• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Sony CEO: "We have not yet decided on PS6 launch timing"

I'd actually welcome that approach with it just playing PS5 games on BC and streaming PS6 games. But given all the hardware features, it would be a total waste of RT and AI silicon. The CPU is certainly the biggest concern, especially given the TDP. Hopefully halving the framerate is all it takes for most games to work. I wouldn't be surprised if 30 fps stops being the floor, for this reason. If all PS6 games run 48 fps or above for quality mode (with 30 fps cap just for TVs that don't support it), then scaling down to handheld should be easy. Half the framerate target to 24 fps and then adjust a few setting to recover 25% so it can hit 30 again.
I'm sure all the engines are going to scale differently. Some will do better than others. Streaming the blockbuster stuff and dreamy specs would be one way to get my attention again as a customer. They've got the IP, the people and the motive to do it. Otherwise, for my tastes the tech ceiling is a little depressing for the next decade. Launch games should be in the polish stages of development so we will find out soon. Interested to see how they make the case to 5 Pro owners in these trying times. I think it's a tough sell when people are going to be breaking down "Intergalactic PS5 vs PS5 Pro vs PS6" in a million videos and it's hard to imagine the difference could be too drastic. I mean I guess they could shaft 5 Pro owners and withhold some effects it would be capable of running.
 
Last edited:
Sony should be more like Nintendo. Don't be scared of 26% or whatever Helix has. Releasing GTA6 and expecting everyone to just jump ship to PS6 seems like a very stupid move. Especially when the PS6 could be close to a grand. They'd have to have a multi tier support system for multiple consoles. That and support avenues for broken consoles. They'd probably go bankrupt if they released it too soon. Nintendo would need a big time Zelda and Mario game on their minimum spec console (or so we believe) and they'd nail the same Switch numbers.

What's becoming worrisome about consoles is that the refresh is becoming less and less appealing. The idea of a "any" box where you get years and years out of your device is a lot more appealing. The PC is basically that device.
Am I reading your post correctly in that you think Sony shouldn't release PS6 any time soon because Helix is 26% more powerful?
 
I don't want it to be that way. Multiplatform games are stuck targeting Switch 2 and PS6P for the next decade. The only chance we have at a big production going after an exciting tech target is if someone takes their shot at streaming.
If you can accept streaming you'll be able to stream gta6 from ps5 pro to ps portal at a much higher quality than any hypothetical ps portable could deliver.
 
I'm sure all the engines are going to scale differently. Some will do better than others. Streaming the blockbuster stuff and dreamy specs would be one way to get my attention again as a customer. They've got the IP, the people and the motive to do it. Otherwise, for my tastes the tech ceiling is a little depressing for the next decade. Launch games should be in the polish stages of development so we will find out soon. Interested to see how they make the case to 5 Pro owners in these trying times. I think it's a tough sell when people are going to be breaking down "Intergalactic PS5 vs PS5 Pro vs PS6" in a million videos and it's hard to imagine the difference could be too drastic. I mean I guess they could shaft 5 Pro owners and withhold some effects it would be capable of running.
If the below comparisons happen on launch games, it will sell pro owners. They are all graphics whores and/or have fat wallets anyway. The differences aren't drastic, but big enough to convince may be 10%.


gkksgik9sg4e1.gif


LG0AiP1ENJdsyp7G.gif


nBflR2Am1VxwgkE2.gif

54tvWmLKoqIY92Km.gif




The rest... yeah it is going to be a rough launch for sure, regardless of whether they nerf the hardware to save $100 or not. I don't think the PS6 will have a roaring launch like the PS5 even at $599. Consoles are starting to become like TVs, instead of phones. The majority just don't bother to upgrade until there is a pressing need for it or if it straight up dies on them.
 
Last edited:
If the below comparisons happen on launch games, it will sell pro owners. They are all graphics whores and/or have fat wallets anyway. The differences aren't drastic, but big enough to convince may be 10%.



The rest... yeah it is going to be a rough launch for sure, regardless of whether they nerf the hardware to save $100 or not. I don't think the PS6 will have a roaring launch like the PS5 even at $599. Consoles are starting to become like TVs, instead of phones. The majority just don't bother to upgrade until there is a pressing need for it or if it straight up dies on them.
It creates a situation for Pro owners, where Intergalactic is almost assuredly a PS6 launch title, that the promise they bought the 5 Pro for never really came to fruition, and finally here are the games you intended to play on your Pro and its old news. You become the stepchild even. PS5 is still the baseline they'll use for handheld and whatever else and PS6 will be the high watermark. Maybe that explains the lackluster Pro support we've seen from devs, they're putting that work into PS6 mode. If the Pro didn't exist this launch would be a lot easier. Maybe since it is the most enthusiastic we're talking about, they're still happy to buy a PS6 on launch day after having bought the PS5 twice so they could play the good, mid gen stream of first party games which seems to be lining up as the PS6 launch window lineup now. Wolverine, GTA6, Naughty Dog etc. SSM is firmly in PS6 territory at this point.
 
It creates a situation for Pro owners, where Intergalactic is almost assuredly a PS6 launch title, that the promise they bought the 5 Pro for never really came to fruition, and finally here are the games you intended to play on your Pro and its old news. You become the stepchild even. PS5 is still the baseline they'll use for handheld and whatever else and PS6 will be the high watermark. Maybe that explains the lackluster Pro support we've seen from devs, they're putting that work into PS6 mode. If the Pro didn't exist this launch would be a lot easier. Maybe since it is the most enthusiastic we're talking about, they're still happy to buy a PS6 on launch day after having bought the PS5 twice so they could play the good, mid gen stream of first party games which seems to be lining up as the PS6 launch window lineup now. Wolverine, GTA6, Naughty Dog etc. SSM is firmly in PS6 territory at this point.
The situation certainly applies to me, so I agree. I'm likely holding off on all of those titles until the PS6 comes out, including GTA 6 :messenger_grinning_smiling:
 
Logic dictates it will.
"Logic" = Hopium.

Memory hasn't become significantly more expensive to produce overnight
Only Partially true, as SKHynix is paying their employees 1M bonuses and they are paying for expedited EUV shipments. Their production costs did increase.

the reason pricing has shot through the roof is that it is supplied constrained
The supply is constrained because demand blew up. AI Memory demand is far behind supply and it's forcing consumers to crunch.

But again the existence of the bimodal market, where they can take consumer supply to AI and vice versa, allows the memory cartel to maintain consumer pricing in the correct band where it's not too expensive for consumers but is very profitable for the cartel. 8-12$/GB is just about right. The consumer side of the market can forever be "supply constrained".

And supply will eventually be fixed, be it via ramping production, scaling back on demand, increased competition, regulatory pressure, or a combination, it will price correct.
Again what you still don't get is that supply can = demand at a different fair market price.

When AI came, the net effect is that the demand curve shifted to the right. Meaning that the fair market price for memory has been permanently and irrevocably increased.

Bozo VRAM maxing is like a religion at Radeon. Which is why Kepler is coping so much that it will become possible again. This memory shortage vindicated Jensen's memory cautious approach and prioritization of software and hardware features over commodities.
 
Last edited:
The rumored specs apparently reach 5080 levels, but I have a hard time believing that. Not when a 5080 still costs $1300 at the low end. There's no way Sony will be able to get that performance even if their system itself costs $1000 new.
 
Memory is largely expected to be pressured cost-wise until well into 2027. Best case is a 2027 holiday release… likely 2028.
 
"Logic" = Hopium.


Only Partially true, as SKHynix is paying their employees 1M bonuses and they are paying for expedited EUV shipments. Their production costs did increase.


The supply is constrained because demand blew up. AI Memory demand is far behind supply and it's forcing consumers to crunch.

But again the existence of the bimodal market, where they can take consumer supply to AI and vice versa, allows the memory cartel to maintain consumer pricing in the correct band where it's not too expensive for consumers but is very profitable for the cartel. 8-12$/GB is just about right. The consumer side of the market can forever be "supply constrained".


Again what you still don't get is that supply can = demand at a different fair market price.

When AI came, the net effect is that the demand curve shifted to the right. Meaning that the fair market price for memory has been permanently and irrevocably increased.

Bozo VRAM maxing is like a religion at Radeon. Which is why Kepler is coping so much that it will become possible again. This memory shortage vindicated Jensen's memory cautious approach and prioritization of software and hardware features over commodities.
You're acting at if this is the first time RAM prices have doubled or even tripled. It's famously known as being cyclical and volatile.

As I said pricing is surging due to demand outpacing supply by a considerable margin, not due to a fundamental shift in production costs overnight. Supply will eventually increase. Staff bonuses easily fluctuate depending on profitability and once profits pull back due to expenditure on scaling out manufacturing you'd see those same bonuses drop.

RAM prices are unlikely to come back to pre 2025 prices in the short term, but pretending that the current status quo will remain that forever is the real hopium.
 
Last edited:
Your issue here is that yu do not understand what the tech is actually being used for, or more so, what kinda tech you need to achieve certain things.
I know what it's being used for. What you don't understand is that pound for pound, dollar for dollar, you will get much more power from the Orion-based PS6 than the handheld. The handheld is devised to be a low power machine, watt for watt, the Orion console can't come close.

Having that as premise, it doesn't make sense to leave the next-gen saturation to the handheld device that is even less powerful than a PS5 in raster, when you have an Orion console that is leaps and bounds more powerful by just being $100 dollars more.

Canis makes sense to exist as a handheld and as a complementary to a moderately priced PS6 home console.

Lets look at this from the two things the PS6 will address. Lets call this device A and device B

A console capable of doing native 1080p-1440p and upscaling that to 4K plugged into a living room TV. This console would run PS6 games at the highest possible fidelity and will offer the overall best experience.
That's not gonna happen with just 30GB of RAM vs 20GB of RAM. You get the best fidelity on PC with everything maxed, but it's not the RAM that takes you there, it's the 500watt GPUs.

You will get the same game with the 20GB PS6 but with less uncompressed audio and texture. New compression techniques will be the new craze.

You won't even need to think of what ifs because even PC GPUs will be conservative when it comes to RAM. Only the $2000 GPUs will have 32GBs.

To ensure that this console can keep doing that for the next 10 years and not become outdated, you need minum, 4Ghz CPU, 3 GHz 56 CU GPU, 160-bit bus, 30GB RAM, which means around 26GB available for devs/games, at least 1TB SSD....etc.
No, you don't need that as a minimum. Canis handheld already exist.

PS6 Pro can exist with 30GB or more for those who want it. Up to Sony. But I think they should do it.

This is your baseline for the next generation. This baseline will cost the consumer around $600-$800. This is device A.
The baseline has to be cheap enough. Sony don't need to absorb unnecessary losses when they can cut the BoM cost by reducing memory components.

The Pro can go bonkers with specs, price doesn't matter.

Now the second thing..

You need a device that can run what device A can run, but instead of targeting lower settings, lower peak framerates, and lower native resolutions. So now you are trying to run games at 360p-540p native and upscale that to 1080p.
The baseline PS6 can also target lower settings just like this low power, low raster, low watt, low bandwidth device.

And you get much more power, raster, path tracing, and pixels pound for pound dollar for dollar in the Orion PS6.

Preferably, primarily used on a 7-9" handheld screen, but giving the option to plug it into a TV. To do this, you need a much smaller GPU, less bandwidth, less RAM, and when you tie all those things together, it will fit into a handheld form factor. This device will cost consumers anywhere betwen $400-$500. This is device B. (to help you with some context... 360p is 16x less than 1440p, some napkin comparison would tell you that the PS6 handheld could technically be about 8-16times weaker than a home console GPU and still be able to handle PS6 games... if they make the right decisions in the right areas).
The handheld is a great idea. But it should not replace the idea of an affordable PS6 console set top box.

This is where you should start making your analysis from.

A "proper" PS6, costs what it costs because that's what it costs to make a proper next-gen console today; you do not cut back on that in such a way that you are releasing an underpowered, incapable machine because you are trying to save a few bucks. The best thing to do is to make two separate products that are in the same family. Hence a home console... and then a handheld. If you do this right, one of them is not holding back the other. You make a product for the people willing to spend no more than $500, and you make a product for the people willing to spend as much as $800.

I am adding this part because I feel I may need to explain everything and trying to avoid that... the right areas are that both PS6 and PS6H are based off the same architecture and have an identical feature set, PS6h will have RAM close to the PS6 in capacity and bandwidth at about a third to half in speed.


The switch has shown that you do not need a 30GB $800 console to sell 100M.

This is the one crucial detail you are ignoring and likely explains why, in all this stuff, you say you talk like the PS6h doesn't exist.

It does, and its for a reason. The PS5 handheld addressing a lower budget, 1080p market, is the sole reason a proper PS6 in the traditional sense can exist in this day and age.
PS6 Pro can exist to fulfill our ultimate PS6 wetdream. But an affordable baseline PS6 should exist before that.

The handheld can't be that "baseline" device because dollar for dollar it doesn't make sense.
 
Last edited:
Don't touch the original PS6 specs, rather, offer the X/S variants like Xbox did this gen. PS6 S can be something in between the PS6 handheld and the full-on PS6.
Wrong. PS6 APU is already very conservative. What you want is this current PS6 specs to be the Series S and another PS6 Pro to be the Series X.

The current PS6 specs can afford to have 20GB of RAM. There are far more GPUs existing and will exist in the future that has more raster, RT, and ML capability than Orion APU but with relatively conservative amount of RAM. PC will not go crazy with VRAM capacity with the current and future pricing. They will follow PS6 footsteps. And devs will have to adjust.

Lower that native res, reconstruct to 4k and you'll be fine.
 
"Intergalactic PS5 vs PS5 Pro vs PS6" in a million videos and it's hard to imagine the difference could be too drastic.
I think the gain in path tracing will translate well in youtube videos. Much more than the zoomed cleaner image of PS5 Pro vs PS5.
 
You're acting at if this is the first time RAM prices have doubled or even tripled. It's famously known as being cyclical and volatile.
"It always went down when it went up, thus surely it will go down again".

Nah, it's not possible that this isn't anything remotely like any of the last cycles. The Semiconductor industry will remain the same despite going from a smol industry to a massive multi trillion dollar one. 🤡

due to a fundamental shift in production costs overnight. Supply will eventually increase
How hard is it to understand that when demand increases, the fair market price (price where supply = demand) increases?
 
Last edited:
PS5 sales are winding down hard. Even the price increase didn't bring in people on the fence, not even with GTA6 on the horizon. Sony will be launching a PS6 sooner rather than later. They don't really have a choice.
 
"It always went down when it went up, thus surely it will go down again".

Nah, it's not possible that this isn't anything remotely like any of the last cycles. The Semiconductor industry will remain the same despite going from a smol industry to a massive multi trillion dollar one. 🤡


How hard is it to understand that when demand increases, the fair market price (price where supply = demand) increases?
How hard is it to understand that the price increase is because demand vastly outstrips supply? That's it. That's the only reason.

Supply increases and you'll find prices dropping again. Why do you think the big three are hesitant about spending billions on building new fab capacity? Reason one, they know there is a strong possibility demand will drop and they will be caught in the situation (again) where they are spending billions on new fabs and yet revenue falls, and the second reason is that increasing supply to match the surging demand basically means prices will fall and they won't have such obscene profits.

You seem to be under the impression these costs increases are for some other reason than the simple truth, it's a supply and demand issue and nothing more.
 
Last edited:
PS5 sales are winding down hard. Even the price increase didn't bring in people on the fence, not even with GTA6 on the horizon. Sony will be launching a PS6 sooner rather than later. They don't really have a choice.
I think your predictions are wrong.

The moment the GTAVI gameplay trailer for the base PlayStation 5 is released and it's confirmed that the PlayStation 5 Pro version will have a 60fps option, sales will skyrocket.

It's the most anticipated game for a huge number of people.

Sony will be collecting monthly revenue due to the increased number of people on PSN+ playing GTA VI online, plus transaction fees.

Technically, this game will boost Sony's coffers.
 
Last edited:
Wrong. PS6 APU is already very conservative. What you want is this current PS6 specs to be the Series S and another PS6 Pro to be the Series X.

The current PS6 specs can afford to have 20GB of RAM. There are far more GPUs existing and will exist in the future that has more raster, RT, and ML capability than Orion APU but with relatively conservative amount of RAM. PC will not go crazy with VRAM capacity with the current and future pricing. They will follow PS6 footsteps. And devs will have to adjust.

Lower that native res, reconstruct to 4k and you'll be fine.

PS6 needs that next gen jump and it can't be down to having a visual edge over PS5, not when you charge close to 1000€ for it.

Having a native LLM implementation could do wonders for interacting with NPCs, player immersion and world building, or sports games that have genuinely dynamic commentary that never repeats - this is the type of next gen advancement people could get excited about and consider investing in such a system despite its high price.

LLMs are very memory-heavy and since this is a console that would last until mid 30s, Sony should not tamper with reduced RAM.
 
"It always went down when it went up, thus surely it will go down again".

Nah, it's not possible that this isn't anything remotely like any of the last cycles. The Semiconductor industry will remain the same despite going from a smol industry to a massive multi trillion dollar one. 🤡


How hard is it to understand that when demand increases, the fair market price (price where supply = demand) increases?

I agree and I don't see memory prices ever going down in any significant way.

This debate actually comes down to a question of whether a person is a AI skeptic or AI optimist in terms of future AI capabilities.

Skeptics argue that the current AI paradigm will saturate long before it can do most of white collar jobs.

AI optimists believe that AI models will gradually take over white collar jobs - and if that is the case, then things like memory and the whole semiconductor industry will change its role from being a human tool (or materials of which the tool is made from) to an active agent that does the work and replaces human labor - it's a whole paradigm shift in economics in which capital starts to replace labor.
 
As long as Sony can make a box that can do path tracing then take the time. Even if its 40FPS path tracing with framegen to 60+ vrr then that should be fine for most people.
 
PS5 sales are winding down hard. Even the price increase didn't bring in people on the fence, not even with GTA6 on the horizon. Sony will be launching a PS6 sooner rather than later. They don't really have a choice.
can't wait to see the price/specs. it's either going to be super expensive or expensive and low powered.

you seen the price of 32GB RAM and 2TB ssds these days? that has to be the minimum for a PS6.

sony are going to need start making games to convince people to cough up the money for the PS6.
 
How hard is it to understand that the price increase is because demand vastly outstrips supply? That's it. That's the only reason.

Supply increases and you'll find prices dropping again. Why do you think the big three are hesitant about spending billions on building new fab capacity? Reason one, they know there is a strong possibility demand will drop and they will be caught in the situation (again) where they are spending billions on new fabs and yet revenue falls, and the second reason is that increasing supply to match the surging demand basically means prices will fall and they won't have such obscene profits.

You seem to be under the impression these costs increases are for some other reason than the simple truth, it's a supply and demand issue and nothing more.
Thank you, this is what I was trying to tell them exactly as well.
 
I agree and I don't see memory prices ever going down in any significant way.

This debate actually comes down to a question of whether a person is a AI skeptic or AI optimist in terms of future AI capabilities.

Skeptics argue that the current AI paradigm will saturate long before it can do most of white collar jobs.

AI optimists believe that AI models will gradually take over white collar jobs - and if that is the case, then things like memory and the whole semiconductor industry will change its role from being a human tool (or materials of which the tool is made from) to an active agent that does the work and replaces human labor - it's a whole paradigm shift in economics in which capital starts to replace labor.
Being an AI optimist or pessimist doesnt determine the economics of demand and supply. You have a point that AI pessimists are banking on AI hitting a wall which would reduce the amount of memory demand. On the other hand I am an AI optimist, I think there is need more memory, larger memory and storage capacity, so we can have smarter longer term memory, easier training runs, and ability to store more data from the real world or generate larger datasets for RLHF. But that doesnt mean manufacturers may not meet or outstrip memory and storage demand over the next 5-10 years. Even if current AI systems continue growing in leaps and bounds, from a technical hw perspective there are only 2 constants

1.) we should be getting better and cheaper memory per storage capacity over the long horizon. The cost of a TB of memory in 2000 compared to a TB of memory in 2026 compared to a TB of memory in 2052 is going to go down whether you're an AI optimist or pessimist.

2.) The second constant is there will always be market up and downs especially in commodities like memory and storage.

When you combine these two its really not rocket science to realize we are in a memory market upcycle thats eventually going to come down even meaningfully. Even if memory prices continuegrowing over the short term, they must come down. Either new entrants into the market will increase supply(CXMT, YMTC) or support industries will make it much easier for current companies to produce enough memory to meet demand.
 
PS5 sales are winding down hard. Even the price increase didn't bring in people on the fence, not even with GTA6 on the horizon. Sony will be launching a PS6 sooner rather than later. They don't really have a choice.
I don't get the logic. Sales are down because price for ps5 have gone up. So they're supposed to release even more expensive hardware?

Confused High Quality GIF
 
That's only part of it

Co3Ynr9g5zCTc6n1.png
You misunderstand the chart.

The chart is more about willingness to buy and sell. What the line is saying what quantity of memory will be supplied and demanded at any given price. the point the two lines meet is market clear price. The demand curve shifted to the right but the supply curve is still the same one and if anything shifted left. (their production costs are up)
 
Last edited:
Yes, there is now structurally more demand for memory than before, that is not up for debate. But that doesn't in of itself mean memory prices will never come down. That would only hold true if demand keeps growing faster than supply forever. In reality it won't, and markets will adapt.

The price floor in the future will be higher than pre-2025, but it's simply not true that prices of memory won't come down from where they are now - in the long term, assuming supply increases. You can pretend all you want, but the DRAM market is not a perfectly competitive commodity market.
 
Last edited:
You misunderstand the chart.

The chart is more about willingness to buy and sell. What the line is saying what quantity of memory will be supplied and demanded at any given price. the point the two lines meet is market clear price. The demand curve shifted to the right but the supply curve is still the same one and if anything shifted left. (their production costs are up)

No, I understand supply and demand just fine. Just doesn't make sense to provide one scenario as if it is the only scenario. Demand isn't only ever going to shift one way, in other words.
 
Last edited:
I don't get the logic. Sales are down because price for ps5 have gone up. So they're supposed to release even more expensive hardware?

Confused High Quality GIF
No. Sales are down. The price increase was announced and even with people getting a couple weeks to buy before it went up, they sold very little. This means the PS5 is at the end of it's lifecycle. There are no more buyers willing to jump in. The price increase makes it even less likely to bring in new gamers.

This is not exclusively the same group as potential PS6 buyers. I am not sure why you think it is. Price is going to be an issue for every platform. Waiting doesn't change that, it just misses the people who are interested in a new gaming system and end up choosing what is available which could be Switch 2, PC, or Helix. What does waiting do? No guarantee component costs will decrease, your current platform is not selling and isn't creating much hype for your business. So you sit down and do nothing? And you do that with new products designed and nearly ready to be manufactured. OK.
 
PS6 needs that next gen jump and it can't be down to having a visual edge over PS5, not when you charge close to 1000€ for it.

Having a native LLM implementation could do wonders for interacting with NPCs, player immersion and world building, or sports games that have genuinely dynamic commentary that never repeats - this is the type of next gen advancement people could get excited about and consider investing in such a system despite its high price.

LLMs are very memory-heavy and since this is a console that would last until mid 30s, Sony should not tamper with reduced RAM.
My whole premise for advocating for a reduced RAM in PS6 is that it's not necessary and that 20GB PS6 can achieve what an RTX 5080 16GB can achieve.

But if LLM is really on the cards for next-gen and an extra 10GB is necessary for that as opposed to just being an anti-PC marketing shock, then by all means, Sony should do it. I want new experiences for next-gen.

But going back to your previous suggestion of a PS6s with power in between PS6 handheld and Orion PS6, I don't think that would be worth it considering Orion APU is already very conservative at only 160W and estimated to cost $110. You won't get anymore meaningful savings there if you pair it with another 30GB VRAM because it is "necessary".

PS6 handheld with cheap ass low watt LPDDR RAM sounds the best way forward for a cheap next-gen console. And even beyond PS6, LPDDR paired with low watt apu sounds the best way forward for an entry level playstation.

That is, of course, if we truly need this much amount of RAM.
 
Last edited:
A "new business model" may mean they will release the PS6 on schedule, whenever that is, rather than delay it but that the price will reflect what it costs to make at the time so with the current high pricing on RAM and storage that could mean the console will be expensive. Maybe, even more expensive than the PS5 Pro is now.

I don't see what else it could mean unless they were going to release the PS6 Handheld and PS6 Console together and that they will now stagger them. The handheld is likely to be expensive anyway so maybe that will be delayed? I would certainly rather have the handheld delayed over the console as I do not have any need for a handheld PS6 personally. Also, I suspect that most people will want the full-fat PS5 console rather than a cut-down handheld.
 
Last edited:
Wanting to release PS6 sooner in 2027 with underwhelm specs (and some expecting PS6 equal to RTX 5070 LMAO) is kind of absurd. Instead, why don't we expect they release in 2028 2029 with far better hardware and solid software first party and exclusives line up?
 
Well it will not be equal in term of specs and raw power, but it will be close just like how base PS5 equal to RTX 2070 Super and closer to RTX 2080 in terms of raster (non RT) with "universal decompression tech (increasing bandwidth) "and 30 GB RAM (2X of RTX 5080 RAM) and software optimization.
 
Last edited:
If the below comparisons happen on launch games, it will sell pro owners. They are all graphics whores and/or have fat wallets anyway. The differences aren't drastic, but big enough to convince may be 10%.


gkksgik9sg4e1.gif


LG0AiP1ENJdsyp7G.gif


nBflR2Am1VxwgkE2.gif

54tvWmLKoqIY92Km.gif




The rest... yeah it is going to be a rough launch for sure, regardless of whether they nerf the hardware to save $100 or not. I don't think the PS6 will have a roaring launch like the PS5 even at $599. Consoles are starting to become like TVs, instead of phones. The majority just don't bother to upgrade until there is a pressing need for it or if it straight up dies on them.
PS5 Pro are able to run F1 EA 30 fps with PT at 1080p on GDC demo. PS6 is doubling in raster of PS5 Pro (3X PS5 base raster raw according to rumor). So others improvement will be likely on tensor capability with will be 2 to 3X of PS5 Pro tensor capability (with the help of neural arrays) and RT equal to RTX 4080 Super. So I can see PS6 is RTX 5070 Ti to RTX 4080 Super level in term of raw specs.
 
Last edited:
But that doesn't in itself memory prices will never come down.
I'm not making the argument that prices will never come down. It's more nuanced than that. I'm saying things will never go back to 2023-2024 levels or to the old memory market. You're misunderstanding two things:

1. 2024 wasn't "normal." It was a bloodbath. Samsung and SK Hynix were operating at massive losses from a post-COVID/Mining glut. At the same time, state-subsidized Chinese CXMT flooded the market with DDR4 below cost. Memory was structurally underpriced. That's the market that let Sony, a minor memory customer, get $2.9/GB on GDDR6-18Gbps and think 30GB of GDDR7 wasn't madness.

2. Every previous memory cycle had the same demand driver: humans buying devices. PCs in the 90s. Smartphones in the 2010s. Laptops during COVID. The endpoints are fixed, and cheaper memory only minimally increases sales because consumers see memory as a peripheral component. Nobody buys a second phone because memory got cheap, and cheap memory doesn't meaningfully change the endpoint price.

The AI inference market is completely different. Memory is a binding constraint. Capacity determines what model size you can run. Speed determines how many customers you can serve. The AI market treats memory like energy. With oil, lower prices drive higher consumption. Same logic applies here: if memory costs drop and inference gets cheaper it reliably leads to more consumption. More applications become economically viable. More agents get deployed. More models get served. More context windows get extended. Demand for intelligence/Automation, like demand for energy, is very elastic to price declines.

On top of that, the AI market is now bigger than the consumer market and acts as a permanent balancing sink. The cartel isn't offering multi-year agreements anymore, it's 3 and 6 month contracts. You want 12 months you pay in advance. They continuously pull supply in and out of the client side into the AI side, meaning if consumer demand collapses, consumer supply will too. And on the AI side, more supply reliably and elastically creates more demand, which suppresses the price drop of getting more supply.

The existence of a the now bigger AI memory market will allow the cartel to change the behavior of the boom and bust consumer memory market. They can trivially adjust supply between the AI market and the consumer market to defend a price band. KeplerL2's case for memory prices to go back to "normal" is a market crash so severe it's worse than 2008. Yes, that's possible.

Above all, the primary argument I am making is that this is completely uncharted territory. Again "GUARANTEE" is the key word when I said:
THERE IS NO GUARANTEE MEMORY PRICES AND AVAILABILITY WILL EVER RETURN TO 2024 LEVELS. NOT IN 2027, 2028 OR EVEN 2035. NOSTALGIA IS NOT A STRATEGY.

To me the only way to break out is by making changes on the Supply side that can substantially decrease the production costs. Thus 60%+ Op margins for cartel can be met at a lower price. I don't see how the cartel can make TSMC sized CAPEX investments without TSMC sized returns on invested capital.
 
Last edited:
I'm not making the argument that prices will never come down.

You did though

"When AI came, the net effect is that the demand curve shifted to the right. Meaning that the fair market price for memory has been permanently and irrevocably increased."
 
Top Bottom