General Next Generation Rumors and Discussions [Post GDC 2020]

If I want to design a game where I can move around a massive, but diverse and dense environment at high-speed, I can't do it on current gen consoles because you can't pull all the assets from storage for these diverse dense environments. You could not, for example, take Spider-Man and make into a Superman game.

If you were flying around at Superman speeds, wouldn't you have to kill all the audio effects until you slowed down or stopped moving?
 
On xbsx the lower speed partition have the same "cost" as the high speed partition on the bus. So to get the close to the maximum b/w they have to use the 10TB partition the vast majority of the time. Which shouldn't be difficult, just some more thinking about how to allocate buffers. But it won't be 560 unless they don't use any memory above 10GB.

Example:
90% requests at 560
10% requests at 336
Effective usable b/w here would be 537

Obviously if it's 99% to 1% the impact is pretty much nothing (557). And if a lazy dev does it randomly it will drop lower (500).

Hmm. Isn't the cost from using the slow speed pool only that it slows down the fast pool in 1 to 1 ratio, but imo doesn't reduce the overall/total bandwidth of the memory system?

I think you can fully saturate the slow pool and still have a total bandwidth of 560GBs. If you fully saturate the 6 2GB chips using their 192bit connection of the total bus running at 336GBs, you can still fully saturate the remaining 4 1GB chips with different data using their 128 connection running at 224GBs.

I do understand that you can't be using the 560GBs of bandwidth on the first 10GB and then add something to the last 6GB without dropping in speed on the first 10GB, but the total speed from 10+6GB should't be lower than on 10GB alone and should still be able to reach 560GBs?
 
If you were flying around at Superman speeds, wouldn't you have to kill all the audio effects until you slowed down or stopped moving?

No, you're superman, he can hear everything. You just play all audio effects at the same time. Hang on, the PS5 can do that. Did Sony build a console just to make the ultimate Superman game!?!? :runaway:
 
No, you're superman, he can hear everything. You just play all audio effects at the same time. Hang on, the PS5 can do that. Did Sony build a console just to make the ultimate Superman game!?!? :runaway:
didn't superman and zod undergo painfulness when they heard the world at once? They needed to reduce their powers to hear just 1 thing or go mentally crazy.
 
An interesting quote from resetera:

————

I'll not share any names as I have some friends but here's a small quote regarding PS5. This is from a major 3rd party dev......

"The data transfer is so damn fast that you don't need to keep shit rendered behind the player. That's just one part. You don't need to have a loud ass console rendering all that. Just in front." **

The rest is personal stuff and just so some realize, it's the same that I've been personally hearing for a long while.

So again, happy times ahead for all.

——————

Although this seems nice, I have a doubt: does rendering behind the player still happen? Or does he mean a safety margin to to the sides? Or is this just croc?
 
Last edited:
On xbsx the lower speed partition have the same "cost" as the high speed partition on the bus. So to get the close to the maximum b/w they have to use the 10TB partition the vast majority of the time. Which shouldn't be difficult, just some more thinking about how to allocate buffers. But it won't be 560 unless they don't use any memory above 10GB.

Example:
90% requests at 560
10% requests at 336
Effective usable b/w here would be 537

Obviously if it's 99% to 1% the impact is pretty much nothing (557). And if a lazy dev does it randomly it will drop lower (500).
I thought only graphics could use the faster RAM or is it just that’s the desirable use?
 
I thought only graphics could use the faster RAM or is it just that’s the desirable use?
No idea how they restricted it, the math are the same regardless. You get maximum bandwidth if the data requests are spread evenly, which statistically requires the data to be spread evenly.
 
Hmm. Isn't the cost from using the slow speed pool only that it slows down the fast pool in 1 to 1 ratio, but imo doesn't reduce the overall/total bandwidth of the memory system?

I think you can fully saturate the slow pool and still have a total bandwidth of 560GBs. If you fully saturate the 6 2GB chips using their 192bit connection of the total bus running at 336GBs, you can still fully saturate the remaining 4 1GB chips with different data using their 128 connection running at 224GBs.

I do understand that you can't be using the 560GBs of bandwidth on the first 10GB and then add something to the last 6GB without dropping in speed on the first 10GB, but the total speed from 10+6GB should't be lower than on 10GB alone and should still be able to reach 560GBs?
That's not possible since each channel is connected to separate part of the data available. They need to have the max bandwidth to each chip to reach the maximum. It means the gpu must have a NEED to fetch equally on each chip. This is done by splitting the space evenly across the channels, which statistically keeps them full (request being random). If you fully saturate half the controllers the other half will deplete the requests targetting their respective region until it slows down to the same rate as the others.

There are tricks to mitigate this, which is what MS did by partitioning the high range as a low demand pool, and keep the intense gpu data equal on all channels. It's a very good solution for an age old problem with data parallelism. But it has the caveats mentionned.
 
Last edited:
It's not about graphics. If I want to design a game where I can move around a massive, but diverse and dense environment at high-speed, I can't do it on current gen consoles because you can't pull all the assets from storage for these diverse dense environments. You could not, for example, take Spider-Man and make into a Superman game. Not unless you slow Superman way down, or you make so many compromises to the environment that it just doesn't feel like a city anymore - just Superman just flying around a bunch of tall bland geometry.
'The Flash' is going to be PS5 exclusive, and f@#%ing impossible to play. :p

Current location - Fort McPherson, Alaska.
Next mission - Montevideo, Uruguay.

You have 48 seconds to run there. Don't crash...
 
It's not about graphics. If I want to design a game where I can move around a massive, but diverse and dense environment at high-speed, I can't do it on current gen consoles because you can't pull all the assets from storage for these diverse dense environments. You could not, for example, take Spider-Man and make into a Superman game. Not unless you slow Superman way down, or you make so many compromises to the environment that it just doesn't feel like a city anymore - just Superman just flying around a bunch of tall bland geometry.

Graphics is the least imaginative thing devs will be leveraging SSDs for in new machines for. Forget graphics. It's a complete distraction.

Think of Cyberpunk 2077 where you can fly around the city at high speed and zoom down to ground near instantly in any place and have all that detail immediately available.

I'm really surprised that people still say things like that and believe it. Make Spider Man with gourard shaded low poly count models and you could do it on the NSW with everything you think the next gen will bring to games.

Gameplay will be exactly the same, the only difference will be the level of graphics. If you wanted to go barebones graphics you could just have low poly wireframe models ala. 80's era Battlezone arcade game and have a massive Spider Man game with warp speed movement in cities or even the entire world of this solar system. Ugly, sure. Gameplay that wouldn't exist even on next gen consoles? Definitely.

No AAA developer is likely to do that because graphics sell, gameplay is secondary to that for AAA developers. If Gameplay was the first priority than HZD would have easily been able to incorporate the flying that they wish they could have put in. But it isn't, they wanted X level of graphics which then determined what level of gameplay they could have on the hardware.

TL: DR - the level of desired graphics is a significantly larger limiter on game design than hardware.

Regards,
SB
 
Last edited:
I'm really surprised that people still say things like that and believe it. Make Spider Man with gourard shaded low poly count models and you could do it on the NSW with everything you think the next gen will bring to games.

Gameplay will be exactly the same, the only difference will be the level of graphics. If you wanted to go barebones graphics you could just have low poly wireframe models ala. 80's era Battlezone arcade game and have a massive Spider Man game with warp speed movement in cities or even the entire world of this solar system. Ugly, sure. Gameplay that wouldn't exist even on next gen consoles? Definitely.

No AAA developer is likely to do that because graphics sell, gameplay is secondary to that for AAA developers. If Gameplay was the first priority than HZD would have easily been able to incorporate the flying that they wish they could have put in. But it isn't, they wanted X level of graphics which then determined what level of gameplay they could have on the hardware.

TL: DR - the level of desired graphics is a significantly larger limiter on game design than hardware.

Regards,
SB
you shouldnt be talking about graphics and gameplay as separate entities in such a way. Gameplay is elevated by graphics and other elements. If you want to play Spiderman in wireframe, so be it. It wont be the same experience. Visuals provide cues and immersions that make you feel and understand the gameplay better.
It is not just because some company wants pretty graphics. We want to experience for example how it is like to be and play as Spiderman in Manhattan. With no graphics there is no Spiderman and no Mangattan. We want to experience an immersive gameplay of riding a horse in RDR2 in large lands or flying through Skyscrapes as Spider Man, or fighting a giant mechanical T-Rex in HZD. We want to be Solid Snake in harsh rainy conditions on a Tanker. We want to use shadows and light in our stealth gameplay. If all these games were Wireframe they wouldnt be the same. The gameplay experience would have sucked and in many cases limited because the developer doesnt cant provide the proper visual cues to support the gameplay elements properly.
Graphics are a big part of the experience. It is not a limiter. It goes hand in hand with the gameplay and game design.
 
you shouldnt be talking about graphics and gameplay as separate entities in such a way. Gameplay is elevated by graphics and other elements. If you want to play Spiderman in wireframe, so be it. It wont be the same experience. Visuals provide cues and immersions that make you feel and understand the gameplay better.
It is not just because some company wants pretty graphics. We want to experience for example how it is like to be and play as Spiderman in Manhattan. With no graphics there is no Spiderman and no Mangattan. We want to experience an immersive gameplay of riding a horse in RDR2 in large lands or flying through Skyscrapes as Spider Man, or fighting a giant mechanical T-Rex in HZD. We want to be Solid Snake in harsh rainy conditions on a Tanker. We want to use shadows and light in our stealth gameplay. If all these games were Wireframe they wouldnt be the same. The gameplay experience would have sucked and in many cases limited because the developer doesnt cant provide the proper visual cues to support the gameplay elements properly.
Graphics are a big part of the experience. It is not a limiter. It goes hand in hand with the gameplay and game design.

You, yourself, just separated them. Gameplay and experience. Yes graphics enhances the experience. The gameplay is the same regardless. Same thing with faster loading leading to a better experience. The gameplay remains the same whether it takes 10 minutes to load a level or 1 second to load a level. The experience, however, is far better if you don't have to wait 10 minutes. The gameplay itself is still the same.

Just like Deus Ex PC has the same gameplay as Deus Ex console. Original Deus Ex, not the new ones. However, the experience on PC was far better because the levels were larger leading to a more cohesive "experience."

Regards,
SB
 
You, yourself, just separated them. Gameplay and experience. Yes graphics enhances the experience. The gameplay is the same regardless. Same thing with faster loading leading to a better experience. The gameplay remains the same whether it takes 10 minutes to load a level or 1 second to load a level. The experience, however, is far better if you don't have to wait 10 minutes. The gameplay itself is still the same.

Just like Deus Ex PC has the same gameplay as Deus Ex console. Original Deus Ex, not the new ones. However, the experience on PC was far better because the levels were larger leading to a more cohesive "experience."

Regards,
SB
In Resident Evil 2 on PS1, all areas were separated to loading sections. Enemies could not follow you from one room to the other. In Resident Evil 2 Remake enemies can follow you through rooms. All areas are connected as one.
As for Deux Ex the level design changed on the PS2. Seamless gameplay and therefore the experience were compromised.
 
Back
Top