Next-Generation NVMe SSD and I/O Technology [PC, PS5, XBSX|S]

With SSD storage and compression getting so fast, it seems that'll be the method of choice and tiled resources will be on the back burner for years to come, if not completely relegated to 'novelty'. It's hard to come up with a game use-case that requires it when you can just stream everything in using the conventional ways. the 'necessity' here is working smarter and more efficiently, and that's never a priority for humans!
 
With SSD storage and compression getting so fast, it seems that'll be the method of choice and tiled resources will be on the back burner for years to come, if not completely relegated to 'novelty'. It's hard to come up with a game use-case that requires it when you can just stream everything in using the conventional ways. the 'necessity' here is working smarter and more efficiently, and that's never a priority for humans!

Using tiled resources would allow for more unique textures and even more efficient use of VRAM would it not?
 
Unreal Engine 5 use virtual texturing and virtual geometry for environment. But you have BVH to stream too on consoles. If a game use for example SDF for raytrace against it is possible for static part of the scene to stream it from the SSD. Other assets to load are animations, sounds, geocache alembics stuff for crazy set pieces.


And after you have portal stuff and it could much more complicated than in R&C Rift Apart like a Dr Strange game one day. The latency is so low in SSD NVME you can load something for next frame if it is not too big and use less memory, load just in time a sound for example when a character die.


storage-sata_vs_sas_vs_nvme-f.png
 
Last edited:
Using tiled resources would allow for more unique textures and even more efficient use of VRAM would it not?
If you cannot develop your own VT system then yes. If so, then maybe not. TR and SFS as well as VRS work on fixed sizes IIRC. If you’re willing to build your own custom systems in compute you will likely have better usage over your hardware there.
 
With SSD storage and compression getting so fast, it seems that'll be the method of choice and tiled resources will be on the back burner for years to come, if not completely relegated to 'novelty'. It's hard to come up with a game use-case that requires it when you can just stream everything in using the conventional ways. the 'necessity' here is working smarter and more efficiently, and that's never a priority for humans!

That's one way to look at it. Another is perhaps SSD's fast access of granular fragmented data might entice devs to make more use of virtual texturing (tiled resources or not) than they did before.

What is the use of having a sophisticated virtual texture system, if in the end your engine is streaming assets in huge packaged chunks to minimise seek times from the HDD.

When you remove that bottleneck, the gains of virtual texturing become even larger.
 
Hence we see relatively limited attempts to leverage the fast storage available.
We're seeing that because broader decisions to not build any proper next gen games seems to be the norm for the industry, at least up til now. You cant exactly build games with this fast storage in mind, and still offer last gen versions of them. At least not without any great effort where they're essentially two very wholly different builds of the game.

These SSD's weren't included in the new consoles as a sort of optional fun feature for devs if they want to use them. They're pretty much there as a replacement for a huge memory quantity increase. It's going to be extremely difficult for developers to make a large, ambitious next gen leap in presentation levels without leveraging these SSD's to a reasonable degree since they're otherwise stuck with an extremely limited memory capacity improvement over what they had before.

Making use of fast storage is going to become imperative going forward. And there will be no going back. It will be a paradigm shift, not something certain devs use while others continue to work in the 'old ways', at least among larger studios making demanding games.
 
It's going to be extremely difficult for developers to make a large, ambitious next gen leap in presentation levels without leveraging these SSD's to a reasonable degree since they're otherwise stuck with an extremely limited memory capacity improvement over what they had before.

I agree that this is the case for high fidelity AAA games, but the success devs and publishers have been finding with cross-gen games, and specially how these expensove new consoles keep selling out despite having close to no exclusively next-gen content, makes me wonder how many contemporary devs will find it necessary to be at the forefront of graphical/engine technology.

I know some studios will forever do, because its in the DNA, but I'm not sure if all the guys at the top of production value of PS4/One gen will fight tooth and nails tonstay in that bracket. Perhaps a resurgance of a Neo-AA kind of studio may be in order.

Graphics don't seem to sell as much these days. We've perhaps finally reached good enough for the majority of consumers.

As a gamer, thats great news actually. As a graphics enthusiast, not so much...
 
In order to implement DS1.1, does a developer still need to implement traditional solutions to support PS4/5, Xbox One crossgen and linux for steam?
If so it sounds like a lot of work for some dev.
 
In order to implement DS1.1, does a developer still need to implement traditional solutions to support PS4/5, Xbox One crossgen and linux for steam?
If so it sounds like a lot of work for some dev.

I don't expect DirectStorage game to release on PS4 and Xbox One. For cross platform game PS5 and Xbox Series will be ok. From what DF said they heard from devs the SSD API is easier to use on PS5 than DirectStorage.
 
I agree that this is the case for high fidelity AAA games, but the success devs and publishers have been finding with cross-gen games, and specially how these expensove new consoles keep selling out despite having close to no exclusively next-gen content, makes me wonder how many contemporary devs will find it necessary to be at the forefront of graphical/engine technology.

I know some studios will forever do, because its in the DNA, but I'm not sure if all the guys at the top of production value of PS4/One gen will fight tooth and nails tonstay in that bracket. Perhaps a resurgance of a Neo-AA kind of studio may be in order.

Graphics don't seem to sell as much these days. We've perhaps finally reached good enough for the majority of consumers.

As a gamer, thats great news actually. As a graphics enthusiast, not so much...
Nah, we're not tapped out on demand for better graphics yet. Never will. Gamers just haven't gotten a better taste of what can be done with the new hardware yet. Usually by now we'd have loads of games showing off the potential of new systems and we're just not seeing that yet. It's been unprecedented, but it wont stay this way for much longer at all.

I'd love for 'AA' to make a comeback, though. But I fear as more true next gen titles come in, people's standards will go up as always, and then merely 'good' looking AA games start getting attacked for their 'bad' graphics again. We'll see.
 
I don't expect DirectStorage game to release on PS4 and Xbox One. For cross platform game PS5 and Xbox Series will be ok. From what DF said they heard from devs the SSD API is easier to use on PS5 than DirectStorage.

That would make sense given what Cerny said in the road to PS5 about developers just having to tell the drive what they want and where they want it and the hardware just does it with little input from them.
 
Development of games was slow down by the COVID 19 pandemics and with current gen console shortage it is not helping to finish the cross gen period. But it seems console shortage will finish soon and from 2023 and 2024 we will see more true current gen game like Plague Tale Requiem.
 
That's one way to look at it. Another is perhaps SSD's fast access of granular fragmented data might entice devs to make more use of virtual texturing (tiled resources or not) than they did before.

What is the use of having a sophisticated virtual texture system, if in the end your engine is streaming assets in huge packaged chunks to minimise seek times from the HDD.

When you remove that bottleneck, the gains of virtual texturing become even larger.
Sure, but that's a latency thing, not a BW thing. You won't need 7 GB/s, just 2 GB/s and 10 us latency...sort of thing. The finer the granularity, the more reliance on low latency rather than max throughput. I imagine in an ideal streamed asset engine, all this high speed storage and fancy compression will be moot. We should be able to have a system with a minimum of RAM, VRAM, and throughput, and just use it all so much more efficiently. While daydreaming, chuck in a VR headset with foveated rendering and the perfect consle might be something like 4 GBs ultra fast RAM and an ultra low-latency storage solution at 2 GB/s. You could then either have a cheaper machine or move the budget on to processing.

Next-gen storage strikes me as a hardware solution to a software problem, as is often the way. And then the hardware leads, the software plays to its strengths, you get an entrenched paradigm. Once you get a point where the software paradigm can shift thanks to the hardware, the opportunity for something new has passed. If it's going to happen, it'll be a long, slow road. Still, it keeps the Tech websites in business making more and more benchmarks!
 
Last edited:
Sure, but that's a latency thing, not a BW thing. You won't need 7 GB/s, just 2 GB/s and 10 us latency...sort of thing. The finer the granularity, the more reliance on low latency rather than max throughput.

I agree that the low latency from SSDs is more of a game changer than the raw max bandwith, and have said so from the beginning. That stays true to virtualized texture streaming systems, which might be more viable and useful now than they were with HDDs.

Are you agreeing with me or not here? I don't think I understand your argument now....
 
My point is more that the pursuit of higher and higher BW via hardware and compression probably isn't necessary if the software was developing differently. The high BW of SSDs and primary debate at the moment (40 GB/s compressed data etc) isn't their main benefit, but the low latency is.
 
Sure, but that's a latency thing, not a BW thing. You won't need 7 GB/s, just 2 GB/s and 10 us latency...sort of thing. The finer the granularity, the more reliance on low latency rather than max throughput. I imagine in an ideal streamed asset engine, all this high speed storage and fancy compression will be moot. We should be able to have a system with a minimum of RAM, VRAM, and throughput, and just use it all so much more efficiently. While daydreaming, chuck in a VR headset with foveated rendering and the perfect consle might be something like 4 GBs ultra fast RAM and an ultra low-latency storage solution at 2 GB/s. You could then either have a cheaper machine or move the budget on to processing.

Next-gen storage strikes me as a hardware solution to a software problem, as is often the way. And then the hardware leads, the software plays to its strengths, you get an entrenched paradigm. Once you get a point where the software paradigm can shift thanks to the hardware, the opportunity for something new has passed. If it's going to happen, it'll be a long, slow road. Still, it keeps the Tech websites in business making more and more benchmarks!

They both matter. The greater utility of big bandwidth isn't only realized at max throughput.

In other words, low cycle latency doesn’t mean much if you have a thin pipe with a lot data to move. You can have a device where a single request performed in isolation may be outstanding but in a sea of data requests, latency may be poor because the device lacks the bandwidth to feed its computational needs.
 
Last edited:
My point is more that the pursuit of higher and higher BW via hardware and compression probably isn't necessary if the software was developing differently. The high BW of SSDs and primary debate at the moment (40 GB/s compressed data etc) isn't their main benefit, but the low latency is.

But the less memory you have the more bandwidth you need, games are more than geometry and texture. I think for example next generation maybe there won't be more RAM inside the consoles.

I need to find the article or blog post again but I think on Direct Storage side, they did some test and they think PCIE 4.0 is not enough. They need PCIE 5.0 SSD and it will be enough for the foreseeable future.

Out of portal or loading time, max throughput per second is not important but the throughput per frame is important.
 
Last edited:
Back
Top