not if the monitor is 1080p.
Just joking.
There's a wide range of gamers, and setups, Lockhart could give xbox a good base to build from both from early to mid timelines in sales.
Xbox Series S [XBSS] (Lockhart) General Rumors and Speculation *spawn*
- Status
Just joking.
There's a wide range of gamers, and setups, Lockhart could give xbox a good base to build from both from early to mid timelines in sales.
Unknown Soldier
Veteran
From what I understand from the XBSS rumours is that it will run a 4TF GPU that has RDNA2 and perform like a Xbox One X with it's 6TF GCN2.0 GPU ala 4K 30fps in certain games and the rest 1440p at likely 60fps. Of course if certain games are going for 4K 60fps with the XBSX then it's likely the XBSS will only do 1440p 60fps, at least that's what I assume.
It would be a joke if the XBSS only does 1080p 60fps if the XBSX does 4K 60fps.
It would be a joke if the XBSS only does 1080p 60fps if the XBSX does 4K 60fps.
Boo-urns.
I guess we'll have to wait and see, but I'm not getting any hopes up for it. :s MS will have to come out and demonstrate DLSS if it exists.I'm thinking the bandwidth will be less (128-bit @ 14Gbps = 224GB/s, which is roughly what a 4TF RDNA1 GPU has albeit unshared). One thing to consider is that for a particular texel density, higher resolution would see better than linear scaling as the texel per pixel increases hence only 3x power for 4x resolution so better cache hits. On the other hand, we expect texture resolutions to scale accordingly, so it could be moot depending on the title. I have no idea how things will turn out. I'm guessing if a developer is seeing performance issues at 1080p, they'll just scale the settings accordingly and be done with it. Given that multiplatform is the bulk of developers, there will be some form of scaling anyway for PC, and LH could be a good lower bound for hitting mass market there.
I think performance modes greatly depend on the CPU, so it just depends if the developer wants to scale down low enough on the GPU to hit 60fps. Given how reconstruction methods work better with higher frame rates, it might not be the worst. e.g. Gears 5 has 60fps on base consoles in Multiplayer/Horde/Escape and they just drop resolution, but in my experience, the reconstruction is good enough while the performance is a much bigger boon to gameplay experience.
It might just become ubiquitous as the console hardware becomes less and less exotic with compatibility and generations becoming more performance-based. Can't say.
Fidelityfx is platform agnostic so I don't see the issue. It runs on my vega gpu and it runs on rdna 1. So i don't see it suddenly not working with RDNA 2. The question is if they have added anything to it in RDNA 2. MS has said they have added ML hardware to the XSX. That hardware could be used to improve Fidelity FX or AMD may have improved it on their own. We will have to wait for hot chips or for amd to announced RDNA 2.
Here it is on a 5700
So like I said , I'm expecting an improved edition of this and with amd's new hardware supporting machine learning i would assume amd would leverage that like Nvidia has.
Now of course this is not as good as DSLL but again it doesn't require any special hardware at this time.
I don't consider a sharpening filter anywhere near what DLSS does for nVidia.
DLSS uses the tensor cores, which are in abundance on RTX GPUs with higher TOPs compared to what AMD CUs would offer. Given how R&D budgets are with nVidia, I'm not sure AMD will come up with anything better with less HW performance.
DLSS is better but it requires specialized hardware. Maybe AMD can't match DLSS but I am sure they can improve upon Fidelity FX and may even have additional hardware to assist it in RDNA 2. I dunno all the features of RDNA2 or what is in the xbox / ps5.I don't consider a sharpening filter anywhere near what DLSS does for nVidia.
But does it matter ? Even if we just get the current Fidelity FX implementation on Lockhart at the end of the day your rending a 4th of the pixels most likely with a third of the power. Fidelity FX is a feature that can allow it to punch further above its weight class.
I'd also wager for AMD getting their successor to Fidelity FX or even just Fidelity FX into Lockhart / XSX games will only improve their lot in the pc game. If they can leverage MS or Sony to improve upon it by writing ML algorithms that leverage the rdna 2 like hardware in those consoles. I am sure AMD would love that also as they can then implement it on the pc side.
What kind of memory configuration are you expecting on Lockhart?
My reasoning, working backwards like Ben Affleck in the John Woo classic Paycheck, is that if the game available memory is 7.5 GB, and you take the same 2.5GB system reserve from Snek, that leaves you with 10GB. Which is an odd number admittedly as none clamshell it would leave you with a 160-bit bus, meaning simply not using half of a 64-bit memory controller.
I suppose you could use five 16 Gbit chips and clamshell two of them .. if the controller could handle that. I suppose another possibility might be a 192-bit bus, with a mixture of 8 and 16 Gbit chips, similar to the XSX's interesting arrangement. So maybe 6 GB fast across 192-bits o bus, and 4 GB slower across 128-bit. I'm quite attached to the 10GB number at the moment. It's my current thing.
--------------------
But yeah, it's a fair point that things can get better utilised as resolution increases, at least in terms of the 3D pipeline rather than compute (afaik). Use of suitable texture and geometry lods should help a lot I'd guess, which PC reviews understandably keep as consistent as possible at different resolutions. Unreal 5's use of compute shaders and "one poly per pixel" stuff should help games to scale well across resolutions in future ... maybe.
Interestingly, comparing TPU reviews it looks like the 5500XT might scale more closely with resolution than the 5600 or 5700XT. Which seems a bit weird as the 5500XT seems to have more CUs per shader engine. Though it's hard to know what role the CPU plays in it.
Hmm. An evening of looking at graphs I think.
Was thinking perhaps clamshell. 6x1GB + 2x2GB for 10GB. Down the road it'd be 3x2GB + 1x4GB.
Hadn't considered that funnily enough, but I suppose the XSX could be mixing different densities on different channels on the same controller, or maybe clamshell can support different densities on either side (never even considered it tbh). But yeah it would allow for the smallest board and for better cost savings down the line so it's got to a prime contender. I think I'll switch my bet to this suggestion.
224 GB/s would be a bit disappointing IMO given that it's got 8 fast Zen 2 cores on there, but OTOH I suppose the 5500XT is 5.2 TF as opposed to 4 so maybe it's not so bad. Hopefully it's a full 8 16-bit channels and all that.
Still holding out for 32 ROPs though!
I don't think so because the CPU can't handle more than 100 GB/s anyways, and the GPU will scale, but who knows what they may come up with. I'm trying to find the CPU technical tidbit.
Discussion about marketing and relative positioning or involving Sony moved to https://forum.beyond3d.com/threads/marketing-next-gen-consoles-2020-spawn.61734/
D
Deleted member 13524
Guest
The bulk of the conversation was about what the resolution of Lockhart / Series S will render at if the game's SeriesX counterpart targets resolutions below 2160p, specifically 1440p to 1800p like we've been seeing this gen.
It makes no sense to move that conversation to the marketing thread.
It makes no sense to move that conversation to the marketing thread.
- Status
