Playstation 5 [PS5] [Release November 12 2020]

[Kaotik]

Well Cerny said it was more or as powerful as the 8 jaguars core from ps4,so he is not wrong here ?

Yeah and no, I just rewatched the part so I wouldn't trust my short but bad memory. Cerny said specificly that it's more powerful than 8 Jaguar cores combined when measuring SIMD power & bandwidth.
So yes, it's right for that specific task, for which (to my understanding) Jaguars were pretty bad to begin with, but that in my books is still quite far from saying the "tempest engine is more powerful than ps4's brains"

 

[Rootax]

Yeah and no, I just rewatched the part so I wouldn't trust my short but bad memory. Cerny said specificly that it's more powerful than 8 Jaguar cores combined when measuring SIMD power & bandwidth.
So yes, it's right for that specific task, for which (to my understanding) Jaguars were pretty bad to begin with, but that in my books is still quite far from saying the "tempest engine is more powerful than ps4's brains"

I understand your point, but I believe at this point is kind of semantic ? And the guy on twitter was an audio guy, so I get his tweet too.

My huge letdown about that is they didn't call it the CELL engine

 

[Karamazov]

no matter how powerful it is, it's freeing the CPU from those audio tasks, juste like XseX has a dedicaced audio chip, and that's good for us.

 

[psorcerer]

PS4's southbridge was connected with PCIe, and that's the start of the process for its jailbreak.
https://fail0verflow.com/media/33c3-slides/#/11

Certified harddrives only.

 

[Betanumerical]

Certified harddrives only.

I honestly dont think that will stop the problem. It will certainly make it harder (and honestly a lot of security is about raising the bar for attackers as well as eliminating actual problems), but the level of attacker that is willing to spend the time to do PCIe based DMA attacks is also the kind who can probably setup a shim to pass on all the required requests and then inject their own afterwards.

Id love to know the details of how this will work but I know that won't happen.

 

[Karmaprof]

I’m not sure. It should have both. Geometry engine I assume is another label for Mesh shader. And VRS is VRS. You can do both to save cycles.

From what i understand MS has its own VRS solution a Patent and not using the Standard Version of VRS from the Desktop Navi Variant.

 

[iroboto]

From what i understand MS has its own VRS solution a Patent and not using the Standard Version of VRS from the Desktop Navi Variant.

Sure, but he’s saying VRS without GE isn’t as good as GE. But all GPUs that support VRS support GE. On AMD and Turing. Mesh Shaders and VRS are core features of DX12U

 

[Dictator]

He did talk about mesh shaders in all but name when he referenced the Geometry Engine and its features. We’re all wondering why he didn’t mention VRS. My guess is the interest of time. His talk wasn’t an exhaustive coverage of the GPU capabilities.



The anti-tamper measures in this generation are fairly extensive:

https://www.crn.com/news/components...on-work-led-to-a-big-security-feature-in-epyc

I am still wondering why Sony said RDNA 2.0 based while MS said RDNA 2.0.
Semantics on a Level, but MS went out of their way to delineate features that are Part of RDNA 2.0 and ones they wanted specifically on the side for the XSX, like the int 8 and int 4.
Is the geometry engine actually 100% the same as the mesh shader? Is there an amplification shader/task shader there?
It would be nice if these questions were headed off by a much more rigorous detailing of the GPU Feature Set, something which I would have liked for a GDC talk.

 

[Karmaprof]

Sure, but he’s saying VRS without GE isn’t as good as GE. But all GPUs that support VRS support GE. On AMD and Turing. Mesh Shaders and VRS are core features of DX12U

I would be Interesting if Nextgen Consoles GPUs support such thing like Tilebase Rendering and Deffered Rendering like PowerVR2.

 

[Rootax]

I love powervr, but my guess is their are good reasons if nvidia and amd haven't done a TBDR gpu yet. Maybe it's technical, or IMG Tech has too many patents for this ? But it's a subject for another topic.

 

[anexanhume]

I am still wondering why Sony said RDNA 2.0 based while MS said RDNA 2.0.
Semantics on a Level, but MS went out of their way to delineate features that are Part of RDNA 2.0 and ones they wanted specifically on the side for the XSX, like the int 8 and int 4.
Is the geometry engine actually 100% the same as the mesh shader? Is there an amplification shader/task shader there?
It would be nice if these questions were headed off by a much more rigorous detailing of the GPU Feature Set, something which I would have liked for a GDC talk.

It makes sense in the context of DirectX 12 Ultimate. MS has a bigger imperative to advertise and enumerate these features as they try to unify and move platforms forward. MS has just been more communicative in general outside of that campaign.

 

[Silent_Buddha]

That’s all valid, but it’s beside the point of whether you should believe a random common from a “developer” on the internet. All those fake spec leaks. Those were “developers” too ;-)

Sure, but people are trying to read more into what Cerny has said than he actually said. Now, what people are speculating him as meaning might be how it is, but it also might not be how it is.

For a GDC talk it was somewhat disconcertingly vague on many points with more of a focus on talking to general consumers versus developers than what I usually expect from a GDC talk. It's always interesting to hear him talk, but I suspect that if the talk had been given at GDC, it would have been more technical than how it turned out. C'est la vie.

Regardless, while I don't trust random posts in comment threads as much as reputable journalists I also don't just immediately disregard what they have to say. Just like I didn't trust what "insiders" were saying about XBSX and PS5 before the respective hardware details came out, but didn't immediately disregard what they were saying.

Take the NX Gamer video, for example, it's about as speculatively right or wrong as random comments on a thread about what Cerny has said. But people will selectively take that as more or less truthful depending on what they want to believe. Like how some people will believe what they saw in the NX Gamer video more than what people from DF have been told by developers working on the PS5. On one hand we have what we know is speculation (which could be right or wrong) and on the other hand we have something that is coming from developers that we know are working on the hardware.

Regards,
SB

 

[Xbat]

Take the NX Gamer video, for example, it's about as speculatively right or wrong as random comments on a thread about what Cerny has said. But people will selectively take that as more or less truthful depending on what they want to believe. Like how some people will believe what they saw in the NX Gamer video more than what people from DF have been told by developers working on the PS5.

https://twitter.com/x/status/1241454613993152512

Isn't he a dev that would of worked on PS5 ?

 

[MrFox]

I'm still puzzled about what lossy format the PS5 supports. He only talked about Kraken which seems to be replacing the old LZ block on the Amiga Blitter, but they didn't mention what lossy image format they will use. For streaming assets this is a pretty important part.

 

[Scott_Arm]

Well Cerny said it was more or as powerful as the 8 jaguars core from ps4,so he is not wrong here ?

That is a very odd comparison to make. PS4 had DSPs for audio, so I’m not even sure what he’s comparing.

 

[Karamazov]

PS5 SSD fast data loading of worlds tech demo:

 

[3dilettante]

If it is like in the patent tamper checking it is working like this.

https://www.resetera.com/threads/ps...nd-sonys-ssd-customisations-technical.118587/

That method should make it much more difficult to get at the data that is encrypted on the drive, since the data over the link would be compressed and encrypted, and there wouldn't be a good way to manipulate the data stream with the checking occuring after the payload has gone on-die and then run through the full decode process. If the PS5 were to use whole memory encryption, that end might be even more secured, though maybe Sony would rely on security measures around the DMA controllers.

Certified harddrives only.

The PS4's jailbreak involved inserting an intermediate device between Sony's own south bridge and SOC. A third party drive isn't a lower-risk option, unless the assumption is that the third party is less likely to have naive firmware or PCIe interface policies than Sony. However, in that case there's the Sony-owned custom SSD controller and APU.

I'm still puzzled about what lossy format the PS5 supports. He only talked about Kraken which seems to be replacing the old LZ block on the Amiga Blitter, but they didn't mention what lossy image format they will use. For streaming assets this is a pretty important part.

Perhaps Sony is relying on the existing GPU-supported texture formats as the lossy portion of the compression? It would likely lag in effectiveness over a new standard that can apply more sophisticated techniques over broader data sets, but there would be space savings over a lossless algorithm. It wouldn't be a new feature, so it could have been taken as a given.

 

[psorcerer]

However, in that case there's the Sony-owned custom SSD controller and APU.

I have no idea how they will integrate external drives with their controller.
I don't see any possibility of achieving the same latency/bandwidth.
Other than maybe that external controller gets disabled and the chips are passed through as-is to the Sony controller.

 

[Globalisateur]

I am still wondering why Sony said RDNA 2.0 based while MS said RDNA 2.0.
Semantics on a Level, but MS went out of their way to delineate features that are Part of RDNA 2.0 and ones they wanted specifically on the side for the XSX, like the int 8 and int 4.
Is the geometry engine actually 100% the same as the mesh shader? Is there an amplification shader/task shader there?
It would be nice if these questions were headed off by a much more rigorous detailing of the GPU Feature Set, something which I would have liked for a GDC talk.

Where did they say (just) that ? On the official channels Microsoft also wraps RDNA2 with specific terms:

RDNA 2-class GPU.

leveraging AMD’s latest Zen 2 and RDNA 2 architectures

Custom RDNA 2 GPU

https://news.xbox.com/en-us/2020/03/16/xbox-series-x-tech/
https://news.xbox.com/en-us/2020/03/16/xbox-series-x-tech/

 

[zupallinere]

PS5 SSD fast data loading of worlds tech demo:

Another optimistic take or maybe more generously a reasonable argument based what certainly could be going on. This part was interesting ...

In particular the idea that PS5 SSD would act as a kind of flexible and fast virtual memory to make it look like "100 GB" of memory reminded me of musings years ago on the subject. If you had say really fast ram with an abundance of non-volatile fast memory, say 48 GB of Crossbar memory somewho melded together and fed by a cheap SSD that you could do some interesting things.
Looking back on my thoughts on the subject I even remembered the Crossbar thing. CRS my butt.
https://forum.beyond3d.com/threads/...-edition-2014-2017.56353/page-86#post-1988653
https://forum.beyond3d.com/threads/...-edition-2014-2017.56353/page-88#post-1989284
Now there is no 3 level hierarchy just memory and fast Non-Volatile memory which makes more sense.
I would add that the praise heaped upon the PS5 might come from quotes before say the whole memory bandwidth things was resolved if such scuttlebut is true.(14 or 16gbs)
Still if the quote from the supplier was accurate it might mean that the PS5 SSD might not be as expensive as one would imagine. Based on the excerpts flashed on the screen by the Youtube guy cheap chips flash chips attached to a big parallel fast cache might mean 399 or 449 isn't out and out crazy.