Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

[BRiT]

That wouldn't stop disc-compatible BC though. Put in the disc - you own the rights to all media on it. XBO supports disc-based DC but only for a handful of select titles.

Because it has to download an entirely new program, hence the license terms about redistribution. It uses the disc only as an ownership key check.

As an exercise, uninstall the BC titles and unplug your Xbox One from the internet. Then pop in that game disc that is BC and see what happens.

@Shifty Geezer

In the future, maybe with the more powerful CPUs, disc-based BC could be a real thing, where the repacking and ahead of time recompilation happens at the console. It wasn't done on Xbox One because those tiny Jaguar cores would take hours or even days to do this task. That's definitely not user friendly.

 

[mpg1]

I mean if Lockhart was extremely close in gpu power to X1X would you need separate BC mode for that? Maybe Sparkman is Lockhart SoC and runs X1X and next gen games in the same mode.

 

[AbsoluteBeginner]

Mental gymnastics people go about now is just...amazing. Reminds me of Durango leak lol

"No way MS goes for 1.2TF, thats 50% less then PS4. Not in a million years"
"Half the ROPs? Yea sure lol!"

If the leak is legit (and it almost certainly is), Oberon is ~15-20% weaker then Arden. It will almost certainly break Sony's TDP "record", so I am very surprised of people saying they are skimping this gen. Its VERY fast, smart chip. I made post about it last month :

I will love it when PS5 is 316mm² with 40CU clocked to hell and back coupled with 256bit bus and fastest chips around for ~530-576GB/s.

Such a smart design that I refuse to believe they are not going with it lol

They could SERIOUSLY undercut MS with that one. Would be PS4Pro vs Xbox One X (only difference in power would be ~15% and not 30%).

https://forum.beyond3d.com/threads/...rits-post-e3-2019.61246/page-122#post-2092393

Taking following assumptions in mind :

1) Sony's BC issue (CU count and clock speeds being very staple of it)
2) Console form factor
3) Single SKU strat

There is very little chance we are getting anything stronger then above. In fact, its stronger then anything Sony pushed at console release relative to TDP limits.

This will be ~200-220W console even on 7P node.

 

[BRiT]

I've been saying for the past couple weeks that I should attempt to format up a more technical thread about "NextGeneration console approaches to Backwards Compatibility". If nothing else it would be a reference point for information linking to current-gen approach on midgen upgrades and what we know of ISA from GCN3 through GCN5 and then transition to RDNA.

However, the current rumors is engaged at crazy-train level speeds, and it's a bit of fun to watch.

 

[Scott_Arm]

Why the assumption that XseriesX has the same number of rops? Looks like RDNA has 4 RBs per shader array. 40 CUs aligns to 4 shader arrays and 60 CUs aligns to 6. It could be a different configuration, but I’m just wondering if there are any numbers to suggest that it is.

edit: confusing codenames

 

[BRiT]

Why the assumption that Oberon has the same number of rops? Looks like RDNA has 4 RBs per shader array. 40 CUs aligns to 4 shader arrays and 60 CUs aligns to 6. Oberon could be a different configuration, but I’m just wondering if there are any numbers to suggest that it is.

Yeah, the other day when @TheAlSpark and I were digging through tech publications, the new RDNA naming convention has really made things extra confusing to me on just how closely do things align between what we knew on GCN3 (current gen) and how things may be layed out on NextGen, at least from a backwards compatibility and timings aspects.

 

[Scott_Arm]

Yeah, the other day when @TheAlSpark and I were digging through tech publications, the new RDNA naming convention has really made things extra confusing to me on just how closely do things align between what we knew on GCN3 (current gen) and how things may be layed out on NextGen, at least from a backwards compatibility and timings aspects.

The 5700xt has 4 shader arrays each with 10 CUs and 4 RBs. So Xbox is either increasing the number of CUs per shader array or increasing the number of shader arrays. Or maybe ps5 has more than 4 rbs per shader array and that’s why it’s rumoured to have a higher fill rate. Just wondering what the source of the rumour is. This whole thread feels like a side conversation to resetera and Twitter. Pretty much impossible to follow because the assumption is I’m reading that stuff instead of just putting links in this thread. Still have no idea where the source of the benchmarks is.

 

[AbsoluteBeginner]

The 5700xt has 4 shader arrays each with 10 CUs and 4 RBs. So Xbox is either increasing the number of CUs per shader array or increasing the number of shader arrays. Or maybe ps5 has more than 4 rbs per shader array and that’s why it’s rumoured to have a higher fill rate. Just wondering what the source of the rumour is. This whole thread feels like a side conversation to resetera and Twitter. Pretty much impossible to follow because the assumption is I’m reading that stuff instead of just putting links in this thread. Still have no idea where the source of the benchmarks is.

Its very easy to follow actually. Documents are now deleted, I don't have them but there was number of SPs for Arden as well as pixel fillrate (think Proelite found it) so it was easy to know how many ROPs.

 

[Scott_Arm]

Its very easy to follow actually. Documents are now deleted, I don't have them but there was number of SPs for Arden as well as pixel fillrate (think Proelite found it) so it was easy to know how many ROPs.

It’s easy to follow if you’ve seen the documents, which I haven’t. Yes, you can reverse engineer the numbers so if they’re correct ps5 will have a fill rate advantage.

 

[Deleted member 11852]

Still have no idea where the source of the benchmarks is.

It was on Github until this morning when, presumably AMD, had it pulled.

 

[Scott_Arm]

How do the fillrate of ps5, series x compare to 4pro and 1x?

 

[mrcorbo]

Yup. Sony's choices for PS4 have made certainly complicated their backwards/forwards strategy that I doubt they thought overly about in 2010-2013.

Even if they did, they may have decided the benefits in the present and near future outweighed the costs they would incur down the line. I think it's hard to fault their reasoning, in that case.

 

[BRiT]

Or...

Most elaborate hoax ever.

 

[RDGoodla]

1. Both PS5 and xsx should use 7nm+ ( we know Arden is 350mm2 with 56CUs).

2. OBR B0 is an old chip because Taiwan leaker said OBR is still continuously changing version and re-tested in November.

Latest OBR may have at least 44~48 CUs in 7nm+
with 300mm2 size.

3. PS5 GPU with 48CUs at 2GHz will have 12.288 TFlops with higher fillrate than xsx. This result will be in line with Kleegamefan’s comment “more flops” and 10% more game performance.

 

[Shifty Geezer]

In the future, maybe with the more powerful CPUs, disc-based BC could be a real thing, where the repacking and ahead of time recompilation happens at the console. It wasn't done on Xbox One because those tiny Jaguar cores would take hours or even days to do this task.

Well precisely. The original Xbox was not sufficiently abstracted such that when presented with a different x86 CPU and a DirectX GPU using DX API calls, the same code could run. MS hasn't done that with their consoles until XBO, where they realised what was needed to ensure future BC was a higher level of HW abstraction. Prior to that, they just put out hardware with an API, same as everyone else.

 

[Deleted member 11852]

Even if they did, they may have decided the benefits in the present and near future outweighed the costs they would incur down the line. I think it's hard to fault their reasoning, in that case.

Could be but I don't even think it was a conscious decision, I just think it didn't occur to them that down the line they want to run PS4 games on different hardware or that it would be so complicated.

 

[AbsoluteBeginner]

I'm struggling listening to "Klee said"

Klee said people will be disappointed with specs back in August. Are people people disappointed by 12-13TF Navi? No, of course not. If that were the case then people would be estatic.

He hasnt said anything of note, just please can we please referrence Klee here? There was nothing technical from here posted...EVER.

 

[AbsoluteBeginner]

How do the fillrate of ps5, series x compare to 4pro and 1x?

PS4Pro - 911 x 64 - 58 GP/s
XBX - 1172 x 32 - 37.5 GP/s
Oberon (rumored) - 2000 x 64 - 128 GP/s
XSX (rumored) - 1700 x 64 - 108.8 GP/s

 

[mpg1]

The only thing that doesn't fit re Sparkman being Lockhart is the 4Tflops rumour. But that always seemed extra low to me.

56CU's@1GHz = 7.16Tflops
56CU's@900MHz = 6.45Tflops
56CU's@850MHz = 6.09Tflops

Again the whole Lockhart/extra sku thing is still not clear as Micrososft hasn't announced anything. There's even been rumors that is was killed.

 

[Deleted member 11852]

The original Xbox was not sufficiently abstracted such that when presented with a different x86 CPU and a DirectX GPU using DX API calls, the same code could run.

That was never the goal of any of the Xbox until, arguably Xbox One. You seem to be equating abstraction with compatibility but abstraction is designed to solve a different set of problems. As for abtracting code to run on different processors, nobody does this. Apple's iOS platform comes closest but this relies to recompilation of code as it's downloaded and/or updated by the App Store.

The difference in execution is tangible. Microsoft used attraction APIs and found an affordable way to run previous generation Xbox games on latter hardware which not only ran better, but looked better. Sony have used close-to-the-metal APIs which which limited their hardware choices and when they did deliver a limited form of B/C, the 'improvements' were limited to screen filters.