• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Next-Gen PS5 & Next Xbox |OT| Speculation/Analysis/Leaks Thread

Next Gen Consoles Power Level


  • Total voters
    222

Norse

Member
Jun 7, 2004
2,903
190
1,550
I'm looking at the next Xbox as possibly having two skus at launch. Scarlett will be compatible with Xbox accessories so why not a core SKU that's just a Scarlett with small SSD and no controller or headset for $100 less than a Scarlett with a sshd, controller and headset? I'd buy it. Be curious to see if PS5 will also use PS4 accessories beyond the VR.
 

Ovech-King

Member
Feb 27, 2017
141
77
240
I think Microsoft should have removed that shot of the gpu chips from that reveal video because they've compromised their final gddr6 model. Sony can definitely take this , modify their final machine and run with the victory
 
  • Like
Reactions: Negotiator

JLMC469

Member
Feb 6, 2018
288
243
240
I think Microsoft should have removed that shot of the gpu chips from that reveal video because they've compromised their final gddr6 model. Sony can definitely take this , modify their final machine and run with the victory
Was it deliberately though? :unsure:
 
  • Thoughtful
Reactions: TLZ

SonGoku

Gold Member
Aug 16, 2018
3,825
3,617
560
Pretty deep dive into Zen 2 here noting the key improvements
The new platform and core implementation is designed around small 8-core chiplets built on TSMC’s 7nm manufacturing process, and measure around 74-80 square millimeters.
Each full CPU, regardless of how many chiplets it has, is paired with a central IO die through Infinity Fabric links.
If the IO will be a separate die then maybe consoles will go with a chiplet based APU, consisting of 3 dies:
  • GPU
  • CPU
  • IO
 

LordOfChaos

Member
Mar 31, 2014
9,120
1,172
710
If the IO will be a separate die then maybe consoles will go with a chiplet based APU, consisting of 3 dies:
  • GPU
  • CPU
  • IO
Is that what people were seeing above thinking it was a dedicated RT chip? I don't know what from Microsoft indicated that, i wouldn't expect it
 

Negotiator

Member
Jun 28, 2011
4,001
636
695
If the IO will be a separate die then maybe consoles will go with a chiplet based APU, consisting of 3 dies:
  • GPU
  • CPU
  • IO
Which consoles? MS has already spoiled Anaconda and its big ass monolithic die.

Do you think the PS5 could adopt a MCM approach to overtake Anaconda?
 

LordOfChaos

Member
Mar 31, 2014
9,120
1,172
710
I quoted the link you posted lol
I know lol, I'm just wondering where the idea that the RT hardware was connected with infinity fabric rather than on a monolithic GPU came from, like did the video show an IO chiplet or something and people assumed it was the ray tracing hardware (not you, just speculating on the chiplets)

IF is fast but not that fast, wouldn't expect RT hardware to be separated from the GPU even with it
 
Last edited:

SonGoku

Gold Member
Aug 16, 2018
3,825
3,617
560
Which consoles? MS has already spoiled Anaconda and its big ass monolithic die.

Do you think the PS5 could adopt a MCM approach to overtake Anaconda?
idk just trying to make sense of it
If Zen 2 has a separate io die how did ms create a monolithic apu

Thoughts?
 

SonGoku

Gold Member
Aug 16, 2018
3,825
3,617
560
The APU semi-custom chip for MS and Sony doesn't need to be chiplet.
The way i understood it is that the IO is on a separate die because it can't be shrunk to 7nm
How does a semi custom design override this limitation?
 

LordOfChaos

Member
Mar 31, 2014
9,120
1,172
710
The way i understood it is that the IO is on a separate die because it can't be shrunk to 7nm
How does a semi custom design override this limitation?
It's in the above AT deep dive

Each full CPU, regardless of how many chiplets it has, is paired with a central IO die through Infinity Fabric links. The IO die acts as the central hub for all off-chip communications, as it houses all the PCIe lanes for the processor, as well as memory channels, and Infinity Fabric links to other chiplets or other CPUs. The IO die for the EPYC Rome processors is built on Global Foundries' 14nm process, however the consumer processor IO dies (which are smaller and contain fewer features) are built on the Global Foundries 12nm process.

Hm, the monolithic shot is curious indeed. It would be a limitation, you can have different fabrication nodes across chiplets but not chips, so either MS fronted the cost to port it to 7nm, or they were being artsy.

(I'm still lost as to where the idea that RT hardware was over infinity fabric came from)
 
Last edited:
  • Like
Reactions: SonGoku

ethomaz

Member
Mar 19, 2013
24,337
3,536
800
37
Brazil
Internet forums lol
From what I understood the chiplet choice of Zen 2 was more about EPYC Rome so they can have a central IO die with how many 8-core Zen2 dies around... so you have scalability with 8, 16, 24, 32, etc core using the same design.

Another issue AMD had is that their IO controller is still designed to 12nm... so instead to delay even more Zen 2 they choose to use chiplet to consumer processor but I'm sure AMD doesn't need that much time to make the IO controller 7nm but with Zen 2 and Navi maybe they had a time issue for that.
 
Last edited:
  • Thoughtful
Reactions: SonGoku

SonGoku

Gold Member
Aug 16, 2018
3,825
3,617
560
From what I understood the chiplet choice of Zen 2 was more about EPYC Rome so they can have a central IO die with how many 8-core Zen2 dies around... so you have stability with 8, 16, 24, 32, etc core using the same design.
Yeah, nailed it.
Found this:

Separating CPU chiplets from the I/O die has its advantages because it enables AMD to make the CPU chiplets smaller as physical interfaces (such as DRAM and Infinity Fabric) do not scale that well with shrinks of process technology. Therefore, instead of making CPU chiplets bigger and more expensive to manufacture, AMD decided to incorporate DRAM and some other I/O into a separate chip. Besides lower costs, the added benefit that AMD is going to enjoy with its 7 nm chiplets is ability to easier bin new chips for needed clocks and power, which is something that is hard to estimate in case of servers.
I see so not really a limitation, AMD just chose to use a separate io die to make it easy to scale up
 

Imtjnotu

Member
Nov 4, 2017
990
1,384
380
san diego
It's in the above AT deep dive

Each full CPU, regardless of how many chiplets it has, is paired with a central IO die through Infinity Fabric links. The IO die acts as the central hub for all off-chip communications, as it houses all the PCIe lanes for the processor, as well as memory channels, and Infinity Fabric links to other chiplets or other CPUs. The IO die for the EPYC Rome processors is built on Global Foundries' 14nm process, however the consumer processor IO dies (which are smaller and contain fewer features) are built on the Global Foundries 12nm process.

Hm, the monolithic shot is curious indeed. It would be a limitation, you can have different fabrication nodes across chiplets but not chips, so either MS fronted the cost to port it to 7nm, or they were being artsy.

(I'm still lost as to where the idea that RT hardware was over infinity fabric came from)
I'm guessing the only way to explain where the hardware goes. If it's using a monolithic Apu then maybe on a separate part of the pcb is the rt hardware. Tho we have no clue what that hardware is currently
 

LordOfChaos

Member
Mar 31, 2014
9,120
1,172
710
I'm guessing the only way to explain where the hardware goes. If it's using a monolithic Apu then maybe on a separate part of the pcb is the rt hardware. Tho we have no clue what that hardware is currently

@llien said Microsoft hinted at this on the last page, where was that?

IF is 42GB/s chip to chip, thats' fast compared to PCI-E but not fast in the context of internal monolithic GPU workings, I was just wondering if that idea had a strong hint towards it. I don't think RT over Infinity fabric sounds that plausible but I don't know what the hint referenced was
 
Last edited:
  • Like
Reactions: ethomaz

CyberPanda

Banned
Mar 4, 2019
8,817
13,355
1,130
From Ree:

Richard Leadbetter in Digital Foundry's Microsoft E3 / Xbox Scarlett initial impressions video had this comment:

"I've spoken to developers about this across the year once this sort of concept [dual-SKU approach w/Lockhart] became known and there has always been to everyone I've spoken to about it, a certain degree of suspicion and worry. If we look back at the console generations the lowest spec machine has always been a pain in the bum to work with. ... [Based on Xbox's E3 conference] Is it [Lockhart] gone? That is the question because it certainly seems like it."

With Microsoft describing it as a single console at this years E3 makes me think the Lockhart has indeed been scrapped. And now Microsoft is no longer promising the strongest console outright which makes me believe the next-gen Xbox might now be a singular device with a price ceiling at $499. They might actually have been betting that Sony was releasing in 2019.
 

TLZ

Member
Oct 20, 2015
6,247
3,557
740
Sydney
Also looks like I was wrong aswell. I thought around 10TFLOPs with the chance of pushing 10.8 TF to beat Stadia, but I guess looking at the 5700 cards it seems that between 8-9.2 is much more likely than over 10 unless those chips shown are not comparable to what we get with the PS5.
Didn't some reliable person on Twitter confirm both consoles will definitely have more than Stadia's TFs?
 

CyberPanda

Banned
Mar 4, 2019
8,817
13,355
1,130
Didn't some reliable person on Twitter confirm both consoles will definitely have more than Stadia's TFs?
Navi is more efficient. 8.6 TFlops of Navi would be equivalent to a 10+ TFlop Vega, which is what Stadia server blade uses.
 

llien

Gold Member
Feb 1, 2017
5,687
2,798
700
@llien said Microsoft hinted at this on the last page, where was that?
The bit with infinity fabric references (and weird patent on what looked like GPU with HBM)

Navi is more efficient. 8.6 TFlops of Navi would be equivalent to a 10+ TFlop Vega
9TF Navi, beats 12.6Tflop Vega 64, as we see in AMD's official slides.
I'd argue against word "efficient" though. AMD simply shifted resources from raw computing power into other beef.
 
Last edited:
  • Thoughtful
Reactions: SonGoku

sinnergy

Member
Jun 16, 2007
2,140
38
985
Like I said months ago, we can rub our hands if one of the consoles get 11 TF.
9 - 10 seems more likely, 10 being the new 11.

The lack of a TF number from both sides tells it all.
 

TLZ

Member
Oct 20, 2015
6,247
3,557
740
Sydney
Navi is more efficient. 8.6 TFlops of Navi would be equivalent to a 10+ TFlop Vega, which is what Stadia server blade uses.
I understand. The guy on Twitter though did not know about RDNA then, and I'm guessing his source didn't either. So I'm hoping these are 10.5tf+ in RDNA numbers.
 

TLZ

Member
Oct 20, 2015
6,247
3,557
740
Sydney
Like I said months ago, we can rub our hands if one of the consoles get 11 TF.
9 - 10 seems more likely, 10 being the new 11.

The lack of a TF number from both sides tells it all.
I think both of them are keeping their cards too close to their chest, both playing their cards right, waiting for one another for that one slip up or impatience so they one up each other, even if it's just half a TF, to claim the most powerful beast PR for years.
 
  • Like
Reactions: Negotiator

Aceofspades

Member
Mar 31, 2015
2,180
396
375
My prediction:
-Both will be north of 11TF.
-Power difference will be marginal.
-Both targeting $499.
-Sony will have more elegantly designed SSD solution.
-RT implementation will be similar (lesser than Nvidia current offerings)
- Sony is doubling down on VR ( new headset)
-Sony will offer Premium subscription as stated in older rumor.
-Sony will offer a lite PS5 ( no disc drive) for $50 less
- MS will offer "weaker" Lockhart (less Ram , less GPU power) for $399
-Dualshock 5 will have better battery life and included camera.

This is just speculation on my part based on the different rumors we had and my own guessing. I don't know shit. Im not an "insider" or whatever.
 

TeamGhobad

Member
Oct 15, 2018
2,284
2,154
440
since both ms and sony are going with navi we can do a simple tflop to tflop comparison between the systems.
 

Lort

Member
Jul 9, 2017
827
555
360
My prediction:
-Both will be north of 11TF.
-Power difference will be marginal.
-Both targeting $499.
-Sony will have more elegantly designed SSD solution.
-RT implementation will be similar (lesser than Nvidia current offerings)
- Sony is doubling down on VR ( new headset)
-Sony will offer Premium subscription as stated in older rumor.
-Sony will offer a lite PS5 ( no disc drive) for $50 less
- MS will offer "weaker" Lockhart (less Ram , less GPU power) for $399
-Dualshock 5 will have better battery life and included camera.

This is just speculation on my part based on the different rumors we had and my own guessing. I don't know shit. Im not an "insider" or whatever.
You forgot to mention the things xbox will do better like...
More powerful ( if only a bit)
Quieter
Better ray tracing
 

ColdToffee

Member
Oct 24, 2017
1,898
1,849
485
Portugal
You forgot to mention the things xbox will do better like...
More powerful ( if only a bit)
Quieter
Better ray tracing
You are basing this on what?
What makes you think that Sony is completely unable to fix the noise problem they had in the past? They did patent a new cooling system since then.
 
  • Like
Reactions: demigod

sinnergy

Member
Jun 16, 2007
2,140
38
985
I think both of them are keeping their cards too close to their chest, both playing their cards right, waiting for one another for that one slip up or impatience so they one up each other, even if it's just half a TF, to claim the most powerful beast PR for years.
I think the TF number doesn’t have much weight this generation, nothing to win in PR.
 

sinnergy

Member
Jun 16, 2007
2,140
38
985
That said, with this whole new architecture, rdna and Zen2 we are in for a treat, even with 9 TF.
 
Last edited:
  • Like
Reactions: TLZ