Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

[goonergaz]

Sony never said 10 times faster of loading times. They say PS5 SSD is faster than all current SSD.

https://www.thesun.co.uk/tech/9119249/ps5-demo-loading-times-under-second/

Sony gave is a video shown to inversters which PS5 loadeded 0.83s while PS4 Pro loaded 8.1s.
Some media speculate that PS5 has 10x faster loading.
.

They gave another demo where is was 15 seconds down to 0.8 (19 times faster);
"He didn’t want to just improve loading times, he wanted to eliminate them. He proved this by showcasing fast travel in Marvel’s Spider-Man on the PS4 Pro, which took 15 seconds. The same fast travel took only 0.8 seconds on a “early low-speed” PS5 devkit."

https://www.playstationlifestyle.ne...w-about-ps5-after-wired-mark-cerny-interview/​

 

[Deleted member 11852]

It's also worth remembering that Sony/Cerny didn't promise to eliminate load times. From the October Wired article:

The PS5 will change that. "Even though it will be fairly fast to boot games, we don't want the player to have to boot the game, see what's up, boot the game, see what's up,"

The goal seems to be get all loading to an absolute minimum allowed by modern tech/architecture.

 

[AlNom]

88 rops sounds fishy.

You're not thinking 4th dimensionally!

This is heavy.

 

[turkey]

It's also worth remembering that Sony/Cerny didn't promise to eliminate load times. From the October Wired article:



The goal seems to be get all loading to an absolute minimum allowed by modern tech/architecture.

This seems like the same thing Microsoft were talking about, multiple games left in a hybernated state. You won't "boot" them but resume playing them.

Not a specific performance goal but and user experience / QOL aim. One that super fast IO can make viable.

 

[MrFox]

They gave another demo where is was 15 seconds down to 0.8 (19 times faster);
"He didn’t want to just improve loading times, he wanted to eliminate them. He proved this by showcasing fast travel in Marvel’s Spider-Man on the PS4 Pro, which took 15 seconds. The same fast travel took only 0.8 seconds on a “early low-speed” PS5 devkit."

https://www.playstationlifestyle.ne...w-about-ps5-after-wired-mark-cerny-interview/​

I wonder how much of that 0.8s was reading the data and how much was setting up the scene and the game engine once it has everything it needs? We cannot know if an infinitely fast storage might level down at 0.1s or 0.6s or something because there's more to do than just reading data. Imagine if global lighting and other slow changing buffers are generated every 10 frames, you get a minimum of 0.3s additional delay for any far jump in the world.

 

[Deleted member 11852]

This seems like the same thing Microsoft were talking about, multiple games left in a hybernated state. You won't "boot" them but resume playing them.

You must boot them initially, and presumably after game updates. Smart speculation, because this is a long-solved problem, is game states will be dumped to disk so gamestates only work as long as the state the game was saved in, remains the same. They may also break after system updates so on PS5 every three weeks after endless "stability and performance improvements" formware updates.

Not a specific performance goal but and user experience / QOL aim. One that super fast IO can make viable.

Yup, it can be a lot better than it is now. It can be a lot better than it is on PC. Suspend/resume on modern consoles is the best for those with busy/unpredictable lives.

 

[BRiT]

I'm assuming the MS approach will be VM Snapshots so there doesn't need to be any support by the game or even the Guest OS, only needs support from the Hyperviser level. They could do it with Hibernation files, but then that requires support on the Guest OS that's already running as a VM, so I'd lean against that in order to make the Guest OS more streamlined.

Not sure how Sony would approach it from a PS4 perspective and I doubt they added any sort of VM / Hyperviser layer for PS5.

 

[bgroovy]

I'm assuming the MS approach will be VM Snapshots so there doesn't need to be any support by the game or even the Guest OS, only needs support from the Hyperviser level. They could do it with Hibernation files, but then that requires support on the Guest OS that's already running as a VM, so I'd lean against that in order to make the Guest OS more streamlined.

Not sure how Sony would approach it from a PS4 perspective and I doubt they added any sort of VM / Hyperviser layer for PS5.

VM snapshots aren't required to suspend apps on Android/iOS. Presumably they will be able to tombstone the gamestate in the same manner on PS5.

 

[PSman1700]

I'm assuming the MS approach will be VM Snapshots so there doesn't need to be any support by the game or even the Guest OS, only needs support from the Hyperviser level. They could do it with Hibernation files, but then that requires support on the Guest OS that's already running as a VM, so I'd lean against that in order to make the Guest OS more streamlined.

Not sure how Sony would approach it from a PS4 perspective and I doubt they added any sort of VM / Hyperviser layer for PS5.

Is that how MS would do it for cross platform games to windows?

 

[Nisaaru]

As nice as these VM snapshots could be for a few games I really wanna see how they perform for persistent MP games. A lot times there you just spend waiting for network states to synchronise.

 

[BRiT]

Is that how MS would do it for cross platform games to windows?

I doubt that benefit will exist for PC Gamers, so no need to be concerned with that aspect of it from MS's standpoint.

Though as BG said, they could take the tombstone approach, but I think that requires explicit support by each game/app since it's part of an API.

 

[PSman1700]

PC gamers doomed, for once console has the upper leg .)

 

[bgroovy]

but I think that requires explicit support by each game/app since it's part of an API.

Even if that's true it could be part of the TCRs for the platform. I wouldn't expect it to be much of a burden.

 

[Silent_Buddha]

I wouldn't expect it to be much of a burden.

It would depend on the game. Game states for console/PC games are far more complex than those for mobile games. I'm not sure it'd be as trivial as you think it would be.

At the extreme's you have something like Skyrim or Fallout 3/4 where you need to keep track of changes to every single item the player has touched in addition to all the game related changes (quest states, world states, etc.).

Most mobile game states likely have to keep track of less things that Skyrim has to track in a single room.

In such a system, a VM snapshot would be easier.

Regards,
SB

 

[BRiT]

Even if that's true it could be part of the TCRs for the platform. I wouldn't expect it to be much of a burden.

Agreed.

I was just thinking why wouldn't current-gen consoles support multiple instant-resumes already. That sort of implies they're not using Tombstoning for that purpose, no?

Tombstone API with forced TCRs doesn't seem like it would put any undue pressure on the developers. Unless it was an issue with getting it into the popular engines in time for 2013 launches?

 

[bgroovy]

I was just thinking why wouldn't current-gen consoles support multiple instant-resumes already. That sort of implies they're not using Tombstoning for that purpose, no?

It probably all down to drive speed. It would take a while to read or write a 5-6GB game savestate to a mechanical drive. PS4 just keeps suspended games in memory and Xbox One has flash provisioned for a single game IIRC. With the SSds it's a lot easier to just set aside a space for like your last 4 games, or whatever.

 

[Metal_Spirit]

That quote would logically be about space a game required and has nothing to do with performance or bandwidth.

I.E data duplication to create chunks of contiguous data to stream to work about physical HDD performance constraints.

Not really. Data duplication exist as you say to create contiguous data to stream, reducing data search times and allowing for sequencial reads. If you eliminate the duplicated data you will need to seek for the data you want, loosing time. But if even so you can be 10x faster, then your read speed must be above 10x more.
Other demos proved that, showing up to 19 times more, on a “low speed” version of the ssd.

 

[bbot]

This person is claiming that 5 times rtx2080ti, per clock, per CU. This is a rumor.

https://twitter.com/i/web/status/1205646265482170368

 

[Rootax]

Guessing performances based on patent, yeah, no.

 

[bbot]

Guessing performances based on patent, yeah, no.

That's why I wish the site owners would ask their sources. Obviously what they find will be rumors. We won't know for sure, until Microsoft discloses the information.