I guess when multiplats get analyzed on both systems we can see what the differences really are when it comes to the I/O systems.
Will the PS5 load at double the speed or more? Or will the software enhancements make the gap alot smaller?
I guess we will have to wait for the comparisons to find out the answer.
However if there's one thing that im sure if is that most of the speed will come from the actual hardware. That's something that can't be changed at this point.
Last edited:
Kazekage1981
Member
Right designing the game from its inception. Wow thanks for the explanation, so it *is* possible to make it look close to Uncharted 4 at 1080/30fps.
loktaiextatus
Member
It would certainly be possible to have Uncharted 4 level textures, and certainly would be possible to have more assets in memory. It depends on what the bottleneck would be.
A better way to think about this would be ... rather than to say it makes a PS3 game look like PS4, or makes a PS5 game look like PS6, is to say that
the PS5 games we get will look like PS5 games- because that is what a PS5 is. And if you were to create a game for a PS5
and there were two alternate universes, one where the PS5 has a platter HDD or a bog standard Sata SSD and say, universe 7. AND here in universe 7
the PS5 has 5.5GBPS raw/10gb compressed or whatever,
in Universe 7 the PS5 games are going to be able to have :
Bigger worlds. a lot more variety in models, textures, higher quality textures per area, faster in-game transition speeds, Smaller install sizes without duplicate assets needed etc.
Last edited:
Ascend
Member
Regarding sampler feedback streaming... I'm not sure people get what it actually does... So I'm going to try and explain things step by step...
First, the transfer value given for the I/O slash SSD is basically a bandwidth value. The 2.4 GB/s raw value means that at most, 2.4 GB of data can be transferred per second.
The compressed value does not magically increase the 2.4 GB/s. What it does is, compress the files to make them smaller. The max amount transferred is still going to be 2.4GB in a second. But when you decompress it again on the 'other side', the equivalent size of the data would have been 4.8GB if you could have transferred it as raw data. So effectively, it's 4.8GB/s, but in practice, 2.4GB/s is being transferred.
Then we get to SFS. First, take a look at what MS themselves say on it;
Sampler Feedback Streaming (SFS) – A component of the Xbox Velocity Architecture, SFS is a feature of the Xbox Series X hardware that allows games to load into memory, with fine granularity, only the portions of textures that the GPU needs for a scene, as it needs it. This enables far better memory utilization for textures, which is important given that every 4K texture consumes 8MB of memory. Because it avoids the wastage of loading into memory the portions of textures that are never needed, it is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance.
That last sentence is important. It is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance. Now what does that mean? If you want to stream part of textures, you will inevitably need to have tiling. What is tiling? You basically divide the whole texture in equally sized tiles. Instead of having to load the entire texture, which is large, you load only the tiles that you need from that texture. You then don't have to spend time discarding so many parts of the texture that you don't need after you spent resources loading it. It basically increases transfer efficiency. Tiled resources is a hardware feature that is present since the first GCN, but there are different tiers to it, the latest one being Tier 4, which no current market GPU supports. It is possible that the XSX is the first one to have this, but don't quote me on that. It might simply be Tier 3 still.
In any case. When tiling, the size of the tiles will determine how efficient you can be. The smaller the tiles, the more accurate you can be for loading, and the less bandwidth you will need. Theoretically, you can be bit-precise so to speak, but that's unrealistic and requires an unrealistic amount of processing power. There is an optimum there, but we don't have enough information to determine where that point is in the XSX. Microsoft is claiming that with SFS the effective mulitplier can be more than 3x. This means that, after compression (everything on the SSD will inevitably be compressed), you can achieve a higher than 3x 4.8GB/s in effective streaming. To put it another way, effectively, the XSX is capable of transferring 14.4 GB/s of data from the SSD. This does not mean that 14.4GB/s is actually being transferred. Just like with compression, the amount of transferred data is still max 2.4GB/s. What it does mean is that if you compare the current data transfer with compressed tiles to loading the full raw uncompressed texture, you would need more than 14.4GB/s bandwidth to transfer the data to ultimately achieve the same result. This also helps RAM use obviously, because you're loading everything from the SSD into RAM, and you would be occupying RAM space that you wouldn't have. Basically, it decreases the load on everything, including the already mentioned RAM, and the I/O, CPU and GPU.
And here I'm going to speculate for a little bit, in comparison to the PS5. Tiled resources has been a feature in GPUs for a while. And the main part that allows this is sampler feedback. Now, you can have sampler feedback, but that does not mean that you necessarily have sampler feedback streaming. That would depend on the I/O. I recall Cerny mentioning that the GPU is custom built, and that they choose which features they wish to include and not include on the GPU. That implies they did not include everything that AMD has to offer in the GPUs. Most likely neither did MS. But if the PS5 still has this feature, then things regarding the SSDs remain proportionally the same between the compressed values in terms of performance difference, 9GB/s vs 4.8 GB/s. However, considering the beefy I/O of the PS5, it is actually quite possible that Sony ditched the tiled resources feature, and instead opted to beef up the I/O to allow the streaming of the full textures instead. If this is the case, then really, the difference in the SSD performance between the two consoles will be quite minimal. Why they would do that is beyond me though, so, most likely it's still in there. Whether they can stream it immediately is another story.
First, the transfer value given for the I/O slash SSD is basically a bandwidth value. The 2.4 GB/s raw value means that at most, 2.4 GB of data can be transferred per second.
The compressed value does not magically increase the 2.4 GB/s. What it does is, compress the files to make them smaller. The max amount transferred is still going to be 2.4GB in a second. But when you decompress it again on the 'other side', the equivalent size of the data would have been 4.8GB if you could have transferred it as raw data. So effectively, it's 4.8GB/s, but in practice, 2.4GB/s is being transferred.
Then we get to SFS. First, take a look at what MS themselves say on it;
Sampler Feedback Streaming (SFS) – A component of the Xbox Velocity Architecture, SFS is a feature of the Xbox Series X hardware that allows games to load into memory, with fine granularity, only the portions of textures that the GPU needs for a scene, as it needs it. This enables far better memory utilization for textures, which is important given that every 4K texture consumes 8MB of memory. Because it avoids the wastage of loading into memory the portions of textures that are never needed, it is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance.
That last sentence is important. It is an effective 2x or 3x (or higher) multiplier on both amount of physical memory and SSD performance. Now what does that mean? If you want to stream part of textures, you will inevitably need to have tiling. What is tiling? You basically divide the whole texture in equally sized tiles. Instead of having to load the entire texture, which is large, you load only the tiles that you need from that texture. You then don't have to spend time discarding so many parts of the texture that you don't need after you spent resources loading it. It basically increases transfer efficiency. Tiled resources is a hardware feature that is present since the first GCN, but there are different tiers to it, the latest one being Tier 4, which no current market GPU supports. It is possible that the XSX is the first one to have this, but don't quote me on that. It might simply be Tier 3 still.
In any case. When tiling, the size of the tiles will determine how efficient you can be. The smaller the tiles, the more accurate you can be for loading, and the less bandwidth you will need. Theoretically, you can be bit-precise so to speak, but that's unrealistic and requires an unrealistic amount of processing power. There is an optimum there, but we don't have enough information to determine where that point is in the XSX. Microsoft is claiming that with SFS the effective mulitplier can be more than 3x. This means that, after compression (everything on the SSD will inevitably be compressed), you can achieve a higher than 3x 4.8GB/s in effective streaming. To put it another way, effectively, the XSX is capable of transferring 14.4 GB/s of data from the SSD. This does not mean that 14.4GB/s is actually being transferred. Just like with compression, the amount of transferred data is still max 2.4GB/s. What it does mean is that if you compare the current data transfer with compressed tiles to loading the full raw uncompressed texture, you would need more than 14.4GB/s bandwidth to transfer the data to ultimately achieve the same result. This also helps RAM use obviously, because you're loading everything from the SSD into RAM, and you would be occupying RAM space that you wouldn't have. Basically, it decreases the load on everything, including the already mentioned RAM, and the I/O, CPU and GPU.
And here I'm going to speculate for a little bit, in comparison to the PS5. Tiled resources has been a feature in GPUs for a while. And the main part that allows this is sampler feedback. Now, you can have sampler feedback, but that does not mean that you necessarily have sampler feedback streaming. That would depend on the I/O. I recall Cerny mentioning that the GPU is custom built, and that they choose which features they wish to include and not include on the GPU. That implies they did not include everything that AMD has to offer in the GPUs. Most likely neither did MS. But if the PS5 still has this feature, then things regarding the SSDs remain proportionally the same between the compressed values in terms of performance difference, 9GB/s vs 4.8 GB/s. However, considering the beefy I/O of the PS5, it is actually quite possible that Sony ditched the tiled resources feature, and instead opted to beef up the I/O to allow the streaming of the full textures instead. If this is the case, then really, the difference in the SSD performance between the two consoles will be quite minimal. Why they would do that is beyond me though, so, most likely it's still in there. Whether they can stream it immediately is another story.
Last edited:
Exactly. With the information we know, the PS5 has a significant I/O advantage so its upto MSFT to prove the XSX can actually hit a 2-3x multiplier on I/O. It would be a coup for the XSX.
Great right up. I think the PS5 will have custom texture filters as well in the GPU but the additional custom hardware in the XSX for finer streaming of textures is what’s going to blow the competition out of the water. If they can pull it off, it would be incredible.
MasterCornholio
Member
Not if it leads to a degradation of image quality.
I don't think there's anything Microsoft can do to nullify the physical differences between the two I/O systems without making sacrifices.
Ascend
Member
Compression can be both lossy and losslessly. I guess it will be up to the developers to optimize for that. SFS is lossless, if it works as advertised.
It’s not just about picking only pieces of textures or lower quality mips. It’s also that with PRTs there were whole textures in RAM that were never used...at all!! MSFT did a multi year study with hardware they had put in the Xbox one X to monitor this. Richard at DF has a good explanation of what happened. This data they collected over 4 years is what’s going to make it hard to replicate what's in the XSX. Unless Sony did something similar or has a trick up its sleave as well.
Last edited:
MasterCornholio
Member
That could be a problem because it's promising alot and we haven't really seen proof that it can help match the PS5s I/O system yet.
Forgive if I seem brash but I don't have much faith in how software is going to make up for the physical differences between the two systems.
Last edited:
The SFS is hardware
MasterCornholio
Member
I must have been getting confused with BCPak.
But still after reading
Ascend
post a second time I don't see how the PS5 wouldn't have a feature like that. It definitely seems like they had something similar in the past and the I/O system in the PS5 would certainly benefit from it. Which is why I don't really believe the differences are minimal. If they were we would hear more about it from Microsoft because right now the official specifications makes it seem like there's a big difference between the two.
Last edited:
Yeah BCPack is one of the compression algorithms being used and it's also hardware accelerated. Devs usually use their own decompression algorithms at the expense of CPU overhead if they are not happy with Zlib or whatever is in the hardware decompression block.
Last edited:
Ascend
Member
Thanks... Just to confirm...;
So conclusion is, the PS5 can most likely do sampler feedback streaming, but would need CPU resources for it, while the XSX does it in hardware.
Last edited:
ToadMan
Member
I don’t know how you can get this - this is the opposite of what Sony said.
In fact all of the misunderstandings about variable clocks is precisely because Sony are designing the console around the thermal output of their hardware in order to provide a cooling solution that won’t intrude during gameplay.
captainraincoat
Banned
Most games are not made to cater for ssds or faster streaming of assets....you only have to look at PCs which have had these for years they target gpu/cpu specs
you will probably also find most third party games will also not cater to this which accounts 90% of a consoles library
you will probably also find most third party games will also not cater to this which accounts 90% of a consoles library
why would devs not utilize the SSDs since they're in both consoles and PCs can by definition have upgradable hardware?
oldergamer
Member
I still think ps5 will on rarely hit its performance maximum with the nvme drive. Sony tends to put out peak numbers and not sustained (example 10x teraflops). Heat will be a factor and that can affect performance.
Im really curious how that will affect the sustained performance.
Im really curious how that will affect the sustained performance.
You understand that in UE5 demo not only the textures are streamed but everything else too?
All LoDs are processed at pixel-level for textures, shadows, meshes, displacements, etc.
captainraincoat
Banned
Because they still target cpu/gpu.
they would have some elements of ssd utilisation but it’s gonna be pretty basic and only 1st party will make use of the customisation it offers
Ascend
Member
I used textures as an example. Everything that is streamed follows the same path. For the XSX textures have a different compression method to other stuff apparently, but I left that out.
But maybe I was unclear on what I meant. That everything is streamed in the UE5 demo does has no bearing on whether they are streaming the full resources from SSD, or only what is needed. I was referring to whether whole textures (as example) are streamed, or only the required portions. In other words, it is unclear if the PS5 is discarding unnecessary data before or after streaming it from the SSD. The XSX is capable of doing (most of) it before, in-hardware, saving SSD bandwidth. If the PS5 does this, most likely it needs to do it with software, costing CPU resources. If they are not doing it, you would use more SSD bandwidth than you would need to use on the XSX.
Last edited:
ToadMan
Member
Indeed because most games were designed around HDD based storage and the vast majority of PCs still use HDDs for bulk storage.
The reason the SSD tech in PS5 in particular is a big deal is because developers can design their games without the need to put in artificial limits to allow for the slow data transfer From HDDs. Every PS5 has the tech.
No more forced walking sections, no more out of place corridors, elevator rides, artificial movement speed limits, cavernous maze sections all designed to slow the player down while the console loads data in the background.
If the designer includes an element in their game it’s because they chose to do so for gameplay, not because of data streaming limitations.
Additionally, engine developers won’t need to spend time on things like LOD, replicating assets to circumvent seek speeds on HDD, patch file induced slowdown or filling up system RAM with data that isn’t needed. In fact the lack of data replication required could lead to smaller game install sizes but of course that depends on how the developers choose to package and distribute.
Whether designers will choose to design their games to this new paradigm depends on the team, but the capability is there if they choose to use it.
Sony 1st party will, that’s why I say PS5 storage access speed is a significant technical bump for this coming gen.
Since MS is following a policy of Xsex and PC development, they’ll have to design to slower PC speeds not allowing them to take full advantage of the concept and thats presumably why they’ve specced a slower storage solution.
Obviously portions. Otherwise it would not fit in RAM.
Both PS5 and XBSX will use 64K blocks for data streaming.
Direct Storage on PC will also use 64K blocks.
You cannot discard stuff if you have no idea whether it will be rendered. Before actually rendering it.
No CPU is used. Only GPU. GPU can calculate things, it's called "async compute".
SFS just simplifies one path: traditional texture MIP filtering.
You load ALL textures into VRAM, render them, produce "mip level map" from the render target, then discard the parts that don't fit the map.
You can obviously render the map using lower resolution mips and then fetch higher resolution ones.
All of it is desctibed in MSFT whitepaper on SFS.
MasterCornholio
Member
Kazekage1981
Member
I think the main barrier to break here is the PERCEPTION of the SSD being more than a storage device, and actually contributing the graphics fidelity. It is not that PS4 and Xbone and PC did not use hard drives for streaming of graphical assets before, they do to some degree but negligible amount or no one really noticed because its not the main thing developers are targeting when making a game. No one designs games around SSDs, they target the: CPU, GPU, RAM. Its only now that the timing is right for not only just to put a SSD drive to eliminate loading times, but to use the SSD that contributes to graphics and works with the GPU.
Interesting to say the least.
Interesting to say the least.
Last edited:
I think MSFT made a good hardware for current gen games.
It can do all the traditional render paths faster.
Sony is building on the new approaches.
It kinda reminds of PS3, but only slightly because it's not hard to program for.
On the other hand using novel tech is a risky path anyway.
Ascend
Member
The 64k blocks is news to me.
As for it being obviously portions, I'm not so sure. You can still stream the whole thing and do culling, then stream the next one etc. It would be a lot more straining on the SSD, but it seems that's exactly what the PS5 was built for.
That's the backwards way. Maybe I didn't explain it like I should. By discarding it before loading from the SSD, I basically meant that only what is required will be loaded. You will not load what will not be used.
Hm... That's how the PS5 likely works, but, the XSX seems different. First line, but I'll put more info here too;
Note that I just googled that. What I wrote before was simply what I logically deducted.
Last edited:
itsjustJEFF
Member
Amazing how many hardware engineers there are all of a sudden. MS has released no details about this specific tech, yet it's being quoted and misused. The guy literally says "We don't have any real details yet, but its possible..." Stop and just wait till MS drops the info. Breathe cause it's going to be okay. At the end of the day, no matter how you slice it. It's doesn't matter if one console has instant access to "100GB" or has "Less bottlenecks" or "an SSD that is far and away... blah blah" there will be minimal differences between these two consoles. They went in two different directions, but they'll end up in the same place.
Last edited:
Self_Destructive
Member
Where are you getting this from? I want the sources.
MasterCornholio
Member
Ok I understand where your going.
It's true that Sony is doing alot of things differently with their I/O solution. And I agree that it is risky since some of it really hasn't been done before. On paper the specifications for the I/O system look really impressive and appear to blow the completion out of the water. But if there's one thing that I've learned is that all that matters is the final result which we will see when these systems release.
It could be possible that Microsoft has a much better I/O system but it hasn't been demonstrated yet. It could also be possible to that Sonys I/O is leagues ahead of the Xboxs. I guess we have to wait and see what happens.
Sampler feedback whitepaper and Sony's SSD patent. Direct Storage will use the same as it should work with SFS too.
Self_Destructive
Member
I'm referring specifically to your claim that DirectSorage on PC will use 64K blocks.
Link me to that please if you can.
Last edited: