CoreyFlawless
Member
What ever this come out of. Gta6 gonna play amazing on it
[MLID] Zen 6 Magnus Leak: AMD's MASSIVE APU for next gen console (+ Medusa Point Specs)
- Thread starter Lunatic_Gamer
- Start date
- Rumor Hype
SegaSnatcher
Member
Very good point!
ExtendEmbraceExtinguish
Member
Software is what they will give. You either buy it or gamepass. We're not getting 50 million new IPs look at the top 20 lists.
They will sell Software but software isn't what you were referring to:
If this is what's important then a $1000 "portable PC" has no chance against a Switch and an equally expensive "nextbox" no chance against a PS5.
Last edited:
ExtendEmbraceExtinguish
Member
100% correct. The devices are your buy in however the software will be the difference.
Loxus
Member
Using the 7900xtx as a reference and taking in to account scaling of all units, moving from 5nm to 3nm should land ~215mm².
This leaves room for other stuff.
But I'm still questioning the memory controllers. PHY don't scale well at all.
For example, here is the 5090, which shows 6 GDDR7 PHY on both left and right sides similarly to MLiD diagram.
This is a massive 750mm² chip.
Now if you zoom in close to one of the GDDR7 PHY, you can see it's split into two 16bit slices. Could he have misinterpreted one of these 16bit slices as a 32bit PHY?
If we look at it that way, 6 × 32bit = 192bit-bus.
A 192bit bus isn't as bad as it seems.
Micron reveals the future of GDDR7 memory
One of the most interesting aspects of GDDR7 isn't its raw bandwidth, it's the memory's planned capacities. 16Gb (2GB), 24Gb (3GB), 32Gb (4GB), 48GB (6GB), and 64GB (8GB) modules are part of the GDDR7 standard.
Possible capacities:
6 chips × 3 GB = 18 GB
6 chips × 4 GB = 24 GB
6 chips × 6 GB = 36 GB
SK Hynix launches GDDR7 with up to 40Gbps speed
And possible bandwidth:
192bit-bus ÷ 8 × 28 = 686GB/s
192bit-bus ÷ 8 × 32 = 784GB/s
192bit-bus ÷ 8 × 36 = 882GB/s
196bit-bus ÷ 8 × 40 = 980GB/s
The benefits of a 192bit-bus is less power, complexity and cost.
Proelite
Member
Using the 7900xtx as a reference and taking in to account scaling of all units, moving from 5nm to 3nm should land ~215mm².
This leaves room for other stuff.
But I'm still questioning the memory controllers. PHY don't scale well at all.
For example, here is the 5090, which shows 6 GDDR7 PHY on both left and right sides similarly to MLiD diagram.
This is a massive 750mm² chip.
Now if you zoom in close to one of the GDDR7 PHY, you can see it's split into two 16bit slices. Could he have misinterpreted one of these 16bit slices as a 32bit PHY?
If we look at it that way, 6 × 32bit = 192bit-bus.
A 192bit bus isn't as bad as it seems.
Micron reveals the future of GDDR7 memory
One of the most interesting aspects of GDDR7 isn't its raw bandwidth, it's the memory's planned capacities. 16Gb (2GB), 24Gb (3GB), 32Gb (4GB), 48GB (6GB), and 64GB (8GB) modules are part of the GDDR7 standard.
Possible capacities:
6 chips × 3 GB = 18 GB
6 chips × 4 GB = 24 GB
6 chips × 6 GB = 36 GB
SK Hynix launches GDDR7 with up to 40Gbps speed
And possible bandwidth:
192bit-bus ÷ 8 × 28 = 686GB/s
192bit-bus ÷ 8 × 32 = 784GB/s
192bit-bus ÷ 8 × 36 = 882GB/s
196bit-bus ÷ 8 × 40 = 980GB/s
The benefits of a 192bit-bus is less power, complexity and cost.
Loxus
Looking at the GDDR6 controllers on the Navi 31 memory dies, you can fold them in half.
Edit: So the GDDR6 phys on the went from 13.7mm2 on TMSC 6nm to 6mm2 on TSMC4nm, that's over 2x in scaling when it should only be 1.26x.
So phy controller scales better than logic.
Can you double check my math? Below is the folded GDDR6 phy controllers on the 6 memory dies, 37mm2 on N6.
Last edited:
thicc_girls_are_teh_best
Member
Test
P Physiognomonics
Why would PS6 not have L3$ for the CPU when PS5 does? I think you meant L4$, and there's still a chance PS6 will have L4$ (V-Cache) shared between CPU & GPU.
nkarafo
I get where you're coming from. While I'm still interested to see what 10th gen systems will bring, I can't pretend like this isn't going to be more of the same in a lot of cases. The next PlayStation and Xbox systems will be largely similar as they're pulling from the same APU vendor's roadmap.
Audiophile
Pretty interesting CPU configuration concept there. I'd like to see something like that if for no other reason than to bring some flavor to the design and to see how devs work within that type of setup.
P Physiognomonics
Why would PS6 not have L3$ for the CPU when PS5 does? I think you meant L4$, and there's still a chance PS6 will have L4$ (V-Cache) shared between CPU & GPU.
nkarafo
I get where you're coming from. While I'm still interested to see what 10th gen systems will bring, I can't pretend like this isn't going to be more of the same in a lot of cases. The next PlayStation and Xbox systems will be largely similar as they're pulling from the same APU vendor's roadmap.
Audiophile
Pretty interesting CPU configuration concept there. I'd like to see something like that if for no other reason than to bring some flavor to the design and to see how devs work within that type of setup.
Last edited:
Loxus
Member
It's not folded in half, it's two 32bit GDDR6 memory controllers on each MCD.
Looking at the GDDR6 controllers on the Navi 31 memory dies, you can fold them in half.
Edit: So the GDDR6 phys on the went from 13.7mm2 on TMSC 6nm to 6mm2 on TSMC4nm, that's over 2x in scaling when it should only be 1.26x.
So phy controller scales better than logic.
Can you double check my math? Below is the folded GDDR6 phy controllers on the 6 memory dies, 37mm2 on N6.
But you're on to something.
It would make sense if a 32bit PHY is folded instead of going long ways.
Sony did something similar with the PS4.
Orbis is "Circle" in Latin.
THE:MILKMAN
Member
So if we assume Magnus is the next-gen Xbox SoC then what timeline for release we looking at? Still late '26 or like both Sony and Microsoft have many times release at the same time in late '27 or '28?
jm89
Member
Keplar believes iirc 2027.
Proelite
Member
Ok in the RX 5090 die shot they are all in one straight lines but in the Navi 31 MCD they are stacked, one 32 bit on top of each other.It's not folded in half, it's two 32bit GDDR6 memory controllers on each MCD.
But you're on to something.
It would make sense if a 32bit PHY is folded instead of going long ways.
Sony did something similar with the PS4.
Orbis is "Circle" in Latin.
6 32bit Phy on each side in two rows should fit the long edge of a 264mm2 die.
Loxus
Member
Keep in mind the 5090 is 750mm².
This is how I envision how it would look.
It's just a quick something using Strix Point (Ryzen AI 9) and RDNA3 die shot.
It's most likely 96CU like Kepler was saying.
Ignore all the SSD stuff, just wishful thinking.
kevboard
Member
Last edited:
thicc_girls_are_teh_best
Member
Personally Team 2026, considering that's also Xbox's 25th anniversary. But it could be a "soft launch" then.
And that's still assuming Magnus is an Xbox APU, but a lot of signs point to it being the case.
Proelite
Member
Nice work. How much bigger is the WGP in your design compared to RDNA 4 assuming the same node.Keep in mind the 5090 is 750mm².
This is how I envision how it would look.
It's just a quick something using Strix Point (Ryzen AI 9) and RDNA3 die shot.
It's most likely 96CU like Kepler was saying.
Ignore all the SSD stuff, just wishful thinking.
Proelite
Member
PS4 Pro codename was neo.
Wow nice jobKeep in mind the 5090 is 750mm².
This is how I envision how it would look.
It's just a quick something using Strix Point (Ryzen AI 9) and RDNA3 die shot.
It's most likely 96CU like Kepler was saying.
Ignore all the SSD stuff, just wishful thinking.
THE:MILKMAN
Member
Yeah Kepler reminded me that the Shakespeare related codenames like Oberon are AMD's and the Matrix ones are Sony's. So from that I take it that Zion being the Sony codename for PS6 is a good possibility.
Loxus
Member
Hard to tell without a proper RDNA4 die shot.
adamsapple
Or is it just one of Phil's balls in my throat?
60% denser Logic, SRAM and Analog are pretty much the same.
Quoting myself from just a few days ago
Die size suggests something like 80 CUs at probably 3GHz or so, which puts Sparse Matrix FP4 performance at almost 2000 Teraflops or 83x faster than the Xbox Series X using FP16 (technically still not the biggest leap ever, but a very large number at least)
Loxus
Member
These are most of the codenames for Playstation from PS4/PS5 generation.
Shakespeare references
(PS5 generation)
Prospero - PS5 console codename
Oberon - PS5 production SoC
Ariel - Early PS5 prototype
Gonzalo - Even earlier PS5 prototype
Viola - PS5 Pro SoC
The Matrix references
(Pro models)
Neo - PS4 Pro console codename
Trinity - PS5 Pro console codename
Superman universe reference
Krypton - PS4 Pro dev kit codename
UK cities
(AMD SoCs for PS4 family)
Liverpool - PS4 launch SoC
Newcastle - PS4 Slim SoC
Glasgow - PS4 Pro SoC
River name
(AMD console refresh)
Monongahela - PS4 Slim console codename
Latin / classical / mythology
Orbis - PS4 base console codename (Latin: "Circle")
Vita - PS Vita retail name (Latin: "Life")
Internal / dev references
Devin - PS5 dev kit codename
NGP - PS Vita codename ("Next Generation Portable")
Last edited:
Viola was just PS5 Pro, PS5 Slim was Oberon Plus.
During development PS5 SoC revisions also used the codenames Bishop, Cardinal and Flute.
RodriGames40
Member
This would be around 4 times faster assuming UDNA architecture. This is coming close to the 4090, and that's 4.5 times faster.
Loxus
Member
Forgot about Flute, I thought that was a random name.
MisterXDTV
Member
Sony won't release a PS6 until they know that developers are nearly done with PS5 as basis....
The right question should be:
"How close are we to the limits of PS5? (also considering they added a PS5 Pro as well a few months ago)
Proelite
Member
Moore law is dead, any console more powerful will cost more and will not get cheaper overtime
More expensive means less public willing or capable of buy which in turn means more risk to publishers to launch games only on the console
So the question remains, if they launch a console called PS6 and all or almost all the games continue to be made and launched for and on the ps5 is this truly a PS6 or is just a PS5 Pro 2?
Last edited:
nial
Gold Member
Sales dictate everything, SIE is expecting to ship 15M PS5 units on the fiscal year ending March 31, 2026.
The generation is over.
Looking at that floorplan 6 SE x 8 WGP config does seem more likely for Magnus. That is one seriously overkill GPU for a console but I guess MS doesn't another PS5 vs Xbox series X situation where power difference is almost non-existent.
neocycle
Member
Really? Ok then.
Loxus
Member
That is more of a render, than an actual die shot.
And I'm not too sure how accurate that is because the WGP are facing in the other direction within the Shader Array.
Usually, two WGP or 2 Shader Arrays face each other via the side with the TMU and RT unit.
But in this case, the TMU and RT unit are to the edge of the Shader Engine, with the SIMD32 from each Shader Array touching each other.
Here is a slightly better image.
Last edited:
jm89
Member
I don't think devs will be done with a ps5 even well into the ps6 life.
Too much risk nowadays to be dumping the previous gen.
Last edited:
TheRedRiders
Member
That's a good point you've raised, I'm curious what the narrative and discourse will be going into "next-gen", and surrounding these consoles. Sony will likely. and based on past history go down an efficient route, like they did with PS5's "fast and narrow design", relatively less CU's but a faster clocked GPU or Xbox's slower but wider approach, more CU's but clocked relatively slower.
If we assume Magnus is Xbox's next chip, then it's likely they'll be sticking with the same design philosophy, but this time ensuring there's a clear advantage in performance compared to the competing PS6. Personally I welcome it, but I'm curious how the market will react to a product will likely could be priced $1000 +
What might also complicate things is the ML technology, PS5 Pro using PSSR, and will soon incorporate into it FSR 4, I'm wondering if Microsoft will also adopt FSR 4 and how this may complicate things considering Project Amethyst. I'm wondering if Microsoft may develop their own ML based upscaling technology.
Barakov
Member
Radical_3d
Member
Another big brain move from Spencer. Make the difference in hardware when it became less and less important due to the quality of AI upscalers. Truly the brain behind GamePass.
Proelite
Member
Considering that PS6 is releasing in 2027+, I don't see how they can guarantee a substantial power difference when PS6 will be on n2p and the Series X2 will be on n3p. It's a 40% density gap. That's not their goal because it's not going to be a realistic goal. Their goal is to be the best gaming PC one can get in a console like hardware / form factor (i.e. APU + soldiered GDDR7 main memory) that can rival RTX 5080 / 4090 performance in 2026. By 2028/2029, they'll have a successor that will trounce the PS6 handily.
Hypothetical PS6 in 2027 will have
8-10 core Zen 6
24GB GDDR7 on 256 bit bus (896GB/s bandwidth)
64-128mb Infinity Cache
68/72 Cus at 3.5 ghz (up to ~61 teraflops)
whichever forward UDNA 2/3 features Cerny decides to incorporate due to Project Amethyst.
~300mm2 die size
$599-$699 (Most likely $699)
such a PS6 will likely perform as well if not better with optimized console SKUs of games.
Last edited:
Proelite
Member
Keep in mind the 5090 is 750mm².
This is how I envision how it would look.
It's just a quick something using Strix Point (Ryzen AI 9) and RDNA3 die shot.
It's most likely 96CU like Kepler was saying.
Ignore all the SSD stuff, just wishful thinking.
Looking at the diagram again, your WGPs are really wide compared to the ones in RDNA3 and RDNA2. Can you probably can easily do 1-2 more columns in every shader array for a total of 120CUs to 144CUs like I had calculated earlier.
HeisenbergFX4
Member
Tobimacoss
Member
Proelite
Member
It'll have proportionally less compared to current gen because the install base of the console + the handheld will not cross 25 million.
Last edited:
Says who?
Proelite
Member
That's disappointing NGL. Sony and MS have always being within one year after volume production of a node since Xb1/PS4 gen.
I'm guessing the AI boom is affecting availability of advanced nodes.
Last edited:
Tobimacoss
Member
Install base is irrelevant though. MS won't be selling Steam/Epic versions, Xbox will be the default front facing store, in order to get a Xbox certification for publishing on Xbox Store, it will have to be an Xbox ecosystem/platform game created by the GDK and/or optimized by the GDKX for the set of AMD APUs.
Besides, it's not just handhelds and consoles, but Xbox PCs, Xbox Laptops, Xbox Cloud all come into play as well.
Unknown?
Member
Or my 30k super computer.
Loxus
Member
It's the same size as RDNA3.
I scaled, then trace over each block.
RDNA3.5 is even wilder.
The WGP is constantly changing.
Like Kepler said, it'll be bigger.
kevboard
Member
well, next gen will need serious horsepower as that generation will be the raytracing generation.
current gen is the beginnings of raytracing, and next gen will be when it will be matured. similar to how gen5 was the first steps into 3D, while gen6 finally was actually good at 3D.
AI upscalers only help so much when it comes to raytracing.
if you play Doom TDA with a very low internal resolution, you'll start seeing very obvious artifacts from the drastically reduced ray count.
this doesn't happen with Doom Eternal as it doesn't rely on RT GI.
so at the same internal resolution Eternal will look better/more stable than Dark Ages, which means you can use more aggressive reconstruction without introducing to many issues.
ray reconstruction can help with with a bit, but can also introduce its own artifacts.
so both manufacturers need to make sure they have enough horsepower that raytracing, and possibly pathtracing run well and look clean.
you don't want to be the system that has constant boiling artifacts or excessive ghosting compared to the other one.
