Support NeoGAF

[demigod]

Regardless of his reporting, and I do trust him, I do think they can achieve something like this if you remember what Mark corny said before PS5 launched and it ended up being true. Not to look into spec too much and the efficiency of the console has proven itself.

Mark CORNY.

 

[EN250]

Sign me up for that . The ps4 was one of my favorite consoles .

Legendary console for sure, hope PS6 is a return to those days

 

[Tobimacoss]

I can see them doing that when the PS6 install base is small at the beginning during that crossgen period but the series s really began to struggle when that crossgen period was over and next gen only games began. That was when devs started asking for the mandate to be dropped.

Simplest solution Sony could do is, Mandate but with an exception only IF streaming rights are allowed via Cloud streaming.

That way the handheld will at least be able to stream game from same license copy.

 

[viveks86]

Series S problem was probably the RAM, the CPU was almost identical to the X.

Right. That's my fear. It will always be something or the other. Such linear scaling on paper has almost never worked in practice when the difference is more than 2x. We are looking at 6x here. I don't know. I'm just gonna choose not to panic till 2027.

I'm not panicking. Does this sound like I'm panicking? Cuz I'm not….

 

[Mr Moose]

Right. That's my fear. It will always be something or the other. Such linear scaling on paper has almost never worked in practice when the difference is more than 2x. We are looking at 6x here. I don't know. I'm just gonna choose not to panic till 2027.

I'm not panicking. Does this sound like I'm panicking? Cuz I'm not….

 

[Venom Snake]

the goal now, in terms of new GPUs in general, is probably to just push raytracing perfomance to a point where it just work without any developer needing to be worried about perfomance.
raster perfomance basically hit its limits, at the very least when we look at PC cards like the 5090, which could probably render anything raster based you could throw at it without issue.

so let's hope AMD's RT perfomance is gonna make some big leaps soon

I mean, if you think about it, game developers are already capable of creating large, dense, geometry-rich environments that don't require high-end hardware to run efficiently. And this has been going on for years now.

AI and RT are basically all we need, besides, there are more problems with modern games that need addressing and i wish the term "next gen" covered that too, as playing the same crap over and over again but with more sugar coating isn't fun for me anymore.

 

[Guardian Monkey]

Isn't like 3x raster of ps5 closer to 3080? Where is the 5080 talking coming from and what amount of power would need to run something like that in a console coming out in two years the powerdraw would still be pretty massive even @3nm no?

 

[Mr Moose]

Isn't like 3x raster of ps5 closer to 3080? Where is the 5080 talking coming from and what amount of power would need to run something like that in a console coming out in two years the powerdraw would still be pretty massive even @3nm no?

No. A 9060XT is close to a 3080.

 

[saintjules]

My guess is that they will try to sell raytracing vs no raytracing this gen. The games will have raytracing support on PS6 and limited or none on PS5.

The xbox one x was $500, the series x and S consoles $500 and $300. I think it will be around $600 for the PS6 if it is cheaper but may end up the same as its launch price at $700. Any cheaper and I will be very pleasantly surprised.

Makes sense. It would be interesting to see pricing at sub $500. I still think it's unlikely.

Alot of people say RT is too expensive and doesn't make much difference. I'd be happy/expecting to see if I saw games that look like this next-gen.

 

[nial]

I don't know if cheaper guarantees a win. Series S sold more than X but not even close to what PS5 sold.
Maybe it's also the software, I don't know.

Also, i hope they will only port the GaaS stuff and let their first party studios focus on Sony consoles and provide the best possible games instead of "losing" time to optimize for S, Switch 2 etc.

Versus a $1000 system? It does.

 

[nial]

I have been saying for months that they would use gfx13 (RDNA5), TSMC N3P node, Canis would have 16 CUs and 15W TDP, that they would come out in 2027, etc.

16 CUs was new to me, huh. Much appreciated!

 

[nial]

Wait. are the two lower bars in the lower right corner due to GaaS or what?

Probably. One might be Bungie.

Sorry for the triple post, but that's for buying Bungie.

 

[Embearded]

This whole gen was narration xss was bad decision as it was problematic to release game both on xss and xsx, imagine if sony will force their first party studios to relase games both on ps6 and handheld with 0.5 power of ps5 . Sony is realy on good path of digging own grave ;d

I think they will be fine if both versions have the same amount of RAM and same CPU.
Surely it will require extra work from the devs but not as much as Series S.

 

[sncvsrtoip]

I think they will be fine if both versions have the same amount of RAM and same CPU.
Surely it will require extra work from the devs but not as much as Series S.

I wont have same amount of ram (128b bus) and same and same cpu power at 15w ;d

 

[ExtendEmbraceExtinguish]

This same talking point was used by certain people to justify the PC ports. You know, PC, that platform with billions of users?

Where are the new 1P internally-developed PS5 games this gen that weren't already in development prior to PS5's launch? How many of them aren't GAAS?

We need to stop using this line of thinking because it's proof now it's bullshit. Any extra money they get from multiplatform sales, are going right to the stocks and fattening CEO pockets & bonuses. They're not using that surplus to develop new games.



OK then. Let's get Disney to premier the new Alien: Earth series on HBO Max and Apple+ Day 1.

Oh wait...



I'm gonna keep it 100%: I don't know if SIE's 1P studios have what it takes to make a TLOU2 even this gen, let alone next gen. I mean purely on the graphical/immersion side of things and feeling like something TRULY pushing otherwise modest hardware to its absolute limits (TLOU2 is...well, a "special" case when it comes to the story....great game mechanics tho).

Games like TLOU2 were made during an era where SIE wasn't blinded by pursuing multiple platforms, meaning that tye of market outlook wasn't getting in the way of creative vision for their games. Which meant those games had even more reason to push limits because they ONLY had the PlayStation install base to depend on for sales, meaning they had to be showpieces for the console by necessity. Without the outlook of multiplatform (even if it's 1-2 years after initial release, a window that'll likely shrink as we've seen with Spiderman 2), those PS4-era 1P AAA games didn't have the associated safety nets. So they HAD to be system pushers even at a technical level, as that's all they had.

One example right there in how genuine exclusivity can be a strong motivator and benefit a game's design & market impact. Now I'm not saying games like Yotei look "bad". However, I also can't shake the fact it "feels" like a game with a very conservative/iterative design (even visually) because the studio knows they have to help in some capacity with a PC version within the year of its release. That's a whole bunch of configs to consider for scope, if you want acceptable performance, and that is going to impact how much a game can push the console's limits. Same with Intergalatic, and again, I think it looks fine visually. Do I think it'll push PS5 the way games like TLOU2 pushed PS4 visually? No.

1) if shareholders are unsatisfied Sony will lose investment and market share (recall the 10 billion loss due to eroding margins pre helldivers2).

2) your tv services comparison is not equivalent. However shows from 1 service appears on others to ensure an audience is found.

 

[Embearded]

I wont have same amount of ram (128b bus) and same and same cpu power at 15w ;d

Then in my opinion it will be a similar case as Series S, which is stupid.

 

[sncvsrtoip]

Then in my opinion it will be a similar case as Series S, which is stupid.

nah will be much worse, difference will be massivly bigger

 

[ExtendEmbraceExtinguish]

I don't know if cheaper guarantees a win. Series S sold more than X but not even close to what PS5 sold.
Maybe it's also the software, I don't know.

Also, i hope they will only port the GaaS stuff and let their first party studios focus on Sony consoles and provide the best possible games instead of "losing" time to optimize for S, Switch 2 etc.

Sony is the mainstream console to the point it is bought for Fortnite/cod/madden/fifa ect.. hence Sony problem is folks not buying ps exclusives vs 3rd party.

So a reasonable jump that is affordable will do wonders vs Msft prebuilt war machine for enthusiasts/pc couch potatoes. Series s is viewed as the poor man's console due to both ps5 & series x existing.

 

[Gameplay Gods Bless]

So, a well-made console that will still get new games 11-12 years after its launch?

What made the PS4 successful wasn't the anemic hw but the lack of competition. You had no choice but to buy bargain bin AMD consoles that gen, there were no other options.

 

[Gameplay Gods Bless]

I don't think these are specs. They are an early proposal in 2023 to Sony from AMD.

This is about as relevant as the ARM Xbox leaks from the FTC trial.

Wtf is your profile pic, looks like cell from dragon ball

 

[Gameplay Gods Bless]

Pro was for Sony fanboy gamers,(including me) marketed as a luxury upgrade.

PS5 Pro was a bad deal for everyone else, it certainly isn't bang for buck. If this gen has taught me anything, it's that we have been in a diminishing returns era for some time now and if you have a Pro, I see no reason to get a PS6 at launch because it's going to be another slow generation before we see the benefits in future games.

PS6 will definitely be much more powerful like the PS5 was, the jump from RDNA2 to RDNA5 (UDNA) is huge but like you said it's going to be wasted on cross gen, and multiplat games that will run on everything including PS5 and PS6 handheld.

 

[Gameplay Gods Bless]

It has to be cheaper than the pro or it won't make it.

How they manage to do this and make a large enough jump over the pro is going to be interesting to see.

It's simple PS5 Pro is still RDNA2 and Zen 2. PS6 will be RDNA5 and Zen 6 so it'll have several generations of microarchitectural improvements and it'll be on a better node (N3). It's similar to the PS4 vs PS3 fiasco where the PS4 wasn't that much more powerful (specially CPU wise), or the Xbox Series S vs PS4 where the Series S uses less power (80W vs 150W on PS4) and was cheaper on release but the Series S had more modern components that made up for its deficiencies.

 

[Tobimacoss]

It's simple PS5 Pro is still RDNA2 and Zen 2. PS6 will be RDNA5 and Zen 6 so it'll have several generations of microarchitectural improvements and it'll be on a better node (N3). It's similar to the PS4 vs PS3 fiasco where the PS4 wasn't that much more powerful (specially CPU wise), or the Xbox Series S vs PS4 where the Series S uses less power (80W vs 150W on PS4) and was cheaper on release but the Series S had more modern components that made up for its deficiencies.

PS5 Pro is Zen 2 and RDNA 3+

Xbox Ally X is RDNA 3.5

 

[Gameplay Gods Bless]

I think he made a good point. All Sony really has to do is maximize 4K 120 displays with better ray-tracing and that would be enough. Then you add an improved version of FSR4 in the mix and it will be a nice improvement overall.

Games will still run below 4k but be upscaled to 4k just as they have for the last decade. The target res doesn't matter to consumers what matters is the experiences.

 

[Gameplay Gods Bless]

Zen 6 + 30 TF + 3x better RT + GDDR7 for $600 will be a bargain even in 2027.

Remember, that the closest competitor will be the Switch 2 at 3 TF at max.

Lmao when there's no competition yea it'll look ok. But AMD hw is usually behind and not all that impressive, let shope RDNA5 beats Nvidia Rubin.

 

[Gameplay Gods Bless]

The one thing I don't believe is the 160W target. Since when Sony is interested into low power consumption in their consoles at the cost of performance? not when low power modes is definitely possible on PS6 too so that would satisfy their recent californian ideology.

PS5 Pro can consume up to 250W in some games and they didn't get sued by anybody because of this.

Lower power lets them cheap out on cooling reducing the cost of the console, like the Series S does. It'll also bring the PS6 console closer in power to the PS6 handheld.

 

[Gameplay Gods Bless]

And those devices does not have 5080-5090
MS would be lucky if they get 1% of those 90M for "extremely powerfull" Magnus chip


They are no longer in competition with Xbox
And Xbox price point will make any power comparison moot and irrelevant

Only high desktops have 5080 power, and only Nvidia provides it, not even AMD has 5080 power.

 

[Gameplay Gods Bless]

The 9070XT is three times faster than the PS5 (and has comparable GPU power to RTX4080). However, with only 40–48 CUs, it will be impossible to match the 9070XT's 64 CUs.

RDNA5 is going to run on a new node so there'll a die shrink, RDNA5 also will have a dramatically rearchitected microarchitecture much better than RDNA4. So RDNA5 is punching above it's weight on 2 dimensions while all RDNA4 had was an improved architecture and monolithic die.

 

[Parazels]

Lmao when there's no competition yea it'll look ok. But AMD hw is usually behind and not all that impressive, let shope RDNA5 beats Nvidia Rubin.

What's the Nvidia Rubin? Switch 2?

 

[Anchovie123]

Not too worried about the specs, FSR4 and frame gen will play a huge role and will be even more improved come 2027-2028

The jump to Zen 6 will be huge as well, here's Zen 3 to 5

 

[Gameplay Gods Bless]

What's the Nvidia Rubin? Switch 2?

RTX 60 series and likely Nvidia N2X chip which will go in laptops and PC handhelds.

 

[Gameplay Gods Bless]

PS5 Pro is Zen 2 and RDNA 3+

Xbox Ally X is RDNA 3.5

It is not, PS5 Pro is RDNA2 specifically RDNA2.X, it's just modified with a few features that came post RDNA2, but it is absolutely RDNA2. Part of the reason why the PS5 Pros performance leap is so anemic is because it's still using half a decade ol GPU tech at it heart.

PS5 Pro deep dive reaction: GPU and RT improvements, PSSR and Sony's new AMD Amethyst partnership

Digital Foundry breaks down the hardware and software of the PS5 Pro, including its GPU, RT improvements, PSSR and machine learning features, and next-gen.

www.eurogamer.net

 

[Loxus]

Lmao when there's no competition yea it'll look ok. But AMD hw is usually behind and not all that impressive, let shope RDNA5 beats Nvidia Rubin.

Have you not seen RDNA4 and FSR4 in action?
AMD stepped up big time.

They just need to release a top end card.

 

[Physiognomonics]

Lower power lets them cheap out on cooling reducing the cost of the console, like the Series S does. It'll also bring the PS6 console closer in power to the PS6 handheld.

They'll compete directly against more powerful Xbox and powerful PCs (PS6 will be a 3rd party machine first, they need a powerful hardware for those games). With PS5 they pushed the hardware in order to have a balanced and powerful gaming machine (maybe in response to XSX fixed 12TF GPU). They have to put out a powerful machine again, they can't aim for XSS low power innefficiency. How's that worked out for MS?

We are talking about a few dollars of plastic for the bigger console and a few dollars more copper for the heatsinks. There is very few copper right now on PS5 Pro, mainly cheap aluminium (and much less than on XSX, even at launch, few people know that). Their cooling solution is already the cheapest of the consoles as it uses volume of the heatsinks + cleverly designed heatsinks + liquid metal cooling (RD on this is already done).

This is their MO since PS5, they'll do the same here. They'll push the silicon to the max with their dynamic clocks. Besides, they'll add a low power consumption mode for people wanting to save electricity. Anybody stating otherwise has forgotten the 8-9TF drama mainly believed and pushed by Xbox fans for obvious reasons. I see the same reasons for pushing for a 160W PS6 now.

 

[Parazels]

RTX 60 series and likely Nvidia N2X chip which will go in laptops and PC handhelds.

Those insects (pc and pc-handhelds ) can't compete with the multi-billion-dollar console market.

 

[Crayon]

I appreciate that


Not yet

Kind of a stupid analogy but have a few rings of people I trust with info and the inner most circle is non repeatable but if anyone from the next ring starts whispering these same numbers then I know its more in the open and I am not going to burn someone

Odd way to put things I guess

Actually, that makes perfect sense.

 

[pacman4000]

RDNA5 also will have a dramatically rearchitected microarchitecture much better than RDNA4. So RDNA5 is punching above it's weight on 2 dimensions while all RDNA4 had was an improved architecture and monolithic die.

Source?

 

[peish]

Ok so here is the scope of what we know so far

Nexbox:
68 CUs (70) RDNA5 or RDNA4/5 mix
192bit bus, 12 memory channels most likely using 2gb clusters so 24gb GDDR7 RAM @32ggps giving us 768GB/s
CPU 11 cores Zen 6, 3 Big cores Zen 6, 8 smaller cores Zen6C
most likely using binned 12 core chips. As the setup will have a chiplet design GPU/CPU
SSD (Speculation) 2TB @+10gb/s
price most likely 599 but MS can eat the loss and sell for 499. BOM is about 600+
launch most likely 2026.

PS6:
40-48CUs RDNA5 (proper) @3Ghz
160-192bit bus 10-12 memory channels most likely using 2gb clusters so 20-24 GDDRR7 @28gpps giving us 640gb/s
CPU 8 cores Zen 6
SSD (speculation) 2TB 10gb/s
launch mid/late 2027

once again Sony has a sleeker simpler more efficient design and MS is going with more custom stuff. powerwise in terms of graphics very comparable. maybe slight advantage ps6. But nexbox has cpu advantage. Also xbox comes out a year before ps6 so its tech won't be as new as ps6. either way very close to power.

Thanks for summary for those lttp with next gen consoles talk

Cant believe both are downgrading from 256bit memory interface. And only 24GB, will not cut it for AI and RT. It is bust, damn shrinkflation and tsmc monopoly

What's more worrying is, despite MS disinterested with Xbox, they seem to target a better specs, and this time the gap is bigger than SX vs PS5… mark corny is losing his touch?

I can see a repeat of kinect, sony having to squeeze in a psportable in their budget and have to use middling specs for ps6…

 

[DinoD]

Because they aren't targeting the same resolution and frame rates? If the target for the PS6 console is 4K60 and 1080p30 for the handheld that's an 8x difference in pixels per second.

My only concern is that handheld CPU becomes the baseline for PS6 games if Sony mandates all games to be released on both. Limiting game logic, physics simulations etc. ?

 

[vaibhavpisal]

I think they will be fine if both versions have the same amount of RAM and same CPU.
Surely it will require extra work from the devs but not as much as Series S.

Apparently portable is using DDR 5 ram that was used in PS4. And is said to be main reason why ROG Ally isn't massively powerful than a steam deck.

It wont be same ram for sure.

 

[Crayon]

My only concern is that handheld CPU becomes the baseline for PS6 games if Sony mandates all games to be released on both. Limiting game logic, physics simulations etc. ?

It probably will and that sucks. But it's being compounded by 8gb gpus being produced to this day. The industry wide baseline is going to be well below even this conservative ps6.

 

[Gameplay Gods Bless]

Have you not seen RDNA4 and FSR4 in action?
AMD stepped up big time.

They just need to release a top end card.

Amd always does this, the problem is they're always playing catch up. Even with FSR they are behind DLSS: still no Ray reconstruction, still no ai frame gen, still no multi frame gen. Think about it this way AMDs brand new 2025 350mm² chip (9070xt) can't even beat Nvidias 370mm² 2022 chip (4080) in raster let alone at RT and it's even worse at PT.

Its never a question of will AMD progress enough to where Nvidia *used to be*, it's whether AMD will catch up where Nvidia is today. I'm sure RDNA5 will finally beat the 4090, but guess what the 4090 came out in 2022 not 2027. By then the 6090 will be out and guess what AMD is going to be behind again, they have been outpaced since Maxwell came out in 2014, it's been 11 years of this shit. Hopefully next gen is different but at this point the boy has cried wolf 1000 times so who's to believe him.

 

[Corporal.Hicks]

RDNA5 is going to run on a new node so there'll a die shrink, RDNA5 also will have a dramatically rearchitected microarchitecture much better than RDNA4. So RDNA5 is punching above it's weight on 2 dimensions while all RDNA4 had was an improved architecture and monolithic die.

The RDNA4 GPUs already use a 4nm process node. The 3nm process node mentioned by MLiD it's not a usuall full node shrink, so I'm assuming we cant expect usuall gains (if I'm wrong someone please correct me). However, there's indeed a theoretical possibility that RDNA5 architecture will offer significant power efficiency improvements.

Matching the RTX 5080 at just 160W would require the biggest architectural leap in AMD's history though just to match 9070XT performance (typical 305W power usage with 350W peaks), yet alone 5080. If AMD could match the performance of the RTX 5080 with only 160 watts, then just imagine what their high-end GPUs with 320-400 watts would do. AMD would not only compete with Nvidia, but crush them.

Guys, I'd like to believe that AMD can surpass our expectations. However, I still remember the dissapointment from PS5 leaks that suggested 12-14 TF. One of the leakers even contacted Neogaf moderators and showed them alleged photos of PS5 to verify his claims. Yet, when Sony made the official announcement, it became clear that the previous leaks were fake and had been created by people with a vivid imagination. For this reason, I just cant believe leakers anymore. I think it's more likely that AMD will continue their tradition of disappointing people. We should temper our expectations until Sony or AMD shares official information about the PS6.

 

[Gameplay Gods Bless]

They'll compete directly against more powerful Xbox and powerful PCs (PS6 will be a 3rd party machine first, they need a powerful hardware for those games). With PS5 they pushed the hardware in order to have a balanced and powerful gaming machine (maybe in response to XSX fixed 12TF GPU). They have to put out a powerful machine again, they can't aim for XSS low power innefficiency. How's that worked out for MS?

We are talking about a few dollars of plastic for the bigger console and a few dollars more copper for the heatsinks. There is very few copper right now on PS5 Pro, mainly cheap aluminium (and much less than on XSX, even at launch, few people know that). Their cooling solution is already the cheapest of the consoles as it uses volume of the heatsinks + cleverly designed heatsinks + liquid metal cooling (RD on this is already done).

This is their MO since PS5, they'll do the same here. They'll push the silicon to the max with their dynamic clocks. Besides, they'll add a low power consumption mode for people wanting to save electricity. Anybody stating otherwise has forgotten the 8-9TF drama mainly believed and pushed by Xbox fans for obvious reasons. I see the same reasons for pushing for a 160W PS6 now.

Yes your point of view makes sense but this time Sony believes Xbox will be more of a PC thing and therefore not direct competition as the only closed garden dedicated console maker will be Sony and Nintendo and Nintendo only makes low power handhelds.

Sony believes the dedicated home console market will be all theirs and only people interested in gaming PCs will buy a next gen Xbox so they don't have to push the PS6 as hard. That and they need a console that can scale down in cost better than the PS5 did so making something closer to a PS4 can help with that.

 

[Gameplay Gods Bless]

The RDNA4 GPUs already use a 4nm process node. The 3nm process node mentioned by MLiD it's not a usuall full node shrink, so I'm assuming we cant expect usuall gains (if I'm wrong someone please correct me). However, there's indeed a theoretical possibility that RDNA5 architecture will offer significant power efficiency improvements.

Matching the RTX 5080 at just 160W would require the biggest architectural leap in AMD's history though just to match 9070XT performance (typical 305W power usage with 350W peaks), yet alone 5080. If AMD could match the performance of the RTX 5080 with only 160 watts, then just imagine what their high-end GPUs with 320-400 watts would do. AMD would not only compete with Nvidia, but crush them.

Guys, I'd like to believe that AMD can surpass our expectations. However, I still remember the dissapointment from PS5 leaks that suggested 12-14 TF. One of the leakers even contacted Neogaf moderators and showed them alleged photos of PS5 to verify his claims. Yet, when Sony made the official announcement, it became clear that the previous leaks were fake and had been created by people with a vivid imagination. For this reason, I just can believe leakers anymore. I think it's more likely that AMD will continue their tradition of disappointing people. We should temper our expectations until Sony or AMD shares official information about the PS6.

No I agree with you (except N3 is a decent node just the early versions sucked). I don't for a second believe the PS6 will beat the 5080 in every way, the PS5 beat the Pascal Titan and matched the 1080ti in raster but beat it in RT and with support for newer features like it's version of mesh shaders. The PS5 couldn't beat the 2080 (launched 2 years before the PS5 ) at Raster, nor RT, nor AI (DLSS, Ray reconstruction) but it beat it in VRAM capacity (2080 only had 8GB and PS5 had 12.5 available for games).


So the PS6 should beat the 4080 and 5080 at some things, either feature set, maybe AI capabilities, maybe VRAM capacity but I don't think it'll beat the 5080 at everything it can do: raster, RT, path tracing, DLSS upscaling, Ray reconstruction, frame gen and multi frame gen, neural shaders etc.

I expect performance to be good and dramatically better at RT, PT and AI but I don't know that the PS6 will be as good vs Old Nvidia hardware like the PS5 was vs Nvidia Pascal (10 series), the PS6 would have to beat the 5090 for that. But it'll be a good machine and certainly better than a PS5 Pro in most ways.

 

[Corporal.Hicks]

The PS5 couldn't beat the 2080 (launched 2 years before the PS5 ) at Raster, nor RT, nor AI (DLSS, Ray reconstruction) but it beat it in VRAM capacity (2080 only had 8GB and PS5 had 12.5 available for games).

Remember that developers must divide the available RAM between the GPU and the CPU. If they would allocate 12.5GB just for GPU (VRAM) there would be nothing left for the CPU. My previous GPU had 8GB VRAM and it run the vast majority of PS5 ports (with at least PS5 graphics settings) without VRAM related problems (stutters / or memory leaks). I only played two PS5 ports where VRAM was clearly an issue: The Last of Us Part 1 and Forspoken. However, TLOU1 was patched soon after, and I could run it with high texture settings without any problems. Based on this experience I think developers on the PS5 must have been allocating between 7-10GB VRAM.

 

[winjer]

Apparently portable is using DDR 5 ram that was used in PS4. And is said to be main reason why ROG Ally isn't massively powerful than a steam deck.

It wont be same ram for sure.

That is so wrong.
The PS4 used GDDR5. Which is a memory optimized for memory bandwidth. Because of that, it's usually used for graphics cards.
The problem is that it uses more energy than normal DDR and has higher latency.

The PS6 portable is probably using LPDDR5. This is the current state of the art memory for mobile systems.
It uses lower power than normal DDR5, so it's great for a portable console.

 

[vaibhavpisal]

That is so wrong.
The PS4 used GDDR5. Which is a memory optimized for memory bandwidth. Because of that, it's usually used for graphics cards.
The problem is that it uses more energy than normal DDR and has higher latency.

The PS6 portable is probably using LPDDR5. This is the current state of the art memory for mobile systems.
It uses lower power than normal DDR5, so it's great for a portable console.

Yes, its not exactly the same RAM.

Point still remains. Its not same level of performance as DDR6X RAM used in current gen systems, let alone DDR7 found in upcoming consoles/GPUs.

Importantly, its considered a bottleneck in handheld systems.

 

[winjer]

Yes, its not exactly the same RAM.

Point still remains. Its not same level of performance as DDR6X RAM used in current gen systems, let alone DDR7 found in upcoming consoles/GPUs.

Importantly, its considered a bottleneck in handheld systems.

There is a lot of confusion there.
Although they are all DDR memory, and have a similar tech base, they are very different products. And each serves a specific need.
Just consider that a chip of GDRR6 can use up to 5W of power. So if we have several channels, which we always have, we can easily spend 20-50W of the system power, just on memory.
On a mobile system that kills the battery quite quickly. And GDDR6X, because it uses PAM, uses even more power.
Meanwhile, LPDRR5 will use around 200mw per chip. This is a fraction of what GDDR6 uses. Or even DDR5.
Usually, systems with LPDDR5 also use several memory channels, to make up some memory bandwidth.
But there is another thing to consider. Memory latency. And in this metric, the order of best is reversed, as the one that has the best one is DDR5, followed by LPDDR5 and in last place GDDR6.

 

[HJuggernaut]

Thanks for summary for those lttp with next gen consoles talk

Cant believe both are downgrading from 256bit memory interface. And only 24GB, will not cut it for AI and RT. It is bust, damn shrinkflation and tsmc monopoly

What's more worrying is, despite MS disinterested with Xbox, they seem to target a better specs, and this time the gap is bigger than SX vs PS5… mark corny is losing his touch?

I can see a repeat of kinect, sony having to squeeze in a psportable in their budget and have to use middling specs for ps6…

GDDR7 runs very hot, they had to for cooling. end of the day the only numbers that matter is pool ram and speed, devs dont touch bus width.