Unreal Engine 5, [UE5 Developer Availability 2022-04-05]

[iroboto]

All I’m trying to say is that the problem with your editor not being as competent as the hardware your designing for is a problem that takes time to solve and an endeavour that each studio will need to tackle on their own. And the closer you get to that I/O barrier the more taxing it will be on PC hardware to emulate.

i don’t expect any game within launch year to be able to particularly push the limits of I/O.

 

[pjbliverpool]

@chris1515 I think the issue is you are arguing with PC elitists who are happy to just spend as much as it takes and use pure grunt to get an upgrade. What Sony has done for the cost is simply amazing, and what MS are doing will bring similar results for Xbox and PC from what I can gather.

No need to pay big bucks for more RAM - the problem the PC guys aren’t considering is that almost no-one would develop a game for such a tiny minority or gamers. So their whole argument is flawed and around scenarios which won’t be happening just so they can some how prove the SSD isn’t a big step forward for games, graphics and developers.

Yeah I'm also having trouble with the "just put in 64GB of RAM silly!" kind of arguments. Obviously that's an option, but one that makes no sense for consoles and basically never going to happen even in the next gen, and no PC game will require that amount anyway, so I'm not sure I understand the point of this line of discussion.

Guys, this isn't the argument I'm making at all. My argument actually has nothing whatsoever to do with PC's (although elements of if are definitely transferrable to that space). This is about whether having more VRAM is of benefit to rendering regardless of whether you have a PS5 class IO system or not. Perhaps I should have just opened with "a PS5 with 32GB RAM would be a more capable gaming machine than the PS5 we got".

I'm sure that would have been a lot less contentious but I thought I try for something a little more discussion worthy by highlighting that a console with 32GB RAM and a SATA SSD, or a spinning HDD with 64GB RAM would not have been universally inferior to the current NVMe + 16GB RAM setup but would rather come with a different set of costs and benefits. The costs being around initial load times and some loss of flexibility in streaming, but that IMO can be mostly mitigate by the larger VRAM pool for pre-caching. The benefits being that you actually have 32GB of RAM to work in rather than 16GB.

 

[pjbliverpool]

https://twitter.com/x/status/1392487201754161154

If this is true I think this should help a lot with the portal mechanic.If the game is loading from the SSD on the fly and only working on the RAM the absolute necessary, then the amount of data it needs to load and render is very small. It'll not lead "the other level", it'll load just what is in front of your character's eyes when it gets there.
Wonder if this is really all true, that the SSD is really being used this way.

This actually gets to the heart of my point.

"That lets us devote all of our system memory to the stuff in front of you right now,"

So if you believe what's being said there then they're using the full 16GB to render the current viewport only. Therefore with 32GB, they could do even more in that viewport.

Of course I don't take what's being said there at face value though given that would pretty much result in you consuming the entire game content by simply turning 360 degrees!

 

[PSman1700]

Arguably, the biggest largest gap from 2013 consoles is SSD performance.

Yes, but a 100x increase in GPU power would result in a larger difference i think, its also hard to compare as gpus tend to come with new features which dont always translate into just numbers increases.

Anyway how some turn this into a pc vs console war..... the pc has got its 7gb/s nvme and with rtx io/ds even faster. Theres no reason for pc vs console claims.

 

[Jay]

So if you believe what's being said there then they're using the full 16GB to render the current viewport only. Therefore with 32GB, they could do even more in that viewport.

Can do that because the ssd is fast enough.
More memory and slower storage may not be able to do the equivalent.
With 32GB may not be able to unload what's behind you as you can turn round, but then may not be fast enough to quickly portal through 2 different locations right after eachother quickly.

More memory you probably would want even faster storage to make better use of it.

 

[iroboto]

Guys, this isn't the argument I'm making at all. My argument actually has nothing whatsoever to do with PC's (although elements of if are definitely transferrable to that space). This is about whether having more VRAM is of benefit to rendering regardless of whether you have a PS5 class IO system or not. Perhaps I should have just opened with "a PS5 with 32GB RAM would be a more capable gaming machine than the PS5 we got".

I'm sure that would have been a lot less contentious but I thought I try for something a little more discussion worthy by highlighting that a console with 32GB RAM and a SATA SSD, or a spinning HDD with 64GB RAM would not have been universally inferior to the current NVMe + 16GB RAM setup but would rather come with a different set of costs and benefits. The costs being around initial load times and some loss of flexibility in streaming, but that IMO can be mostly mitigate by the larger VRAM pool for pre-caching. The benefits being that you actually have 32GB of RAM to work in rather than 16GB.

Exponential asset sizes is the general argument against just having a larger vram pool. It’s not efficient because it each time you want to increase asset sizes vram must increase proportionally to your cache sizes. In this sense it’s an inferior solution to having a very fast nvme feed the GPU JIT or the frame before.

 

[megre]

Typically this is why development hardware is usually several magnitudes beyond what's on console, but the hard drive streaming problem is a little awkward. Because we don't have directstorage APIs resolved yet, and we don't have GPUs with huge amounts of VRAM. So you're left to get the fastest NVME drive possible and perhaps bulk up with 64BG of VRAM, and maybe then your editor will work as to how it would work on PS5, but you've got this really complex job now of downgrading the game to fit within < 16GB of total memory.

I have an idea but I don't know if it makes sense.

128GB of main RAM simulating the SSD. The data inside the main RAM are decompressed already so check in and all other processes are done already. The only remaining process will be CPU sending the data to VRAM when needed.

 

[pjbliverpool]

Can do that because the ssd is fast enough.
More memory and slower storage may not be able to do the equivalent.
With 32GB may not be able to unload what's behind you as you can turn round, but then may not be fast enough to quickly portal through 2 different locations right after eachother quickly.

More memory you probably would want even faster storage to make better use of it.

Let me give another scenario. If R&C truly is able to load data behind the character from the SSD in real time then presumably turning speed must be limited in line with streaming capacity. This is similar to how last gen games limit traversal speed in line with last gen streaming capabilities. So you couldn't, for example have a "look backwards" button, or enable mouse control. However with 32GB of memory you could keep the data behind as well as in front of the character in memory, thus allowing the above scenario's because there's no dependency on the SSD to load that data in when the character turns around/looks behind them. This is just one pretty basic example but I'm sure good developers could come up with lots of scenario's where it's better to have more RAM, than ultra fast streaming speeds. In fact I'd bet that ultra fast in game streaming won't play a particularly big part in many games this generation because of the storage limitations. I've said it countless times before on this forum but you're not going to be regularly streaming at the PS5's maximum IO rate because if you were, you'd be consuming your entire game content in a matter of seconds. There will of course be in game scenario's when you want the SDD to be maximising it's transfer speeds but unless we expect every game this generation to use Rift like portal mechanics then I expect them to be quite a bit rarer than many here are assuming.

 

[Jay]

I've said it countless times before on this forum but you're not going to be regularly streaming at the PS5's maximum IO rate because if you were, you'd be consuming your entire game content in a matter of seconds.

Maybe not.
But you still have to stream in more data from a slower drive to fill the buffer.
If you can jump to a whole new level, then you have to fill the larger buffer from a slower drive.
That inherently means possible change in game design.

 

[Arwin]

PS4 could do Rift Apart for a big part with the new IO system. The reverse is not true. You may want to think that over. As the tweets from the guy from Naughty Dog suggests, the biggest bottleneck for using higher quality assets at all times is the speed of getting them from disk. There are some amazing graphics and levels in Uncharted 4 at points where speed isn’t needed, and they hint at what will be possible when games are designed for the new tech.

In addition, you should see the data on disc not just as blocks filling one environment and simply calculate 16GB of memory means I blow through a 40-80GB game in 3-5 seconds. Even apart from compression, levels are built from building blocks that can be reused a lot even without seeming to be the same block as before. Uncharted is really good at this. And that means that if you have 80GB of *parts*, they can be combined in any possible way partly thanks to the fact that you can get at them super fast no matter their location at any part of the game.

I think some people fail to understand that the bottleneck being solved here is not like bottlenecks you have seen a lot previously, simply because older games entirely constrained by this bottleneck only show some framerate or resolution improvements on newer and faster machines.

 

[Silent_Buddha]

SSDs are a substitute for ram not processing power as i understand it. There is nothing the SSD affords that more ram would not. Processing power however is irreplaceable.

Basically, yes. Years and years ago in the one of the next gen console threads in this forum when people were talking about next gen consoles (this current gen) coming out with 24+ GB of storage, I'd mentioned that this was not practical unless we had a massive increase in storage speeds. With 8 GB of memory, load times on consoles were already getting to to point where they were insanely long and mechanical HDDs had basically plateaued WRT speed (there are minor year over year increases) due to R&D in mechanical storage devices not seeing an easy path to cost reductions and long term reliability with next gen HDD platter innovations (HAMR for example) that would make them economically feasible for the consumer market.

I'd postulated that the next gen consoles would need at a bare minimum of an SSD in order to support more memory. However, there was a significant counterpoint to SSDs being included in consoles at the time and that was cost per GB. I felt that at the rate SSDs were cost reducing at the time that it would be feasible that it could potentially be used in the next generation of consoles (this generation), but others obviously felt differently. Then we had a relatively massive decrease in the rate at which NAND was cost reducing a few years back and I stopped talking about it.

And here we are. SSDs in consoles which allow for the usage of larger pools of memory without massively increasing (doubling or tripling) load times compared to the previous generation of consoles. However, 2 things.

• Memory only doubled (not even doubled going from XBO-X to XBS-X), limiting to some extent how much we can benefit from SSDs if I/O subsystems remained the same
• SSD storage capacity is greatly limited, thus limiting the size of a game and thus the fidelity of the textures and art assets used in the games.
  • XBS systems get around this via Hot Swappable storage modules, effectively giving the console unlimited storage if you can afford multiple storage modules.
  • PS5 systems will eventually get the use of NVME drives, but maximum storage will be limited by the maximum available NVME drive that can be used in the system.
  • Greater compression ratios in this generation of consoles also help with this.

That said, for the first bullet point above, both Sony and MS have come up with changes to the I/O subsystem (hardware and software) that allow for lower latency streaming of data from the SSD which amplifies the effect of memory capacity. IE - when compared to previous I/O subsystems the current generation of consoles have a greater than 2x increase in effective memory capacity since you can stream in some data on demand to be used in rendering the next 1-2 frames.

As to the second bullet point. While this will reduce overall what the current gen consoles and I/O subsystems are capable of (the size of the game and thus assets is far more limited due to the size of the SSD drives), it's still going to be very significantly improved compared to the previous generation of consoles.

All of which means that games will have greater graphical fidelity combined with increased convenience (greatly reduced loading times versus 2-3x increase in loading times) than if we'd kept using mechanical HDDs and used those savings to instead increase the transistor count of the SOCs with an attendant bump in the GPU (don't forget the cooling system needed for an SOC that would consume even more power).

As I said long before the consoles came out. The rumor that the current gen consoles would be using SSDs was far more exciting than if they had more powerful GPUs. The other thing that excited me was the audio, but thus far the only next gen audio experience I've had is with Cyberpunk 2077 ... on PC. So ... hopefully some developer really uses the audio hardware that Microsoft and Sony have put into their SOCs at some time.

Regards,
SB

 

[Silent_Buddha]

https://twitter.com/x/status/1392487201754161154

If this is true I think this should help a lot with the portal mechanic.If the game is loading from the SSD on the fly and only working on the RAM the absolute necessary, then the amount of data it needs to load and render is very small. It'll not lead "the other level", it'll load just what is in front of your character's eyes when it gets there.
Wonder if this is really all true, that the SSD is really being used this way.

Thinking about this, I wonder how this would hold if you had on PC controls with PC mouse sensitivity settings?

On console you are limited by how quickly the game allows you turn the camera view when the controller stick is at maximum distance to the right or left. This will generally take multiple frames on console (usually around 5-10 for really fast action games like COD with max sensitivity and 20-30 frames for many other games) versus many twitch gamers on PC which set their sensitivity so they can make a 180 degree turn in 1 to 2 frames. Those numbers are for 60 FPS games to give a reference. For the games on console that allow KBM to be used, I'd imagine they have similar limitations on how much sensitivity you can give the mouse.

Still really impressive, but people should keep in mind that when comments like that come out, the developer may not be saying that they load all of that data in 1 frame of rendering (16.7 MS for a 60 FPS game) while simultaneously rendering the scene.

Regards,
SB

 

[VFX_Veteran]

Again the geometric level of detail inside the demo is too big for a game not because they can't render it but because the size of the data is too big. And Brian Karis told performance is not a problem and read the tweet thread he created, he explains performance is improving because this is a new technology.

I don't think you understand what I'm saying. The GPU isn't an empty void where you can throw any amount of triangles at it and it just renders it easily no matter what resolution, no matter what complex algorithm you introduce (i.e. RT lighting for example), no matter what advanced algorithm you are using in your game. That's just objectively not possible. The PS5 (as well as any other GPU) has a rendering limit that is completely independent of the speed of the SSD->VRAM transfer rate. If it didn't, then Nvidia and Sony's job would be done. We've achieved the perfectly limitless GPU everyone has been wanting.

If you think the faster loading is more of a limit, I question, why not render that UE5 demo at native 4k/60FPS. My answer? The GPU was limited to that particular bandwidth (ala 1440p/30FPS) with those particular rendering parameters and that particular scene complexity.

 

[PSman1700]

I don't think you understand what I'm saying. The GPU isn't an empty void where you can throw any amount of triangles at it and it just renders it easily no matter what resolution, no matter what complex algorithm you introduce (i.e. RT lighting for example), no matter what advanced algorithm you are using in your game. That's just objectively not possible. The PS5 (as well as any other GPU) has a rendering limit. If it didn't, then Nvidia and Sony's job would be done. We've achieved the perfectly limitless GPU everyone has been wanting.

If you think the faster loading is more of a limit, I question, why not render that UE5 demo at native 4k/60FPS. My answer? The GPU was limited to that particular bandwidth (ala 1440p/30FPS).

Just look where DF sticks likes This whole debate only shows us one thing, some are letdown by what the gpu specs where.

 

[manux]

I don't think you understand what I'm saying. The GPU isn't an empty void where you can throw any amount of triangles at it and it just renders it easily no matter what resolution, no matter what complex algorithm you introduce (i.e. RT lighting for example), no matter what advanced algorithm you are using in your game. That's just objectively not possible. The PS5 (as well as any other GPU) has a rendering limit that is completely independent of the speed of the SSD->VRAM transfer rate. If it didn't, then Nvidia and Sony's job would be done. We've achieved the perfectly limitless GPU everyone has been wanting.

If you think the faster loading is more of a limit, I question, why not render that UE5 demo at native 4k/60FPS. My answer? The GPU was limited to that particular bandwidth (ala 1440p/30FPS) with those particular rendering parameters and that particular scene complexity.

Unreal5 is designed to render pixel sized geometry derived from original art asset. UE5 uses streaming to achieve required asset quality without load times. UE5 requires ssd, what speed of ssd depends on compression and resolution. Both gpu and ssd is needed in case of ue5. It's not one or the other kind of proposition.

 

[Globalisateur]

Thinking about this, I wonder how this would hold if you had on PC controls with PC mouse sensitivity settings?

On console you are limited by how quickly the game allows you turn the camera view when the controller stick is at maximum distance to the right or left. This will generally take multiple frames on console (usually around 5-10 for really fast action games like COD with max sensitivity and 20-30 frames for many other games) versus many twitch gamers on PC which set their sensitivity so they can make a 180 degree turn in 1 to 2 frames. Those numbers are for 60 FPS games to give a reference. For the games on console that allow KBM to be used, I'd imagine they have similar limitations on how much sensitivity you can give the mouse.

Still really impressive, but people should keep in mind that when comments like that come out, the developer may not be saying that they load all of that data in 1 frame of rendering (16.7 MS for a 60 FPS game) while simultaneously rendering the scene.

Regards,
SB

That's some impressive goalposts moving there. I applaud you sir.

 

[VFX_Veteran]

No GPU in 10 years will be able to render a trillion visible triangles in a frame at 60fps.thus LOD is needed, could the above R&C city scene vista contain a trillion triangles? Yes it could (perhaps it does?)

That's a confusing statement since the PS5 has a GPU. Are you saying no PC GPU will render a trillion visible triangles at 60FPS but the city scene vista (which clearly has a background layer that's unplayable) is rendering a trillion triangles? The GPU renders triangles to the screen not the SSD.

 

[VFX_Veteran]

Unreal5 is designed to render pixel sized geometry derived from original art asset. UE5 uses streaming to achieve required asset quality without load times. UE5 requires ssd, what speed of ssd depends on compression and resolution. Both gpu and ssd is needed in case of ue5. It's not one or the other kind of proposition.

I'm aware of what the UE5 demo does. But no graphics programmer is going to tell you that the SSD renders any triangles. The GPU does. It may be designed to render pixel-sized triangles, but when you rasterize those triangles, you have all the assets in VRAM and now have to run shaders which is completely done by the GPU. UE5 demo could switch to a RT lighting paradigm (their code is already in 4.26) and how much would that cost for the PS5 GPU? Answer: ALOT. The lighting/shading of a pixel within a triangle will always be a bottleneck. That's what separates a visuals from realtime to offline.

 

[Arwin]

The UE5 demo on PS5 may well have had a higher resolution and/or framerate with a stronger GPU and the same IO subsystem, depending on the VRAM bandwidth bottlenecks VS the CU bottlenecks.

And that’s what you’ll see across the board. Way more variation and density and better and more textures, regardless of GPU, but the stronger or weaker GPU will show more or less pixels and framerate, but have no further impact on the quality and detail of what is shown.

In the case of this demo, it would have shown a black screen with that same stronger GPU but 50MB/s HDD / old IO system. Wake up man

 

[Karamazov]

I remember an old comment from a dev at the time of the ps1 when sony announced it could render 1.5 million polygons flat per second. The dev said "what's the point of being able to render that much polygons if its memory system can only handle 150 000 of them ?"