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

[Rootax]

If this is true (and I don't believe it is), it look like a "phone release". Like OnePlus and others do, OnePlus 7 vs 7Pro, for exemple. The second one is a little stronger, but by a slim margin (except ram/storage size).
Maybe it make financial sense, if some of the SoC are "too good" for the ps5, why not create a PS5+ or whatever and sold them with a higher price, instead of putting them in regular ps5 ?

 

[Rangers]

If this is true (and I don't believe it is), it look like a "phone release". Like OnePlus and others do, OnePlus 7 vs 7Pro, for exemple. The second one is a little stronger, but by a slim margin (except ram/storage size).
Maybe it make financial sense, if some of the SoC are "too good" for the ps5, why not create a PS5+ or whatever and sold them with a higher price, instead of putting them in regular ps5 ?

Those soc wouldn't be "too good" enough to make a meaningful differentiation. They'd be a few more CU's enabled and maybe a little higher clocks. Like 10-20% difference. Nothing enough to call it a Pro or that it would be noticable.

 

[Rootax]

Those soc wouldn't be "too good" enough to make a meaningful differentiation. They'd be a few more CU's enabled and maybe a little higher clocks. Like 10-20% difference. Nothing enough to call it a Pro or that it would be noticable.

Oh I agree. But if they can sell a 10% faster console, and why not with a 1.5tb ssd instead of 1.0, and made more margin with that than PS5...
Again, I'm not saying they will do it, but I can see why If they do something like that.

 

[Shifty Geezer]

So what? Honestly, who cares?

Those trying to have a more interesting conversation than, "here's a thing > disproven by Github, here's a thing > disproven by Github, here's a thing > disproven by Github, here's a thing > disproven by Github."

A couple of changes have been brought about to this thread. Please everyone drop the subject of the 'GitHub gospel' and just discuss rumours as they come, and we'll see if good conversation can be restored.

 

[iroboto]

Sounds to me guys that are bringing constant rumors from GAF and pastebin are also the ones referrencing Github "gospel" the most. I am mostly checking this thread without posting because some posts are simply tiring.

Most of them are bad though. We should be able to dispute the validity without needing Github as a crutch.

One of them mentioned RDNA 1.9 after Phil Spencer made the official statement on RDNA 2 for instance. That’s a huge red flag. The amount of information a person would need to determine the difference between 3 unreleased products (ps5, xsx, the next AMD card) to make that claim would require extensive in-depth knowledge of all three at a micro level and they would have to know how and which features they are back porting. It’s insanity why people would make such audacious claims without any method to prove these things; we would have a challenge proving it even post release.

his greatest defence is that MS is lying.

 

[Shifty Geezer]

his greatest defence is that MS is lying.

Or rounded up.

 

[upnorthsox]

his greatest defence is that MS is lying.

Or exaggerated by say .1

 

[iroboto]

Or exaggerated by say .1

Who would this work on really?
He’s the only person that came up with 1.9, he’s made up his own decimal system for an architecture. That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

who would be fooled by this trolling.

 

[goonergaz]

Who would this work on really?
He’s the only person that came up with 1.9, he’s made up his own decimal system for an architecture. That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

who would be fooled by this trolling.

me, lol - we’ll not fooled but I can see what he’s trying to say.

it’s like when people said it’s v1 with some features from V2, so people started calling it 1.5. I read this as it’s not quite got all the v2 features- it’s very close but a little removed. I suppose like some have suggested the CPU is a little cutback on the cache ( IIRC).

of course I’m not as clever as you guys though

 

[upnorthsox]

Who would this work on really?
He’s the only person that came up with 1.9, he’s made up his own decimal system for an architecture. That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

who would be fooled by this trolling.

Are you serious? You do realize that RDNA2 is not even a thing. For all it matters, Sony can claim the PS5 is RDNA3. Not sure where that would leave AMD but if it mattered to them they would've declared it by now.

 

[BRiT]

Are you serious? You do realize that RDNA2 is not even a thing.

Really?



AMD will “talk more” about RDNA 2, its next-generation GPU architecture, during its upcoming Financial Analyst Day. Building on the RDNA architecture introduced with AMD’s Navi graphics cards over the past seven months, the chipmaker is promising to spill the beans on its 7nm, ray-traced successor on March 5, 2020.

https://www.pcgamesn.com/amd/rdna-2-reveal

 

[fehu]

The rumor is that PS5 will have a very cheap single model, but that can be upgraded at launch.
This explains the confusion about different performance level, and marketing strategies.
The single model will have a gpu at 7TF, but it can be stacked on top of one other, and linked together with a proprietary hight speed / low latency / coherent port, doubling to 14TF.
It must be noted that in this configuration it will be able to access 16x2 virtual thread, and 12GBx2 of GDDR6.
It's expected a middlegen refresh that will add a third stack on top.

 

[Picao84]

The rumor is that PS5 will have a very cheap single model, but that can be upgraded at launch.
This explains the confusion about different performance level, and marketing strategies.
The single model will have a gpu at 7TF, but it can be stacked on top of one other, and linked together with a proprietary hight speed / low latency / coherent port, doubling to 14TF.
It must be noted that in this configuration it will be able to access 16x2 virtual thread, and 12GBx2 of GDDR6.
It's expected a middlegen refresh that will add a third stack on top.

The early 90s Sega called saying it wants it's ideas back!

 

[Kaotik]

Who would this work on really?
He’s the only person that came up with 1.9, he’s made up his own decimal system for an architecture. That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

who would be fooled by this trolling.

Actually we do have those too, GCN1.0-1.4.1 are GCN (Vega 20 aka Vega with DLops is 1.4.1) and Navi is GCN1.5 (and with DLops 1.5.1) even though it's now called RDNA.
AMD was never consistent with this, with drivers etc saying one thing, marketing other and so on

 

[iroboto]

Actually we do have those too, GCN1.0-1.4.1 are GCN (Vega 20 aka Vega with DLops is 1.4.1) and Navi is GCN1.5 (and with DLops 1.5.1) even though it's now called RDNA.
AMD was never consistent with this, with drivers etc saying one thing, marketing other and so on

Sure but that isn’t how we announce things publicly in marketing statements. Nor do I expect under any reasonable condition that odium can produce driver statements indicating that both are 1.9.
Be reasonable here; the probability extremely low

 

[fehu]

The early 90s Sega called saying it wants it's ideas back!

Yeah, but it's a jolly solution that covers:
- being slower than xsx
- being faster than xsx
- having a monolithic soc
- having multiple chiplet
- having a single model at launch
- having multiple models at launch
- having a refresh down the road
- not having any refresh

Anyone is right and we are all happy

 

[upnorthsox]

Really?

View attachment 3607

AMD will “talk more” about RDNA 2, its next-generation GPU architecture, during its upcoming Financial Analyst Day. Building on the RDNA architecture introduced with AMD’s Navi graphics cards over the past seven months, the chipmaker is promising to spill the beans on its 7nm, ray-traced successor on March 5, 2020.

https://www.pcgamesn.com/amd/rdna-2-reveal

Ok then, I'm wrong.

Thanks

 

[Kaotik]

Sure but that isn’t how we announce things publicly in marketing statements. Nor do I expect under any reasonable condition that odium can produce driver statements indicating that both are 1.9.
Be reasonable here; the probability extremely low

Of course, was just pointing that AMD does use that kind of version numbering too, but obviously RDNA1.9 isn't part of any such numbering scheme

 

[Scott_Arm]

@BRiT AMD planned for RDNA2 to come out this year, and Microsoft announced an RDNA2 based product this year, so we all know what the most obvious conclusion is ... Microsoft is clearly LYING.

 

[Deleted member 13524]

One of them mentioned RDNA 1.9 after Phil Spencer made the official statement on RDNA 2 for instance. That’s a huge red flag.

It's definitely not one. I've seen more than one leaker / leaker claiming that the new console(s) have a featureset above RDNA1 but below RDNA2.

Who would this work on really?
He’s the only person that came up with 1.9, he’s made up his own decimal system for an architecture. That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

Again, you're trying to pull apart all the tiny details in his language like he had carefully chosen his words as if he had written a corporate press release... when in reality he actually didn't put that much attention to it.
He didn't write "the PS5 is using what AMD calls RDNA 1.9". Using the 1.9 number simply means it's almost RDNA2 but not quite. Don't over-analyse it..
I believe in a previous post someone had said something along the lines of the new consoles getting new instructions above RDNA1, but not to the point of reaching AMD's RDNA2. This could be because the iGPU is older than e.g. Navi 2x (e.g. development had to start and finish before Navi 20 was finalized), and the result is either a less advanced architecture or just an architecture built over RDNA1 that gained new customer-specific instructions to cover some additional needs (say, raytracing and ML instructions), but they're not within RDNA2's set (assuming RDNA2 is a superset of RDNA1).

Who would this work on really?That’s like saying we had GCN 1.9 instead of GCN 2. Or GCN 3.3 instead of GCN 3 or 4. He would have been laughed straight out of the building.

Adding to what @Kaotik wrote, you even have different instruction sets under the same GCN moniker. Vega 20's ISA is a superset of Vega 10's that adds 4xINT8 instructions (among other) for ML, and they're both Vega / GFX9 / GCN5. I'd colloquially call it GCN5.2 for example.
Then AMD has other even weirder cases, like the Vega M on Kaby Lake G. They call it a Vega but it's actually a GCN4 / GFX8 chip with some functionality of the succeeding architecture like HBCC.
In this case, I think it would be fair to call it GCN4.5, or GCN4.7 or whatever number I felt like calling it at the time. You may disagree on how I'd phrase it, but I would use that term in a forum and so would many others IMO.