Support NeoGAF

[Radical_3d]

Using the info from MLiD.
PS6: (48CUs × 64 × 2 × 3GHz) ÷ 1000 = 18.43 TFLOPs

This is impossible. That's like 1 more TF than the Pro. I know TF don't tell the whole story but it'd be so sad… I hope a bare minimum of 33TF.

 

[Gaiff]

PS6 looks around ~9070 XT and Xbox around ~5080

That's a negligible difference.

 

[Panajev2001a]

That's a negligible difference.

I think that both would perform a little bit better than that (especially compared to 9070XT at RT and PT workloads) as they will have a lot of the missing HW features efficient RT/PT depends on (on top of FSR4 and the low when it comes to upscaling but also ray-reconstruction).

Knowing UDNA/RDNA5 is so forward looking features set wise makes me happy about both new console lines.

 

[dgrdsv]

This is impossible. That's like 1 more TF than the Pro. I know TF don't tell the whole story but it'd be so sad… I hope a bare minimum of 33TF.

It is very possible if they want to sell the thing at lower price point than they sell the Pro at.
And I'd imagine that this is exactly what they want because $800 is not really a mass market pricing.
Also consider that they'll have loads of ML h/w in the thing with FSR4+ and AI based framegen helping to push more out of the same base processing power. RT h/w will likely be improved to Nvidia/Intel's levels as well (finally).
It should be able to beat the Pro while being on the same TFs. And if that would launch at say $500 then this is a good product - even if not exactly what we'd expect from a "next gen console" going off our previous experiences. But these times have passed, the whole industry is at the silicon scaling wall now, there are no exceptions and there won't be any "4X" type h/w advancements either.

That's a negligible difference.

~20% or so which would be similar to PS5 vs XSX.
At a similar retail price that would be fine.
MS will fuck up with selling it though, as they've done a number of times already.

 

[Physiognomonics]

Like I predicted earlier, it's going to be about 5070 vs 5080, so not a big difference like this gen but Magnus will have 2 more CPU cores so it could have noticeably better framerates in some games.

Now the last thing we don't know is if PS6 APU will be chiplet or not. I am expecting it won't be chiplet as it will be less flexible but cheaper overall. But Sony don't need to have MAGNUS setup as they'll keep the same GPU for the whole PS6 life, until the PS6 Pro.

 

[Panajev2001a]

It is very possible if they want to sell the thing at lower price point than they sell the Pro at.
And I'd imagine that this is exactly what they want because $800 is not really a mass market pricing.
Also consider that they'll have loads of ML h/w in the thing with FSR4+ and AI based framegen helping to push more out of the same base processing power. RT h/w will likely be improved to Nvidia/Intel's levels as well (finally).
It should be able to beat the Pro while being on the same TFs. And if that would launch at say $500 then this is a good product - even if not exactly what we'd expect from a "next gen console" going off our previous experiences. But these times have passed, the whole industry is at the silicon scaling wall now, there are no exceptions and there won't be any "4X" type h/w advancements either.

We have been against that wall for a while hence why I was not really looking forward to a mid generation console to begin with. It makes launching a true next gen platform only 3-4 years later at a lower price very complex (some people may wait for PS6 Pro at a premium price to get what they expected initially).

Still, kind of for the same reason Pro consoles leave a bit of a bitter taste in some people's mouth, there are advantages when you have a different expectation in devs' minds as you launch a new console.

- PS6 is likely to have considerable higher bandwidth RAM (and I still think 32 GB of it) while PS5 Pro could be considered a bit held back bandwidth wise… if they bet on an AI/ML angle they will need a fair bit of RAM dedicated to it and lots of bandwidth too (I think we can expect a non trivial effective SSD bandwidth, compression factor and base SSD raw speed included, boost too)

- UDNA/RDNA5 is a rather large jump efficiency and features wise (looking at KeplerL2's suggestions) over PS5 Pro's hybrid RDNA2.x roots… TFLOPS do not tell the whole story. RT/PT workloads might have a much larger than expected efficiency gains on RDNA5 compared to RDNA2 even when considering the RDNA4 derived RT units (SER, ray reconstruction / AI based denoising, etc… all bring a large efficiency delta… anything improving the performance of incoherent workloads can have massive gains there… I see Cerny investing time there rather than pure CU count to increase FLOPS)

- AI/ML: after their work on PS5 Pro's GPU and PSSR as well as the Amethyst project collaboration with AMD they kicked off, you can bet it will be a major focus of the new HW. Rendering wise sparse is kind of the new buzzword / keyword… less and less at native resolutions and more and more inferred by AI algorithms… again goes back especially to the RT/PT point above

- CPU wise we are expecting a rather large jump forward again from Zen 2 to Zen 6 (although a mix of Zen 6 and Zen 6C but I would rather have heterogeneous cores even if with some minor cuts like they did for Zen 2 on PS5, heterogeneous cores performance wise just bring headaches to game devs)… again something else that was unchanged from PS5 to PS5 Pro but would get a big change with a proper full generation upgrade

- You are expected to write new code and to have things changing under the hood and you know there is gain to be had there (PS5 Pro held back technology that would have required devs direct involvement and/or to have separate game packages / shader binaries too)

- You can have PS6 exclusive software and a cost effective next generation that also unlock a new even cheaper form factor would give people more incentive give to tackle it head on and get people To upgrade (the next portable being based on a limited but very similar tech as the big brother would still get people optimising for true same arch / features)

 

[peish]

~20% or so which would be similar to PS5 vs XSX.

I think it looks worse next gen. Ps6 will be around ~ 50% worse off, if it's only getting 160w tgp

 

[miller152]

We have been against that wall for a while hence why I was not really looking forward to a mid generation console to begin with. It makes launching a true next gen platform only 3-4 years later at a lower price very complex (some people may wait for PS6 Pro at a premium price to get what they expected initially).

Still, kind of for the same reason Pro consoles leave a bit of a bitter taste in some people's mouth, there are advantages when you have a different expectation in devs' minds as you launch a new console.

- PS6 is likely to have considerable higher bandwidth RAM (and I still think 32 GB of it) while PS5 Pro could be considered a bit held back bandwidth wise… if they bet on an AI/ML angle they will need a fair bit of RAM dedicated to it and lots of bandwidth too (I think we can expect a non trivial effective SSD bandwidth, compression factor and base SSD raw speed included, boost too)

- UDNA/RDNA5 is a rather large jump efficiency and features wise (looking at KeplerL2's suggestions) over PS5 Pro's hybrid RDNA2.x roots… TFLOPS do not tell the whole story. RT/PT workloads might have a much larger than expected efficiency gains on RDNA5 compared to RDNA2 even when considering the RDNA4 derived RT units (SER, ray reconstruction / AI based denoising, etc… all bring a large efficiency delta… anything improving the performance of incoherent workloads can have massive gains there… I see Cerny investing time there rather than pure CU count to increase FLOPS)

- AI/ML: after their work on PS5 Pro's GPU and PSSR as well as the Amethyst project collaboration with AMD they kicked off, you can bet it will be a major focus of the new HW. Rendering wise sparse is kind of the new buzzword / keyword… less and less at native resolutions and more and more inferred by AI algorithms… again goes back especially to the RT/PT point above

- CPU wise we are expecting a rather large jump forward again from Zen 2 to Zen 6 (although a mix of Zen 6 and Zen 6C but I would rather have heterogeneous cores even if with some minor cuts like they did for Zen 2 on PS5, heterogeneous cores performance wise just bring headaches to game devs)… again something else that was unchanged from PS5 to PS5 Pro but would get a big change with a proper full generation upgrade

- You are expected to write new code and to have things changing under the hood and you know there is gain to be had there (PS5 Pro held back technology that would have required devs direct involvement and/or to have separate game packages / shader binaries too)

- You can have PS6 exclusive software and a cost effective next generation that also unlock a new even cheaper form factor would give people more incentive give to tackle it head on and get people To upgrade (the next portable being based on a limited but very similar tech as the big brother would still get people optimising for true same arch / features)

We may end up with a scenario where those currently on base models will upgrade when the new gen comes out, while those on Pro models will upgrade when that one releases.
Although most Pro owners want the best experience regardless of price, so I expect they will get the new gen anyway.

 

[TheRedRiders]

I think it's good to remind people that we'll need to be looking beyond rasterization and pure shader performance going into the next-generation, ML will play a significant role in how these consoles perform and I it's clear that Mark Cerny and Playstation have invested a lot of effort and resources in this regard.

I also suspect the PS6 will have several specific customisations in this regard maybe to accelerate the performance or help with implementation. I'm not talking about "secret sauce" or anything, it's just based on Sony's past behaviour.

 

[SplunkyMunkey]

I had this few days ago.

Rumor  Post in thread '(Grain of Salt) Moore's Law is Dead: PS6 SoC codenamed 'Orion', PS6 Portable SoC codenamed 'Canis''

Jul 30, 2025

I wonder if Sony surprises us with the clocks again. It's looking like RDNA5/UDNA + N3P is shaping up to do well on clocks already.

I recall everyone being shocked by 2233MHz on the PS5, most were expecting 1500-1800, perhaps with a slightly bigger die. Some time back I was expecting 2.8-3.0GHz with 72CU for around 27TF Single Issue FP32 (+architectural gains), but maybe we see something well into the 3GHz range. If the clocks are good, the thermal solution will still likely be cheaper than a big die.

3.5-3.6GHZ will be the upper end. For 68CUs (you need 4 disable for binning)...

• Proelite

• 

Don't think there will be v cache because it's going to be challenging to cool as the chip heights will be different. V cache is directly on top of the CPU cores.

What if it doesn't have to sit under the cpu what if its like another 132 mb v cache die thats shared between the cpu and gpu like a controller and uses that ram when it need it

 

[Loxus]

This is impossible. That's like 1 more TF than the Pro. I know TF don't tell the whole story but it'd be so sad… I hope a bare minimum of 33TF.

I think it'll be 60CUs just like the PS5 Pro.

The Pro consoles are like a stepping ground to the the next console.

Desktop PCs going to 12 cores in a single CCD makes me believe the PS6 will also have 12 cores.

 

[Loxus]

I'd argue that the info MLiD has confirms chiplets for Magnus, but he has nothing for Canis/Orion to suggest that. The AMD presentation to Sony was to try and get them on board with using chiplets (obviously AMD want this as it allows them to take advantage of re-using stuff they already have in flight) - my guess is that Sony will go with a single chip. MS are 99% going with the chiplets as they will not put in enough orders to get economies of scale by themselves - but if they take an off-the-shelf chiplet that is already going into a GPU...

We'll see as more info becomes available - these are just my gut feelings based on how Sony/Cerny have approached things in the past. Cerny is not afraid of making changes away from the normal GPU/CPU IP for making things work better in a console. Microsoft are more interested in staying as close to standard GPU as possible as they want to be directx compatible etc (this makes perfect sense, but it means they are far less likely to ask for changes - instead they'll tell AMD what will be in directx and they'll negotiate together which parts of that spec will be hw optimised vs purely sw).

This leads to nonsense commentary like "XBox is the console with full RDNA" stuff - yeah because parts of that RDNA spec are literally directx - which makes no sense for Sony as they would actually have to carry wasted transistors or have some sub-optimal mapping between their APIs and directx.

Chiplets in large volumes are more cost effective from a yield standpoint.

Instead of throwing away a whole monolithic die on a costly 3nm node, as a 2-3 part chiplet design, only the bad chiplet part would be discarded. Resulting in a cheaper overall console.

Though, I do believe they'll go monolithic again. Chiplet design had advanced and are more cost-effective than a monolithic die the same size in high volumes.

 

[Tiago Rodrigues]

This is impossible. That's like 1 more TF than the Pro. I know TF don't tell the whole story but it'd be so sad… I hope a bare minimum of 33TF.

And a Pro costs 800 euros where i live.

I'll say this right here: Most if not almost all of the PS6 gen will have its games released on the PS5 as well. I mean...2013 PS4 is still getting plenty of games released to it 12 years later. With the portable coming and everything i really think Sony is aiming for the biggest MAU of their career by having a cheaper console (PS5) and its successor (PS6) + an affordable portable which will allow for PS5 games (and therefore PS6 that will barely get any exclusive, so they could have a PS5 version downscaled for a portable).

They would be able to have a cheaper entry point (PS5) + a portable solution and a more expensive, more future-proof product (PS6).

All this will allow PS platforms to surpass 150M MAU until 2030.

 

[GoldenEye98]

So PS5 backwards support isn't guaranteed.... but PS6 support is guaranteed because.... we just say it will?

Hard believe honestly. The gap between a 15W device and 160W device is significant.

 

[asdasdasdbb]

So PS5 backwards support isn't guaranteed.... but PS6 support is guaranteed because.... we just say it will?

Hard believe honestly. The gap between a 15W device and 160W device is significant.

Just MLID things.

 

[GoldenEye98]

Just MLID things.

I think he is saying it will happen because Sony will mandate it...but from a technical perspective it's objectively a wider gap than Series S and X.

 

[Felessan]

I'll say this right here: Most if not almost all of the PS6 gen will have its games released on the PS5 as well. I mean...2013 PS4 is still getting plenty of games released to it 12 years later. With the portable coming and everything i really think Sony is aiming for the biggest MAU of their career by having a cheaper console (PS5) and its successor (PS6) + an affordable portable which will allow for PS5 games (and therefore PS6 that will barely get any exclusive, so they could have a PS5 version downscaled for a portable).

With advanced AI upscaling and mandatory RT in games PS5 will soon become poor man's platform as it's performance/IQ will be abyssmal.
It's not PS4 to PS5 transition, even though this transition also accelerated when engines adopted SSD, now we have cost-cutting tech and tech that directly improve IQ at low cost, those will quickly widen the visual gap between generations.

 

[Radical_3d]

It is very possible if they want to sell the thing at lower price point than they sell the Pro at.
And I'd imagine that this is exactly what they want because $800 is not really a mass market pricing.

And a Pro costs 800 euros where i live.

Because the Pro is not a subsidized hardware. Sony is the last standing company doing this (until Pro) and I hope they do it with the PS6. Although the competition situation doesn't help.

But these times have passed, the whole industry is at the silicon scaling wall now, there are no exceptions and there won't be any "4X" type h/w advancements either.

Bruh. 18TF is not even 2X.

I'll say this right here: Most if not almost all of the PS6 gen will have its games released on the PS5 as well. I mean...2013 PS4 is still getting plenty of games released to it 12 years later. With the portable coming and everything i really think Sony is aiming for the biggest MAU of their career by having a cheaper console (PS5) and its successor (PS6) + an affordable portable which will allow for PS5 games (and therefore PS6 that will barely get any exclusive, so they could have a PS5 version downscaled for a portable).

They would be able to have a cheaper entry point (PS5) + a portable solution and a more expensive, more future-proof product (PS6).

All this will allow PS platforms to surpass 150M MAU until 2030.

This makes a lot of sense. But has nothing to do with the power spare to play games in the PS6.

 

[mrBarrelNut]

Using the info from MLiD.
PS6: (48CUs × 64 × 2 × 3GHz) ÷ 1000 = 18.43 TFLOPs

Well the leak said 40 - 48 CUs. 48 in the best case scenario, and that's assuming no disabled CU's out of those 48.

Anyways, Magnus is rumoured to have 40% - 70% more CUs than Orion. And I don't see 160W console clocking CU's to the max.

 

[Chittagong]

A power difference of the scale discussed in this thread is completely inconsequental. PS3 and Xbox 360, PS4 and Xbox One, PS5 and Xbox Series X had similar differences and it made no practical difference.

Only time it really made a difference was PS2 and Xbox, that was a difference that attracted players to jump over.

 

[THE:MILKMAN]

Well the leak said 40 - 48 CUs. 48 in the best case scenario.

Anyways, Magnus is rumoured to have 40% - 70% more CUs than Orion. And I don't see 160W console clocking CU's to the max.

The 160W is just an early number AMD threw in back in 2023. I'm sure that has/will change. I mean PS5 Pro pushes 240W at the wall on N4P so I doubt even on N3P PS6 will be much, if anything, below that at the end of the day.

 

[ecosse_011172]

A power difference of the scale discussed in this thread is completely inconsequental. PS3 and Xbox 360, PS4 and Xbox One, PS5 and Xbox Series X had similar differences and it made no practical difference.

Completely this.
Almost anyone who is that desperate for noticeably more power will get a PC.
Let's face it, only the hardest of hardcore Xbox fans will be buying their box next gen,XBox studios will only be publishing 2-3 games per ten at this rate nd they'll come to PS6/Switch 3 anyway.

 

[Bojji]

Dense Geometry Format (basically HW level Nanite): WO2025085121 DENSE GEOMETRY FORMAT WO2025085120 INTERSECTION TESTING ON DENSE GEOMETRY DATA USING TRIANGLE PREFILTERING

Streaming Wave Coalescer (out-of-order execution): US20250068429 STREAMING WAVE COALESCER CIRCUIT US20250130811 Spill-After Programming Model for the Streaming Wave Coalescer

Workgroup self-launch (reduced CPU and GPU-frontend bottlenecks): WO2025144455 LOCAL LAUNCH IN WORKGROUP PROCESSORS

Many improvements to RT cores (beyond current uarchs): US20250182377 CONFIGURABLE RAY/EDGE TESTING FOR CONVEX POLYGON GROUPS US20250200890 PRISM VOLUMES FOR DISPLACED SUBDIVIDED TRIANGLES US20250200865 RAY TRACING OF DISPLACED MICRO MESHES USING A BOUNDING PRISM HIERARCHY WO2025144454 SYSTEMS AND METHODS FOR DETECTING RAY INTERSECTIONS WITH DISPLACED MICRO-MESHES

PS6 looks around ~9070 XT and Xbox around ~5080

It does for RDNA5

RDNA5 is shaping up to be very impressive. I wonder if any of those features will be in DX13 (or whatever it will be called), that hardware level nanite as standard on consoles could be a game changer (of course devs will take years to use this).

It's amazing how you start making accusations when you have nothing else to say.

AMD and Mark Cerny confirm it's RDNA2.

The hysterical circus that the lunatic Xbox fanboys created on this forum was so unbearable that there was a period of time when any accusation of "PS5 rdna1.x" was banned from the PS5 hardware discussion thread.

=========================

And he ends the post by saying he'll repeat the same rhetoric from the current generation in the next.

The PS5 is RDNA5, obsolete, and only the Xbox is the new true UDNA.

https://en.wikipedia.org/wiki/Dunning–Kruger_effect

Obviously that discussion is not banned anymore. And I don't see (m)any Xbox fanboys around, MS killed most of them during this Xbox generation.

So PS5 backwards support isn't guaranteed.... but PS6 support is guaranteed because.... we just say it will?

Hard believe honestly. The gap between a 15W device and 160W device is significant.

PS5 games so far were designed to use 100% of the hardware, that device (0.5 of GPU power, slow memory etc.) won't be able to play those games in playable form.

Now with PS6 they want to force devs (from the start) to make games playable on 6x worse GPU, this is exactly what MS did this gen with Series S and X but two times worse. Of course assuming that MLiD is right about this rumor and this is not just bullshit...

 

[Mr.Phoenix]

We have been against that wall for a while hence why I was not really looking forward to a mid generation console to begin with. It makes launching a true next gen platform only 3-4 years later at a lower price very complex (some people may wait for PS6 Pro at a premium price to get what they expected initially).

Still, kind of for the same reason Pro consoles leave a bit of a bitter taste in some people's mouth, there are advantages when you have a different expectation in devs' minds as you launch a new console.

This is a you problem though, not a product problem.

The bottom line is that whatever the PS6 will be, it would be considerably more powerful than the PS5pro. To the tune of 80% - 100% more powerful. While the physical TF numbers may not say so, real world performce will. Because if anything is clear now, it is that just like we hit the sprite and polygon wall, we have hit the TF wall too. Where basing your conclusions of how powerful something is strictly based on its TF number would be myopic at best..... ignorant at worst. Even something as simple as having 3-4 times the cache within the APU vs the PS5/pro is enough to be a significant architectural difference.

But more specifically, this kinda concern is not even a real thing. There is absolutely nothing wrong with having a divergent product line. Just looking at the PS5... over 50M people bought it, likely knowing that the PS5pro would come to market at some point. It would be no different here. This notion that having one hurts the sales of the other is not a real thing.

- PS6 is likely to have considerable higher bandwidth RAM (and I still think 32 GB of it) while PS5 Pro could be considered a bit held back bandwidth wise… if they bet on an AI/ML angle they will need a fair bit of RAM dedicated to it and lots of bandwidth too (I think we can expect a non trivial effective SSD bandwidth, compression factor and base SSD raw speed included, boost too)

I will be shocked if the PS6 has anything more than 24GB of GDDR7 RAM. People keep making the mistake of looking at RAM usage in a desktop similarly to that of a console. It just doesnt work that way. As far as total memory goes,

192bit translates to just 6 channels instead of the typical 8. which means a bandwidth of 768GB/s with 32GT/s mem. And as of today, GDDR7 chip production is only focused on 2GB and 3GB chips. Though the spec allows for up to 4GB and 8GB modules. But no one (at least not a mass market OEM), is going to want to use 3/4GB chips. Unless they are being used in premium flagship devices. System design would always prioritize memory bandwidth over memory capacity. Meaning, sony would sooner go with 8 channels and use 2 or 3GB chips, than go with 6 channels and use 4 or 8GB chips.

But lets run with rumor as is....

192bus (which translates to 768GB/s of bandwidth) using 32GT/s mem with 2GB chips can have only 12GB of RAM total. with 3GB chips that is 18GB total. You will need 4GB chips to get to 24GB of RAM. And that is assuming that between now and sometime in 2028, Micron, SK and Samsung start making 4GB chips. That's a BIG ask. Possible, but iffy.

 

[Proelite]

What if it doesn't have to sit under the cpu what if its like another 132 mb v cache die thats shared between the cpu and gpu like a controller and uses that ram when it need it

That's not a 3d v cache then. 3d v cache has to be a specific layout on top of the entire CPU complex.

 

[TheRedRiders]

192bus (which translates to 768GB/s of bandwidth) using 32GT/s mem with 2GB chips can have only 12GB of RAM total. with 3GB chips that is 18GB total. You will need 4GB chips to get to 24GB of RAM. And that is assuming that between now and sometime in 2028, Micron, SK and Samsung start making 4GB chips. That's a BIG ask. Possible, but iffy.

Sheesh

 

[Proelite]

Chiplets in large volumes are more cost effective from a yield standpoint.

Instead of throwing away a whole monolithic die on a costly 3nm node, as a 2-3 part chiplet design, only the bad chiplet part would be discarded. Resulting in a cheaper overall console.

Also the bad chiplet might not be discard either, but rather binned more aggressively for the discreet budget desktop market.

 

[Proelite]

So PS5 backwards support isn't guaranteed.... but PS6 support is guaranteed because.... we just say it will?

Hard believe honestly. The gap between a 15W device and 160W device is significant.

Same reason the series s has series X compatibility despite not having Xbox One X compatibility.

 

[StreetsofBeige]

I will be shocked if the PS6 has anything more than 24GB of GDDR7 RAM. People keep making the mistake of looking at RAM usage in a desktop similarly to that of a console. It just doesnt work that way. As far as total memory goes,

Yup.

When it's fun time to guess specs, ram is always overestimated by gamers There are times the type of ram was wrong (8 gb GDDR5 in PS4 was a great surprise), but the amount of ram is always either overestimated or spot on. I dont think there are too many times, people underestimated the amount of ram and console makers pleasantly surprised gamers with an extra 25% or 50% ram.

I remember gamers thinking PS5 or X would have 24 gb. I think there was even chat at 32 gb. I think what gamers do isnt so much look at PC amounts of ram, but if you look at the spec trend in console gaming, it often goes in ~4x multiples. Sometimes 8x in the past. Ram from 360/PS3 gen to Xbox One/PS4 was 16x. So just take whatever spec is in old gen and just multiply by X to get new gen. So some gamers will think going from 16 to 24 looks off as that's only +50%.

 

[Loxus]

This is a you problem though, not a product problem.

The bottom line is that whatever the PS6 will be, it would be considerably more powerful than the PS5pro. To the tune of 80% - 100% more powerful. While the physical TF numbers may not say so, real world performce will. Because if anything is clear now, it is that just like we hit the sprite and polygon wall, we have hit the TF wall too. Where basing your conclusions of how powerful something is strictly based on its TF number would be myopic at best..... ignorant at worst. Even something as simple as having 3-4 times the cache within the APU vs the PS5/pro is enough to be a significant architectural difference.

But more specifically, this kinda concern is not even a real thing. There is absolutely nothing wrong with having a divergent product line. Just looking at the PS5... over 50M people bought it, likely knowing that the PS5pro would come to market at some point. It would be no different here. This notion that having one hurts the sales of the other is not a real thing.

I will be shocked if the PS6 has anything more than 24GB of GDDR7 RAM. People keep making the mistake of looking at RAM usage in a desktop similarly to that of a console. It just doesnt work that way. As far as total memory goes,

192bit translates to just 6 channels instead of the typical 8. which means a bandwidth of 768GB/s with 32GT/s mem. And as of today, GDDR7 chip production is only focused on 2GB and 3GB chips. Though the spec allows for up to 4GB and 8GB modules. But no one (at least not a mass market OEM), is going to want to use 3/4GB chips. Unless they are being used in premium flagship devices. System design would always prioritize memory bandwidth over memory capacity. Meaning, sony would sooner go with 8 channels and use 2 or 3GB chips, than go with 6 channels and use 4 or 8GB chips.

But lets run with rumor as is....

192bus (which translates to 768GB/s of bandwidth) using 32GT/s mem with 2GB chips can have only 12GB of RAM total. with 3GB chips that is 18GB total. You will need 4GB chips to get to 24GB of RAM. And that is assuming that between now and sometime in 2028, Micron, SK and Samsung start making 4GB chips. That's a BIG ask. Possible, but iffy.

If Sony improves the storage architecture that the GPU can access more data stored on the SSD directly and only use the GDDR7 for latency heavy data. We may actually see less RAM next gen.

 

[RAIDEN1]

The PS6 will eat the competition up! .....Oh wait..there isn't any..

 

[Radical_3d]

Only time it really made a difference was PS2 and Xbox, that was a difference that attracted players to jump over.

It was its worst generation. What matters are (and Phil couldn't get his head around this) games. As demonstrated in the glorious early 360 era.

 

[pacman4000]

But lets run with rumor as is....

192bus (which translates to 768GB/s of bandwidth) using 32GT/s mem with 2GB chips can have only 12GB of RAM total. with 3GB chips that is 18GB total. You will need 4GB chips to get to 24GB of RAM. And that is assuming that between now and sometime in 2028, Micron, SK and Samsung start making 4GB chips. That's a BIG ask. Possible, but ififfy.

2GB GDDR7 will be relatively cheap by 2027. Go with clam shell config (just like they did with PS4), 2×12=24GB, and then wait for 4GB modules.

 

[StreetsofBeige]

Just curious when it comes to ram configs, how does lets say 2 x 12 = 24 vs 4 x 6 =24 affect the gamer or dev?

Is there literally zero effect for gamers and devs? And it's more about company's COGs or form factor? Or there could be a gaming effect?

 

[LimanimaPT]

I own a Ps5 and I have absolutely no reason to get a Ps6... i feel I could have skipped the Ps5

 

[Det]

It was its worst generation. What matters are (and Phil couldn't get his head around this) games. As demonstrated in the glorious early 360 era.

Phil's excuse that "11/10 games don't make people switch platforms" is a typical loser's excuse.

He doesn't know how to make quality games, and that's why he says quality games don't matter.
It sounds like an incompetent worker screwing up at work and making up excuses to avoid getting fired.

Obviously that discussion is not banned anymore. And I don't see (m)any Xbox fanboys around, MS killed most of them during this Xbox generation.

All those people were also part of the Discord group that was spreading FUD.

First, all the users who had been spending time "building credibility" by pretending to be "impartial" disappeared.

What remained were the uncoordinated fanboys posting nonsense and making a fuss about absolutely every PS5 spec, even the delusions that the PS5 SSD was peak performance, burst performance, and that it wouldn't be "sustainable." LOL.

So, I wouldn't find it strange that there are already people doing this work today, gaining credibility by pretending to be impartial here, and then being in a good position to spread FUD.

 

[TheRedRiders]

MLiD claims he knows a lot more about the PS6 but isn't allowed to say much, and now claims that the ray-tracing performance is between 5 and 10x faster than the PS5 Pro. I suspect this will vary on a game by game basis but the uplift should be huge.

EDIT : I'm not as suspicious of this claim, I think Cerny gave it away during the PS5 Pro seminar when he said he expects several quantum in ray-tracing performance in the future.

 

[Gameplay Gods Bless]

RDNA5 is a pretty forward looking uarch. They have every feature Blackwell has and then some.

I don't see AMD catching up to Nvidia unless the 60 series is somehow worse than the 50 series was.

 

[Gameplay Gods Bless]

Dense Geometry Format (basically HW level Nanite): WO2025085121 DENSE GEOMETRY FORMAT WO2025085120 INTERSECTION TESTING ON DENSE GEOMETRY DATA USING TRIANGLE PREFILTERING

Streaming Wave Coalescer (out-of-order execution): US20250068429 STREAMING WAVE COALESCER CIRCUIT US20250130811 Spill-After Programming Model for the Streaming Wave Coalescer

Workgroup self-launch (reduced CPU and GPU-frontend bottlenecks): WO2025144455 LOCAL LAUNCH IN WORKGROUP PROCESSORS

Many improvements to RT cores (beyond current uarchs): US20250182377 CONFIGURABLE RAY/EDGE TESTING FOR CONVEX POLYGON GROUPS US20250200890 PRISM VOLUMES FOR DISPLACED SUBDIVIDED TRIANGLES US20250200865 RAY TRACING OF DISPLACED MICRO MESHES USING A BOUNDING PRISM HIERARCHY WO2025144454 SYSTEMS AND METHODS FOR DETECTING RAY INTERSECTIONS WITH DISPLACED MICRO-MESHES

PS6 looks around ~9070 XT and Xbox around ~5080

So PS6 is faster than a 5070 and 9070 but slower than a 5070ti and 4080?

 

[Mr.Phoenix]

If Sony improves the storage architecture that the GPU can access more data stored on the SSD directly and only use the GDDR7 for latency heavy data. We may actually see less RAM next gen.

That's actually possible. Extremely unlikely, though.

But if the APU is goingto be a chiplet, where the CPU and GPU are mostly separate components, then an argument can be made that we could also have a clear distinction between VRAM and system RAM. Where like 12GB of GDDR7 is used for VRAM exclusively, and like 8-12GB of LPDDRx is used for the CPU.

Sounds more expensive and complicated than just using 4GB GDDR7 chips, though.

2GB GDDR7 will be relatively cheap by 2027. Go with clam shell config (just like they did with PS4), 2×12=24GB, and then wait for 4GB modules.

read post above... the clamshell design isnt as simple as it sounds. Its not just having more mem chips, its also have a far more complex PCB.

It would be cheaper to use an 8 channel APU with 8 x 3GB mem chips to get 24GB, than to use a 6 channel APU with 12 x 2GB cause technically you have to have 12 mem channels for the clamshell to work (6 on either side of a PCB), and thats a far more expensive PCB in addition to just having way more mem modules. Also the added risk on increased potential points of failure complicates manufacturing.

The only time anyone goes for clamshells, is never to save cost, its when there is just no other way to get the required memory capacity. In this case, there are already going to be cheaper paths.

 

[SplunkyMunkey]

So PS6 is faster than a 5070 and 9070 but slower than a 5070ti and 4080?

People debating about the differences between the ps6 and ps5 pro. I personally think with those specs it ain't gonna be much powerful compared to ps5 pro its only gonna be stronger in certain areas like Ai or ML even if its going to be clocked at 3ghz to 3.5ghz with 45 CU's its still going to output the same teraflop performance to ps5 pro

 

[Proelite]

Based on K KeplerL2 's leak that PS6 SOC is taping out next quarter, I don't think in the end Sony went with chiplets.

Because it they did they'll just be using off the shelf AMD discrete parts and taping out won't be a step anymore in console SOC development.

Maybe the tape out is only for Canis.

 

[viveks86]

MLiD claims he knows a lot more about the PS6 but isn't allowed to say much, and now claims that the ray-tracing performance is between 5 and 10x faster than the PS5 Pro. I suspect this will vary on a game by game basis but the uplift should be huge.

EDIT : I'm not as suspicious of this claim, I think Cerny gave it away during the PS5 Pro seminar when he said he expects several quantum in ray-tracing performance in the future.

Exciting if true. Just another step towards making Path Tracing a viable standard (at least for a 30 fps quality mode). Given that even games like Persona are starting to use RT, it's all just a matter of time before the rendering paradigm shifts altogether.

Based on K KeplerL2 's leak that PS6 SOC is taping out next quarter, I don't think in the end Sony went with chiplets.

Because it they did they'll just be using off the shelf AMD discrete parts and taping out won't be a step anymore in console SOC development.

Maybe the tape out is only for Canis.

If MLiD is to be believed, then isn't chiplet already confirmed? Would be nice to see Kepler confirm that as well as I still can't fully trust MLiD.

 

[GoldenEye98]

Same reason the series s has series X compatibility despite not having Xbox One X compatibility.

As I said....the gap here is so much larger.

 

[Lagspike_exe]

MLiD claims he knows a lot more about the PS6 but isn't allowed to say much, and now claims that the ray-tracing performance is between 5 and 10x faster than the PS5 Pro. I suspect this will vary on a game by game basis but the uplift should be huge.

EDIT : I'm not as suspicious of this claim, I think Cerny gave it away during the PS5 Pro seminar when he said he expects several quantum in ray-tracing performance in the future.

Press X for doubt meme.

 

[yogaflame]

Just give me 4k/60 fps with RTGI as standard, and that suffices. I'm more curious and interested on what will be the new features of ps6 especially on the new dual sense, sound, connectivity with PSPnextgen, and new version of PSVR.

 

[nial]

So PS5 backwards support isn't guaranteed.... but PS6 support is guaranteed because.... we just say it will?

Hard believe honestly. The gap between a 15W device and 160W device is significant.

Just MLID things.

I think he is saying it will happen because Sony will mandate it...but from a technical perspective it's objectively a wider gap than Series S and X.

Of course stuff natively running on more powerful hardware isn't guaranteed (PS5 games in full power mode), as opposed to stuff directly targeting Canis (PS6P games).
Read what he said again, PS5 games in low power mode will be BC with Canis.

 

[mayham2199]

i feel I could have skipped the Ps5

Hell no. I agree about PS5 to PS6 but the PS4 is garbage. Games take way longer to load up, store is slow as hell, pretty much just using the OS in general is way slower and to slow IMO. Plus all games at 30 FPS? I don't know how anyone can be on PS4 still honestly.

 

[nial]

Hell no. I agree about PS5 to PS6 but the PS4 is garbage. Games take way longer to load up, store is slow as hell, pretty much just using the OS in general is way slower and to slow IMO. Plus all games at 30 FPS? I don't know how anyone can be on PS4 still honestly.

I was just playing Bloodborne (runs the same everywhere) and it was fine.

 

[Sethbacca]

Now with PS6 they want to force devs (from the start) to make games playable on 6x worse GPU, this is exactly what MS did this gen with Series S and X but two times worse. Of course assuming that MLiD is right about this rumor and this is not just bullshit...

As long as parity is optional and not required then I'm fine with it. The parity requirement is what really fucked MS with the S/X debacle imo, but saying from the get go that some games may not see release on the Series S would probably have left it even more dead in the water.

At least with a handheld you can handwave away some of the parity discussions by saying it's a handheld and it shouldn't be expected to run full fledged PS6 games.