By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Dark_Feanor said:

So, our friends from Semaccurate wanted to get back into the spotlight. And are using unsupported claims they were the first to leak XOne and Switch specs.

https://semiaccurate.com/2018/04/03/semiacccurate-gets-playstation-5next-details/

The PS5 might exists and might be launch this Hollyday season. However, there is zero chance of it´s supporting a Navi GPU and may be close to zero to use Zen CPU cores.

First, there has never been a console with a Radeon GPU architecture that was the top tear of the year. And we even have to suppose those Navi cards are coming this year. 

Second, there is no new mass market node production below 14nm that PS4-Pro and XoX use. The Scorpio chip is the biggest one could get at a $500 machine. And Zen cores are far bigger than Jaguar, supposing Navi CU are also bigger than Polaris and we simply don´t have enough "size budget" to create the SoC.

Third, even if the PS5 GPU is bigger, how the hell will it leverage bandwidth speed? GDDR5X?  That would be a minuscule increase for an extra price. And the PS5 will need to have at least 12GB to reach full 4K most of the time. That is more $$$ over more $$$.

Conclusion.

There is no reason Sony would bet on new architecture at this point of the generation. If they feel the need to leapfrog the XoX, Sony can easily do that with more polished and overclocked Polaris+Jaguar pairs and 12GB of GDDR5. And may be they could get it at no more than $450.

Thoughts? 

Mostly correct, though there are a couple things I want to point out:

1. Navi doesn't explicitly mean the high-end variant - especially not if the rumors of a modular design akin to Epyc and Treadripper are true, in which case a console version would have less chips under the heatspreader than an high-end desktop GPU.

2. By 2020, 7nm would probably be possible instead of 14/12nm AMD is using right now. But even then the process wouldn't be perfected yet and thus yield rates would be bad and prices expensive. So unless Sony would go with the expensive risk production a PS5 would probably come in 14/12nm or 2021 earliest. A release this year is right out without Navi (as the text says; with a Vega chip it would technically be possible) and next year still highly improbable due to technological constrains. A 14/12nm chip would be technically possible, but it would be huge (and thus very expensive) and hard to cool (resulting in low clock speeds and thus low power), and thus hardly worth producing.

3. Bandwith is much less a problem as you make it out. Instead of the standard 256bit connection, 384 or even 512 bit could be used with GDDR5 and the bandwith problem mostly evaporates then. The Mainboard will cost about 5-20$ more because more lanes means more layers for the board but otherwise everything would be fine. Then there's the alternatives like HBM2 or HMC which both can provide more bandwith to a console, though at an elevated cost.

Oh, an 12 GB RAM would just be too small for a next-gen console. 16GB will be very much needed by then