News & Rumours: Playstation 4/ Orbis *spin*

Status
Not open for further replies.
Sony have this already, it's called GNM and according The Crew developer, "At the lowest level there's an API called GNM. That gives you nearly full control of the GPU". I'm betting this is the March's surface tiling/detiling optimisation news after several rounds of repeated retells and embellishments.

What's crazy is SCEE's own website is reporting this. :rolleyes:
It's not SCEE reporting it, it's the community blogs, this area of the website is the public playstation forums.
 
The presentation makes that very unobvious. Sure looks like an official blog article!
Yeah I hate that, when you click on root posts in the forum, it has this "blog" layout with replies under it, but you know perfectly well it's a forum because that's where you arrived first. Then if someone spread the link on the internet (or you arrived there from a google search), it looks like an official page :LOL:
 
Cerny talked early on about replicating the SPURS stuff they created for Cell being replicated in certain ways on PS4 using compute on the GPU. The reports could be referencing progress on that front.
 
Cerny talked early on about replicating the SPURS stuff they created for Cell being replicated in certain ways on PS4 using compute on the GPU. The reports could be referencing progress on that front.
I don't know, wouldn't that be just maximizing the cores utilization with parallel tasks, and making it easier to optimize highly parallel tasks? I mean it's not like this can double the frame rate or something, unless the typical game on PS4 is currently leaving more than half of the CUs idle on average. Also, anything they do with SPURS is just software, and it can probably be done equally well on XB1. If it gets them an edge it would be only temporary until the other guy catch up.

Nay, I think it's the dGPU being finally unlocked ! *duck*
 
It is just software, but if it gets significant boots from the addition compute pipes, the volatile cache tag and onion/onion+ bus options it may not be easily replicated, or provide as large a boost on Xbox One. After all, a 40% boost on a 1.8TF GPU is way bigger than a 20% boost to a 1.3TF GPU.
 
They could do a combination of things too. Certainly system reservation giveback and API optimization are almost givens. Not sure what they could get in optimizations but considering it's version 1.0 of a built from scratch API, I'd say a 20-30% API performance increase is reasonable. A release from reserve of one CPU core, 1.5GB of RAM, and half the GPU time slice (whatever that is) could be doable. The one core alone would add 17% to your base CPU capacity.

Then there's upclocking. The big question is can it be done via system update. If so, then a cpu/gpu increase to 2.0Ghz/900Mhz is within current running conditions for those parts. Otherwise, a 20% increase chip-wide, 1.92Ghz/960Mhz plus a bandwidth increase, is possible though IMO 960Mhz would be a little too hot.

Combine those all into a big performance release and I'm sure they could throw some pretty gaudy numbers up.

Or they could just switch to DX12 and take that fabled 2x increase....................
 
For an up clock to work, you have to guarantee that all of the PS4s that are on the market will continue to work and can handle it. If that's the case, why not just ship at the higher clock.
 
If there's even a small percentage of consoles sold where the upclock will not work (ASIC just can't handle it) then the whole idea becomes infeasible. You can't screw over customers by having hardware updates deliverd as software that won't install on a subset of all systems (it makes the people left behind feel angered and justifiably so), and stratifying your user-base into "turbo-PS4s" and non-turbo goes against the whole console idea of a unified platform.

Considering Sony has been struggling to match demand I think we can safely assume they've built consoles around every single working ASIC they could have gotten their hands on, even those that will not clock very far above default.

Besides - no major performance gains could be derived from an upclock anyway, the improvement would be so fractional as to be irrelevant, if not completely imperceptible in the big scheme of things. Certainly by users who aren't hardware nerds...
 
Not really, you just need to be willing to replace those that can't and considering how underclocked the PS4 is, that may be an acceptable risk. Why not from the start? The obvious is that they didn't have to. And in waiting their data set on reliability went from a few hundred machines to over 7 million (plus they benefited from observing several million XBO running higher). They also benefit from finding out if devs were going to actually use the extra power in the PS4 or just go the parity route. Then there's risk savings chipwise by being able to run thru a significant number of chips at lower clocks, it's not a stretch to think there was real cost savings there. Finally there's the PR chess game with MS, so after MS scrambles to get whatever increases they can Sony turns around and trumps it with their own increases.
 
I think knowingly pushing out an update that will break some consoles, even if you're willing to replace them, probably has certain legal implications and I bet would open Sony up to a class action lawsuit. I doubt they want to mess with that.
 
Anyway ICE only deals with GNM API, right?

Not hardware related stuff. So it can only be software API optimizations like the 10-100x gain on tiling stuff seen previously.
 
Software optimisations should be applicable to XB1 too though, hence the infeasibility of the claim. Where do you get a software upgrade that XB1 can't match? Maybe they are taking low level to a new, lower level, below the OS such that MS's VMs can't go there, that does Magical Things exploiting clock timings and register overflows?
 
Software optimisations should be applicable to XB1 too though, hence the infeasibility of the claim. Where do you get a software upgrade that XB1 can't match? Maybe they are taking low level to a new, lower level, below the OS such that MS's VMs can't go there, that does Magical Things exploiting clock timings and register overflows?

It could be some software that exploits the hardware better via a change that was made by Sony.
 
I think knowingly pushing out an update that will break some consoles, even if you're willing to replace them, probably has certain legal implications and I bet would open Sony up to a class action lawsuit. I doubt they want to mess with that.

How do you know it will break consoles? Is it breaking XBO consoles? For that matter, doesn't any update have the potential of breaking consoles?

I have no idea if Sony is even considering an upclock but I wouldn't be too quick to rule one out.
 
Well if it's a new binary assembly code only available with the licensed/NDAed GNM, I am not sure they could even easily replicate it on the XB1, even if they legally could.
 
How do you know it will break consoles? Is it breaking XBO consoles?
MS decided on the upclock prior to release, and tested the machine for the upclock. Conceptually I can see a CPU+GPU upclock as possible as all Sony ought to do to handle that is increase the fan speed. The problem is how close to the limit these components are currently clocked, and if some will plain break/fail at a small clock increase. They might need a higher voltage to be stable, for example, and that might not be a feature of the motherboard if it wasn't designed for variable clockrates. If a handful of boxes fail, Sony could probably do it and provide a free replacement for those who's consoles get bricked by the update. If it's a larger number, it just wouldn't be worth thre cost. TBH it probably wouldn't be worth the cost anyway, unless it's a significant upclock, not so much due to competition where Sony are out in front, but maybe to help with Morpheus.

Still incredibly unlikely though. If it happens, it must have been designed in as an option from the start. Discussions with AMD about clocks and target maybe settled on an 1.6 GHz/800 MHz option with a switch for 1.8 GHz/900 MHz, or maybe just a CPU upclock. Rumours once were that PS4's Jag cores ran at 2 GHz.
 
How do you know it will break consoles? Is it breaking XBO consoles? For that matter, doesn't any update have the potential of breaking consoles?

I have no idea if Sony is even considering an upclock but I wouldn't be too quick to rule one out.

Not really, you just need to be willing to replace those that can't and considering how underclocked the PS4 is, that may be an acceptable risk.


You're the one who suggested that they could replace consoles that won't work with the update? Why would they replace working consoles?

There's a difference between an unknown potential to break a consoles with an update and one that that's guaranteed to break X percentage of consoles (assuming they're not all guaranteed to work at a potential up clock).
 
Unlocking 'turbo' modes would make kinda sense. With the system reservation you always have some idle'ish cores, shuffling the reserved cores to the hottest cores/modules would help with keeping the turbo potential high (and trashing the cache?).
 
Status
Not open for further replies.
Back
Top