Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

[psorcerer]

Assuming the Player does not play with high control Stick Sensitivity, or the game offers mouse Support. I do not think that was a good example.

270ms for 2GB is pretty good.
I would say that places it in the low speed DDR3 ballpark for gathering read.
Maybe they indeed wanted to get to DDR3 speeds.

 

[Shifty Geezer]

Worse if RT is involved. The wider GPU will win here owing the to more RT cores.

It'll be proportionally the same for RT as rasterising. If PS5 is 80% of XBSX when rasterising, it'll be 80% of XBSX on the GPU side when raytracing. And if rendering 80% of the pixels, it'll need cast 80% of the rays.

 

[scently]

Its not elegant its a cost cutting design, and whilst it is a smart way to drive down costs theres no performance benefit i can see to doing this.

These are consoles so cost is obviously going to be an issue otherwise we would have 32gb and so on. In that light I think they made a smart decision based on budget and their objectives. Unified memory with 10gb@560gb/s for GPU load which is the subsystem that consumes the most bandwidth, and 3.5gb@336gb/s for other processes.
Its a cost saving design it allows them to achieve their objective. Its a smart design.

 

[Shifty Geezer]

https://richg42.blogspot.com/2016/08/rads-ground-breaking-lossless.html
A benchmark of Kraken the solution Sony uses.

Best. Graph. Ever!

 

[Shifty Geezer]

The memory speed hardware specs actually seem to be really close, if you average the total bandwidth of the 16GB on Xbox you get 476GB/s vs the 448GB/s on the PS5. So it’s going to be very interesting to learn what the advantages of doing this end up being.

Doesn't PS5's get more impacted by contention though? I'm completely at a loss how that works!

 

[London Geezer]

Best. Graph. Ever!

That's the most beautiful thing I've ever seen without understanding what it is.

 

[milk]

Its not elegant its a cost cutting design, and whilst it is a smart way to drive down costs theres no performance benefit i can see to doing this.

The benefit is being able to ship with 16Gbs of RAM thanks to those trade-offs. Maybe without it, they would have had to stick with 12Gigs to hit their budgets. Would you find that more elegant?

I always apreciate these more fine-grained solutions. They show the design was near the boundaries of executability. I wish ps4 had done something similar with its RAM and then could have had even more memory in there for the same cost.

 

[DuckThor Evil]

Microsoft's memory config is great. It's 320Bit vs 256Bit where it matters.

There is no penalty with the asymmetrical design in any relevant scenario as it has more than enough of fast speed memory. The function of it is simply to keep the memory amount at 16GB.

 

[anexanhume]

10 GB seems pretty generous for the high-speed partition. I presume that Microsoft's profiling indicated that there would be enough data that was not the critical footprint of the most demanding processes, at least for this stage in the generation's life cycle. If developers are able to move the majority of their bandwidth-limited work into that window, the average tilts significantly towards the maximum. At least until there are engines that aren't cross-gen and have been built to really push the architectures, it seems like they can generally do better than the average.

Microsoft's memory config is great. It's 320Bit vs 256Bit where it matters.

There is no penalty with the asymmetrical design in any relevant scenario as it has more than enough of fast speed memory. The function of it is simply to keep the memory amount at 16GB.

The key thing is that if memory doesn’t have the right texture data, they said it’s only 1-2 frames before they load it, and they blend the texture before that. I think that’s a very minimal impact and a smart compromise. It’s a lot less severe version of texture pop-in that we’re used to now.

 

[scently]

The key thing is that if memory doesn’t have the right texture data, they said it’s only 1-2 frames before they load it, and they blend the texture before that. I think that’s a very minimal impact and a smart compromise. It’s a lot less severe version of texture pop-in that we’re used to now.

The 1-2 frames is in relation to their velocity engine the ssd and not the main memory.

 

[Barrabas]

That's the most beautiful thing I've ever seen without understanding what it is.

Abstract art: "The artist communicates something that is almost intangible yet a potent message gets through".

 

[iroboto]

It does tie up with the feedback/rumours we got from some developers, who were raving about this supposed 'texel per second' (or whatever it was), meaning that textures on screen at any give time are pretty much always 'perfect'.

This SSD thing is my number 1 most exciting aspect of the PS5. The rest, whatever, it means nothing. Bandwith/TF will mean resolutions will be what they need to be.

The SSD wil make an actual, obviius difference apart from obviously faster loading.

Imagine Spiderman 2 on PS5 without the limitations that the game had to work around because of the ancient HDD.

Hell, any game with 'levels' would be revolutionised and annoying corridors (Hello, Bloodborne!).

It's the end of levels.

Who cares if Devs want levels.

Sony to ban all Levels in games. The end. Order reinstated.

I feel like PS5's SSD is the only bespoke piece of console hardware in the last 2 generations. It can only work in the console space, it can't be back ported to PC. It's specific for console needs. Yea it's definitely the most important piece.

For obvious reasons, MS could not go that route. If that hardware doesn't exist on PC, then they have to be really careful about bringing it to console. Their solutions are I/O based application layer solutions, so i'm thinking SFS and DirectStorage API's make up a bulk of XVA's application layer, with likely HBCC or some form of custom HBCC making up the physical layer of it.

Where Sony developed hardware to have pure power in hard disk transfer potential, MS aimed to improve throughput with less transfer performance. The latter being portable to the PC space the former not.

I think despite being down on GPU grunt, it's still possible for PS5 to pull out a win graphically. The SSD could enable them to have higher fidelity textures at further out distances which could result in better overall graphics despite falling back on resolution or frame rate. Having next to no pop in with immaculate detail is in itself graphical prowess we've not seen before.

 

[anexanhume]

The 1-2 frames is in relation to their velocity engine the ssd and not the main memory.

I don’t understand the importance of the distinction given the transfer from that velocity engine to main memory is on the order of 1-2 frames given the drive speed. We’re talking about a texture, not flushing the whole RAM.

 

[Dictator]

I feel like PS5's SSD is the only bespoke piece of console hardware in the last 2 generations. It can only work in the console space, it can't be back ported to PC. It's specific for console needs. Yea it's definitely the most important piece.

For obvious reasons, MS could not go that route. If that hardware doesn't exist on PC, then they have to be really careful about bringing it to console. Their solutions are I/O based application layer solutions, so i'm thinking SFS and DirectStorage API's make up a bulk of XVA's application layer, with likely HBCC or some form of custom HBCC making up the physical layer of it.

Where Sony developed hardware to have pure power in hard disk transfer potential, MS aimed to improve throughput with less transfer performance. The latter being portable to the PC space the former not.

I think despite being down on GPU grunt, it's still possible for PS5 to pull out a win graphically. The SSD could enable them to have higher fidelity textures at further out distances which could result in better overall graphics despite falling back on resolution or frame rate. Having next to no pop in with immaculate detail is in itself graphical prowess we've not seen before.

Do you honestly see texture pop-in in modern game engines not on console anymore? On PC if I see that I always just look at the vram amount utilised and notice the game is actually completely under utilising it or is using an old system built around GPUs with little to no vram.

I think the mesh shader is going to be the actual 'no more popin' / way more detail thing that next gen GPUs give us.

One Thing to Note about the ssd is that the data it is bringing in is static, which means more unmutable disc data. That Sounds cool for a Single Player game where every bit and bolt is predesignes (if you want to spend your time doing that), but does not seem like a perfect fit for the open world game development where procedural solutions requiring less artist and dev time increase the game's scope and detail.

Just had all this conversation with the Star Citizen devs about how Real time prcoedural things make their detail and scope possible for development.
Hence why they use a prcedural damage System instead of a Hand made premade one that swaps into memory.

 

[scently]

I don’t understand the importance of the distinction given the transfer from that velocity engine to main memory is on the order of 1-2 frames given the drive speed. We’re talking about a texture, not flushing the whole RAM.

Because we were talking about the choice of having some of the RAM at a lower speed. I am not sure what that has to do with the velocity engine in relation to the smoothing within a frame or 2.

 

[iroboto]

Do you honestly see texture pop-in in modern game engines not on console anymore? On PC if I see that I always just look at the vram amount utilised and notice the game is actually completely under utilising it or is using an old system built around GPUs with little to no vram.

I think the mesh shader is going to be the actual 'no more popin' / way more detail thing that next gen GPUs give us.

One Thing to Note about the ssd is that the data it is bringing in is static, which means more unmutable disc data. That Sounds cool for a Single Player game where every bit and bolt is predesignes (if you want to spend your time doing that), but does not seem like a perfect fit for the open world game development where procedural solutions requiring less artist and dev time increase the game's scope and detail.

Just had all this conversation with the Star Citizen devs about how Real time prcoedural things make their detail and scope possible for development.
Hence why they use a prcedural damage System instead of a Hand made premade one that swaps into memory.

hmm, this might be my fault because I run all my games off a Western Digital Black as opposed to a SSD, my SSD is so tiny i keep it for OS only.

But for instance, Apex Legends or any BR title where you drop into the zone, even at 1080p and reduced graphic settings to keep my framerate above 120, it's a blurry mess for a while until everything loads in. That can also be a function of having a weaker CPU as well (i5 6600). I don't honestly know, but I have 24 GB of system memory so somehow my setup is not being taken advantage of.

I agree, that the challenge for maximizing the bandwidth on SSD will be on the developers to produce the content, which is a different type of challenge. This might be where outsourcing becomes a bigger thing, or more reliance on purely texture pack based companies (aka what we saw with Death Stranding).

I agree with going procedural generation for open world in the future, it just makes economical sense given the size and scopes that people want. But at the same time, Sony has been more than happy to pay out to it's studios to get the work done.

 

[anexanhume]

Because we were talking about the choice of having some of the RAM at a lower speed. I am not sure what that has to do with the velocity engine in relation to the smoothing within a frame or 2.

And about the impacts of being constrained to 10GB. That was my point.

 

[Kaotik]

And about the impacts of being constrained to 10GB. That was my point.

You're not constrained to 10GB, it's just faster to use.

 

[DavidGraham]

For obvious reasons, MS could not go that route. If that hardware doesn't exist on PC, then they have to be really careful about bringing it to console.

PCs don't care, they can always brute force their way out of any storage problem with more RAM, beyond 2020 most above average PCs will have 24GB of RAM or 32GB, problem solved.

DirectStorage will be much helpful there too.

it's still possible for PS5 to pull out a win graphically. The SSD could enable them to have higher fidelity textures at further out distances which could result in better overall graphics despite falling back on resolution or frame rate.

Textures need meshes to be put onto, if you are underpowered in CPU and GPU you will use less of these far meshes than Xbox, otherwise your performance will drop.

 

[iroboto]

I mean, even in this demo here:

When you see the XSX load, you still see pop-in happening. I'm not sure if this is a result of poor code or what not, I'm open to suggestions, but even with the SSD and the CPU and the GPU; it's still present even if for a split second.