-
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.
- Status
Bo_Hazem
Banned
Well, FF15 was one of the shittiest looking games this gen with the worst draw distance ever, and played it on performance instead of resolution on PS4 Pro. Still, I give it 9/10 and squeezed every bit of it!
Imtjnotu
Member
the demo of it looked so damn good too. i see why they had to downgrade it
GetSchwifty
Banned
I haven't played it yet, but I saw the texture pop-in in the DF video two days ago and It did look pretty bad... You could almost feel the slow HDD seeking. Next-gen can't come soon enough.
Bo_Hazem
Banned
BGs
Industry Professional
They sleep. I can't.
David Ricardo
Member
EventHorizon
Member
It's not just the speed that's important. It's the latency. It doesn't matter if your bandwidth is enough to stream an asset off the SSD if it can't do it when you need it. That's why the PS5's SSD controller has 6 priority channels while the XSX likely only has 2. The XSX simply wasn't designed to be used as a substitute for RAM. It was designed to load ram fast, not replace it.
If you can't get the data when you need it from the SSD, you have to cache it ahead of time in RAM so you can retrieve it from there instead. In fact that is exactly what Microsoft says is a good use for its 3.5 GBs of slower memory used by games. Meanwhile Cerny made a specific use case of its fast SSD being the timely streaming of assets like audio directly from it when the game requires it. Because the PS5 can get SSD data much quicker, it doesn't have to waste as much RAM to preload assets that it might need.
Finally, the overall SSD speed does matter. Data requirements aren't uniform. A game might require new audio while at the same time a player goes through a door requiring the game to load all the assets for a new part of the level. That's going to need far more bandwidth than the loading of any single asset. To compensate for a slower I/O, a game will have to do some combination of the following...
- Cache more data in memory, divide the level up, and force a slow transition between each part
- Reduce the data density in the level to the point where the I/O can handle the worst case scenario
SonGoku
Member
SSD didn't fix it, seems to be an issue with UE4 borked LOD transitions
Imtjnotu
Member
yes it does matter. check the road to ps5 12:08 time of road to ps5
youre able to load textures in an out in a half second. no need to have all the textures loaded for a big area if they dont need to be
SonGoku
Member
Guaranteed throughput is 5.5GB/s for PS5 and 2.4GB/s XSX
The rest depends of how well assets compress on the disk/install, devs don't have to worry about that since they can stream at 5.5GB/s on demand.
David Ricardo
Member
What I'm saying is that the theoretical max speed or even the most common speed of the SSD doesn't matter, because those are not sustained numbers you can rely on. What matters is the minimum speed guaranteed, because that's the speed at which you can actually be sure the data is going to be available.
If you design your game around having 8 GB of assets streamed every second and sometimes the speed is 7.something, there are assets that are not going to be there on time. You would be seeing textures appearing in front of you.
Last edited:
GetSchwifty
Banned
I thought SSD on PS4 doesn't improve loading speeds because the SATA and USB share a controller.
Anyway, it does look like something that can be fixed through a patch... I'll hold off getting it until they do. Thanks for the headsup.
Last edited:
SonGoku
Member
That too but this seems like a engine issue, i agree it should be fixed
EventHorizon
Member
A developer isn't going to leave performance on the table by coding to minimum I/O bandwidth specs. What they'll do is use memory as an SSD cache to average out I/O bandwidth requests. That way the game can be designed around the average I/O bandwidth. That's why bandwidth is so important. The more of it you have, the less RAM you need to waste being used as a cache.
MasterCornholio
Member
BGs
Industry Professional
It is always important to have a good margin of maneuver. Especially in a system like PS5 designed to be as efficient as possible from the beginning. Variable speeds, SmartShift, ...
Edit.- Tomorrow from the computer I can make a better explanation. Morpheo owns me right now.
Last edited:
Gamerguy84
Member
Excellent read, for me anyways because Im trying to learn about this, from Toms Hardware about the difference between GCN, RDNA, RDNA2.
It also made me realize just how powerful these consoles are going to be. We are going to see amazing games that will blow our minds.
The part that really blew my mind was the comparison between Radeon VII (GCN 13..8 TF) vs 5700XT(RDNA1 9.75TF). The 5700 was very comparable to the more powerful Radeon.
Last edited:
This patent describes how priorities work. Are you fluent in Japanese?
EventHorizon
Member
Data priorities allow urgent data requests to jump to the head of the line. It allows the SSD to be more fully utilized. As a simple example, lets compare two phone systems. One with two priority levels and the other with only one. Now imagine you are expecting an important call. The caller knows its important so if they call and the line is busy, they'll call again.
Now if your phone only has one priority level, you'd not want to use it much so that it'd be free when that important call comes in. When you do use the phone, you'd want to keep the call short so that if you're unlucky and the important call comes during that time, there won't be too much of a delay. Even though you didn't know the call came in, shortly after you hung up, you'd get the important call.
On the other hand, if you phone had two priority levels, you could make all the calls you wanted on the low priority line and talk for as long as you liked. The important call would come over the high priority line and you'd get it immediately no matter what. The extra priority level means there is less delay between when something important happens and when it is dealt with. It also means you can more easily make full use of a resource.
Sony gave an example of streaming in level data at a low priority level when the player's actions require an audio clip to be played immediately. That would go on the high priority level that jumps it to the head of the line. Without that ability, either the level data couldn't be streamed to keep the SSD ready for any audio requests, or the audio file would have to have been loaded ahead of time and sitting in memory just in case it was needed.
Edit:
Another way to look at this is that the PS5 will allow normal 2 priority level SSDs to be hooked up to it. Its software can simulate the 6 priority levels it needs on top of the two real ones...except that in order to do that, the 2 priority level SSD has to be faster than the 6 priority level one in order to have the same performance. In other words, the 6 priority levels of the PS5's SSD actually make it perform like an even faster drive.
Last edited:
GermanZepp
Member
Guys, with the power of both nextgen consoles we get 4k assured, probably 60 fps too *if the devs want it. The CPU jump is massive.
So with both consoles hitting 4k60, 4k30, 3800p/60 and with improved checkerboard solutions resolution is not going to be a problem at all. The 60 fps games are going to be a lot more common too.
My point is, with both consoles running the games almost the same way, the power advantages of the xbox aren't gonna make a big difference. Perhaps the x is a more stable console with better raytracing.
That Sony SSD though, I mean, damn. If it is like the rumors suggested, like, no loading, zip, zero, nada and games can be designed around that. We need to see proof first but I'm starting to think Sony make the right calls with the ps5. I mean, we can expect:
4k60
Faster SSD
Dualsense full of *newtech stuff to experience
PSVR2
If Playstation is going to push VR to the next level the ps5 then was probably designed to an extent to high performance, high locked framerates. How the fastest SSD can impact on VR games. I'm excited.
So with both consoles hitting 4k60, 4k30, 3800p/60 and with improved checkerboard solutions resolution is not going to be a problem at all. The 60 fps games are going to be a lot more common too.
My point is, with both consoles running the games almost the same way, the power advantages of the xbox aren't gonna make a big difference. Perhaps the x is a more stable console with better raytracing.
That Sony SSD though, I mean, damn. If it is like the rumors suggested, like, no loading, zip, zero, nada and games can be designed around that. We need to see proof first but I'm starting to think Sony make the right calls with the ps5. I mean, we can expect:
4k60
Faster SSD
Dualsense full of *newtech stuff to experience
PSVR2
If Playstation is going to push VR to the next level the ps5 then was probably designed to an extent to high performance, high locked framerates. How the fastest SSD can impact on VR games. I'm excited.
Last edited:
Trainsmove
Banned
Where has anyone said that? Please post a link?
Trainsmove
Banned
I think you missed this part.
"We call this standard memory. GPU optimal and standard offer identical performance for CPU audio and file IO. The only hardware component that sees a difference in the GPU."
Anything working with the GPU is at max speed, and only the CPU, io, and audio is using the lower speed.
GPU to cpu 560GB/s
GPU to io 560GB/s
GPU to audio 560GB/s
CPU to io 336GB/s
560 is @320bits
336 is @ 320bits
446 IS @ 256bits PS5 max
So any way you look at it it's clear that the Xsx is on another level.
Last edited:
Tripolygon
Banned
When you have near just in time access to any data in the 825GB SSD, you don't need a 100GB partition of virtual memory.
This combination of hardware and software to enable near instant access of data is what Microsoft termed "Velocity Architecture" only Sony does not give it a fancy name and PS5 is faster at it based on specs of the storage and IO.
lepermessiah
Member
, hard to glean context in this place sometimes lol
MiyazakiHatesKojima
Banned
Last edited:
Bo_Hazem
Banned
I just played 2 hours of FF7, it's surprisingly amazing in terms of graphics and had zero issues on PS4 Pro. Which one are you using (OG, Slim, Pro)? I think it's running at around 1800p on the pro.
lepermessiah
Member
I think it's clear the I/O and SSD set up in PS5 is manifestly superior.
megreotsugua
Banned
XSeX SSD and I/O is manifestly a lot slower than the one in the PS5. The Crytek developer affirmed this when he confirmed that XSeX loading is not instant and still at the realm of about 5-10 seconds (he used the xbox official video as an example to avoid NDA).
The same bottleneck in the PC is present in XSeX (probably sped it up a little).
PS5 has 6 custom chips to ensure loading is INSTANT. XSeX only has one custom chip, the decompressor, which is not even as fast as the one in the PS5.
Xbox Velocity Architecture is a marketing talk. No wonder developers are not talking about it and not excited about it. It's the same slow ass architecture in PC sped up a little.
rnlval
Member
AMD GPUs use the "combined scatter" and "combined gather" methods.
Last edited:
rnlval
Member
This is XSX's CG render die shot and PCB. This is not the real world PCB and die shot.
My guess is eight normal 32 bit physical PHY GDDR6 controllers (like NAVI 10) with two extra 32 PHY GDDR6 controllers above the two CCX CPU modules.
I placed 2GB chips closer to the CCX CPU modules i.e. locality design rules?
XSX's CG render die shot reminds me of X1X's die shot layout.
NAVI 10 PCB example
Last edited:
pasterpl
Member
Thanks for your correction.
Seriously? He responded to you yesterday. He said he have not seen it. Stop spreading lies and making up sh*t.
MiyazakiHatesKojima
Banned
Damn, right on the money!
rnlval
Member
XSX's SSD area has two hardware decompressors which are
ZLIB path for general data
BCPack path for texture data
Last edited:
pasterpl
Member
Xbox Series X 'Texture Compression BCPack' Reportedly Better Than The PS5's Kraken
Microsoft revealed the complete specs for Xbox Series X earlier this month. Today, we might have something really interesting - Xbox Series X BCPack.
www.thegamepost.com
rnlval
Member
From DF, PS4 Pro's 1620p is checkerboard.
FeiRR
Banned
I think the way it's going to work will be: your friend is playing a game, you go into their profile. On PS4 you can see what game they're playing and in some games what mode/level (doesn't work that well too, sometimes it's outdated info). To see the game itself, you have to ask for shareplay. On PS5, the system will quickly load the game your friend is playing (provided you have it too, of course) and show you exactly what they're seeing. Then you can join or shareplay. The difference is, this won't be streaming video you see but in-engine graphics. This would mean, with good internet connection the lag would be exactly the same you have in multiplayer games, so not noticeable. And for people who have datacaps - huge savings in transfer.
We can go even further (if the game editing rumour is true): you can also see a game you don't have. Which will, of course, require a certain package of data (say, engine and one level worth of assets) to be send over Internet. In this case, your friend's console is sharing this portion of the game with you on p2p basis or it comes from the Sony storage server (I think the latter is a more probable solution). Technically speaking it's doable but would require some system-level management of install packages, like: package1 (engine), package2 (level1), package3 (level2) and so on. This would also allow you to use that 'create' button to make a demo for friends. Let's say you'll be allowed to share one level of the game or 30 minutes of gameplay. You decide which level (map) you like the best, you put it in a 'demo' package and send to your friend. After playing the level for 30 minutes/once, etc. (publishers decide, of course), your friend will be prompted with PS Store question if he wants to buy the game.
HOWEVER, if SSD is so ground-breaking as devs say, we might see the departure from the traditional level/map system in games. Imagine a military FPS MP where you don't play on maps but certain areas of, say, Europe at war. When one side (this could be thousands of players now thanks to upped CPUs) has an advantage and pushes past the frontline, you don't load another map like, for example, in Battlefield V. You can run/drive a tank/fly a plane from one end of the whole virtual battlefield to the other because now the technology allows it.
I think the implications of fast data access are much more exciting than upping the graphics to 4k in all games or even locking FPS at stable 60. We're going to like it a lot.
Last edited:
Xbox Series X 'Texture Compression BCPack' Reportedly Better Than The PS5's Kraken
Microsoft revealed the complete specs for Xbox Series X earlier this month. Today, we might have something really interesting - Xbox Series X BCPack.
What’s MS holding back to also use Kraken and switch now to it?
D
Deleted member 775630
Unconfirmed Member
Why would they if their own BCPack is better?
For Zlib, as it seems kraken is better suited for general purpose.
draliko
Member
kraken is proprietary from RAD Game Tools, Sony bought a licence to pack it inside the ps5 (a price you as the user will pay). And we don't really know in reality how much better it will be, as always there is a diminishing return after a treshold, so maybe the real world difference (not some synthetic bench) will be irrelevant. Until we see some games, this is all speculation.
FranXico
Member
In hardware?
farmerboy
Member
Which game do you think will show it off best?
pasterpl
Member
why would they? Bcpack is much better for textures, it was designed to handle these specifically
bhunachicken
Member
I'm still mightily confused as to how a GCN @ 13.8TF can be worse than an RDNA1 @ 9.75TF.
To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.
Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.
Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.
What else is going on to affect the performance?
To me, it's like someone telling me that a ton of bricks is heavier than a ton of feathers.
Is a teraflop really such a bad measurement of things? After all, a flop is the measurement of floating point operations per second. A terraflop is a billion floating point operations per second.
Therefore, is one piece out hardware is outputing 13.8TF per second, while the other is outputting 9.75T per second, it should follow that the one with the higher number is handling more of these calculations per second and therefore is the better device.
What else is going on to affect the performance?
lepermessiah
Member
First-party titles will flaunt it the most I'm sure, just thinking what Naughty Dog will do has me salivating already.
- Status