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

[Scott_Arm]

I have low expectations of insiders as well, but where are the legit gaming journalists doing the actual sleuthing work on next gen? Are they complicit in trying to prevent leaks for Sony?
...

I don't know if any legit journalists would actually try to expose a companies trade secrets. At best they analyze leaks, because leaks are already in the public. Tech journalism has always walked a fine line.

 

[AbsoluteBeginner]

This is an interesting quote from Jason Schreier after his > Stadia comment.

"Anyway, I don't know for sure what next-gen specs will look like. I have no idea how many teraflops there'll be. A dev recently told me that a lot of people within studios are just making assumptions based on limited information, and that one of the assumptions is that everything's going to be twice as powerful. "

That tells me they dont really know all that much and, as in case of PS4Pro and XBX, devs might not have actual specs 1+ year in advance.

 

[N2O]

Well if PS5 early devkits comes with Radeon VII i won't be surprised, the Radeon VII is 13 something TF and almost same power as 5700XT.

 

[Proelite]

This is an interesting quote from Jason Schreier after his > Stadia comment.



That tells me they dont really know all that much and, as in case of PS4Pro and XBX, devs might not have actual specs 1+ year in advance.

Earliest Xsx target specs:

4x CPU
1-1.5x RAM
1.5-2x GPU
40x SSD

 

[anexanhume]

Well if PS5 early devkits comes with Radeon VII i won't be surprised, the Radeon VII is 13 something TF and almost same power as 5700XT.

We have no evidence a dev kit ever existed with Vega.

 

[PSman1700]

We have no evidence a dev kit ever existed with Vega.

We have no evidence that it didn't, either. Most likely it did, as development would be very short if all had to wait till AMD had Navi ready for devkits.

 

[AbsoluteBeginner]

If Sony had dev kits out in late 2018, in big silver towers, chances are it was Vega inside. Funny there was mention of Vega20 in Oberon data but I dont think it proves anything...

 

[anexanhume]

We have no evidence that it didn't, either. Most likely it did, as development would be very short if all had to wait till AMD had Navi ready for devkits.

Proving a negative is generally a hopeless task. It’s usually incumbent upon someone trying to make a claim to substantiate it.

If Sony had dev kits out in late 2018, in big silver towers, chances are it was Vega inside. Funny there was mention of Vega20 in Oberon data but I dont think it proves anything...

Developmental Navi silicon was out at that time as well. Given that Ariel had shown up in benchmarks in January, it’s logical to presume it could exist a few months prior to that benchmark.

 

[BRiT]

Proving a negative is generally a hopeless task.

Is there any proof of that?

<benny hill>

 

[PSman1700]

Proving a negative is generally a hopeless task. It’s usually incumbent upon someone trying to make a claim to substantiate it.

So you think Navi has been there before 2018 in devkits, testlabs etc?

 

[Proelite]

If Sony had dev kits out in late 2018, in big silver towers, chances are it was Vega inside. Funny there was mention of Vega20 in Oberon data but I dont think it proves anything...

Sony's earliest devkits could have had in 2018:

Vega 64.
Preview Vega 7.
Navi 10 early samples.

I was tipped to Vega 64 for early 2018 kits. We've seen evidence of Navi 10 used in another Github repository for Ariel execution benchmarks.

 

[N2O]

Now you talked about Vega 64, i just find out it's 12.6TF.

 

[PSman1700]

Now you talked about Vega 64, i just find out it's 12.6TF.

And Vega64 performs equal to the 5700/5700XT which comes in at around 8/9TF, throw in RDNA2 optimizations, higher clocks, and were close to 9.2TF.
Vega64 probably was the closest they could get at the time.

 

[AbsoluteBeginner]

And Vega64 performs equal to the 5700/5700XT which comes in at around 8/9TF, throw in RDNA2 optimizations, higher clocks, and were close to 9.2TF.
Vega64 probably was the closest they could get at the time.

Vega64 is nowhere near XT. In fact vanilla 5700 outperforms Vega by about ~10%, let alone XT.

 

[PSman1700]

Vega64 is nowhere near XT. In fact vanilla 5700 outperforms Vega by about ~10%, let alone XT.

Vega64 was probably the fastest they could get their hands on at the time. It would never match the final navi 9.2TF though. Much has happened over at AMD since Vega.

 

[N2O]

I also remember during 2019 E3 a journalist "heard" PS5 is better than XSX, this could be due to PS5 using Vegaf 64 or Radeon VII which is 12.6/13.4TF, in math it's "better" than XSX's target.

 

[Proelite]

I also remember during 2019 E3 a journalist "heard" PS5 is better than XSX, this could be due to PS5 using Vegaf 64 or Radeon VII which is 12.6/13.4TF, in math it's "better" than XSX's target.

12.6-14 > (2x XB1X === 12).

https://forum.beyond3d.com/posts/2095179/

 

[dobwal]

Outside of Vega, what choice did AMD have before Navi? The only AMD based gpus that come close to the proposed specs in terms of TFs and memory bandwidth is Vega and Fury. Both sport ~500 GB/s worth of bandwidth but Fury is limited in terms of TF.

 

[BRiT]

Another possibility is the Devkits used Nvidia RTX ... Slim and None. But that's the other possibility, assuming you entirely ignore driver issues and vendor extensions etc.

 

[Proelite]

Another possibility is the Devkits used Nvidia RTX ... Slim and None. But that's the other possibility, assuming you entirely ignore driver issues and vendor extensions etc.

I can seee this from MS, but not Sony.