Support NeoGAF

[saintjules]

Somewhat off-topic. Put this latest page in Google Gemini Pro and it's able to read all 17 pages of this thread and give me a summary.

 

[MCH2024]

- AT4/24CU for 12GB LPDDR5X RX 10060 ($300) and 16GB LPDDR5X RX 10060XT ($350)

- AT3/48CU 24GB LPDDR5X RX 10070 ($550)

- AT2/68CU 18GB GDDR7 RX 10070XT ($700)

- AT0/96CU 36GB GDDR7 RX 10090XT ($1200)

AT3/4 dGPUs won't exist outside of AI. 48/44CU SKUs are AT2.

Nobody would want them either. Only actual angle with them is AI inference really.

You are mixing things up. AT0 has 192 CUs or 96 WGPs depending on how you count them.

 

[MCH2024]

No, they are not. While possible, they are avoided for a reason. And even when used, they are used within the same family of RAM, eg DDR5/LPDDR5... and never when combining RAM of different families i.e DDR + GDDR.

They are a thing. And you can have different memory controllers within the same die. 160b GDDR7 bus + 32b LPDDR5X bus was the right thing to do.

 

[MCH2024]

We also happen to already know SKU List.

PTX 1080: 154CU, 384b, 36GB, 380W
PTX 1070 XT: 64CU, 192b, 18GB, 275W
PTX 1070: 48CU, 160b, 15GB, 235W
PTX 1070 GRE: 44CU, 128b, 12GB, 215W

 

[Physiognomonics]

No one has done die shot analysis of Viola so...

I first thought they could have used the I/O complex bandwidth like they similarly did on PS4 Pro with the southbridge limited bandwidth.

But this time the DDR5 is at the opposite side of the I/O complex. So it had to be some DDR PHYs, right?

 

[Shogmaster]

AT3/4 dGPUs won't exist outside of AI. 48/44CU SKUs are AT2.

Nobody would want them either. Only actual angle with them is AI inference really.

You are mixing things up. AT0 has 192 CUs or 96 WGPs depending on how you count them.

I know you are going by that initial MLiD RDNA 5 leak, but if AT2 is binned down to 48/44 CUs from 70 in enough quantities to drive whole segments, that means TSMC 3nm has gone extremely wrong. LOL

And that chart is clearly outdated since Magnus has 68 CUs from 70, yet it lists only 64 CUs for AT2. I seriously doubt AT2 in Magnus will have more CUs than the AT2 in discrete cards.

 

[Loxus]

You gonna have to explain the whole PS5 Pro RAM set up for us plebs then.

Here you can see the DDR module in the red rectangle. It was done during the chip development.

The PS6 chip is done. You can't just go back and add a DDR PHY without serious layout modifications, just to have less GDDR size. It's a waste of R&D.

 

[Stafford]

Isn't this a shit source tho?

 

[Sethbacca]

The only thing that worries me is backwards compatibility: I really hope that all PS6 consoles are 100% backwards compatible with the PS5 (and PS4).

I think that's a foregone conclusion. At this point they've got enough power to emulate every generation of Playstation (and they should).

 

[Shogmaster]

Here you can see the DDR module in the red rectangle. It was done during the chip development.

The PS6 chip is done. You can't just go back and add a DDR PHY without serious layout modifications, just to have less GDDR size. It's a waste of R&D.

I already saw that imagebwhen they did the teardown. But that Pic dont explain crap.

The SoC has 256bit bus. Thats 8 GDDR6 RAM modules with 32bit bus each. WHERE DA FUG DOES THAT 2GB DDR5 GO?!? Its waaaay off to the side! Which part of bus is it on? Is it sharing a 32bit channel with one of the GDDR6 modules? How the fuck does that work?

SOME BIG BRAIN EXPLAIN IT TO MEEEEEEEE~

 

[Loxus]

For PCIe x16 cards, I was guessing:

- AT4/24CU for 12GB LPDDR5X RX 10060 ($300) and 16GB LPDDR5X RX 10060XT ($350)

- AT3/48CU 24GB LPDDR5X RX 10070 ($550)

- AT2/68CU 18GB GDDR7 RX 10070XT ($700)

- AT0/96CU 36GB GDDR7 RX 10090XT ($1200)

Well, according to MLiD, Destop AT2 has 64CUs.

I don't know if that is the full die and desktop AT2 is somehow different from the AT2 in Magnus. This is based on 64CUs being AT2 XT in MLiD image, XT always seems to be the full die.

96CUs is only possible as AT1.
AT0 has 154-184CUs, possibly as 2 AT1s. (This is just my guess)

 

[Shogmaster]

Well, according to MLiD, Destop AT2 has 64CUs.

I don't know if that is the full die and desktop AT2 is somehow different from the AT2 in Magnus. This is based on 64CUs being AT2 XT in MLiD image, XT always seems to be the full die.

96CUs is only possible as AT1.
AT0 has 154-184CUs, possibly as 2 AT1s. (This is just my guess)

I already addressed this brah. Keep up.

 

[Exede]

Up to 550 and its called budget Model is ahm yes stupid.
Like the budget idea is stupid as a whole!

 

[Xyphie]

I already saw that imagebwhen they did the teardown. But that Pic dont explain crap.

The SoC has 256bit bus. Thats 8 GDDR6 RAM modules with 32bit bus each. WHERE DA FUG DOES THAT 2GB DDR5 GO?!? Its waaaay off to the side! Which part of bus is it on? Is it sharing a 32bit channel with one of the GDDR6 modules? How the fuck does that work?

SOME BIG BRAIN EXPLAIN IT TO MEEEEEEEE~

Presumably there's a DDR5 PHY, but only 16-bit wide (instead of 128-bit which you'd see on a typical dual-channel CPU) as it only connects to one chip and thus very slow. iFixit has the chip ID as H5CG46AGBDX017N, which is a 5600MT/s 16-bit DDR5 chip. So we even know the bandwidth from that 5600 * 16/8 = ~11GB/s. Take one of the 32-bit PHYs from Strix Point and cut it in half and it would be about that size in terms of die area.

 

[Shogmaster]

Presumably there's a DDR5 PHY, but only 16-bit wide (instead of 128-bit which you'd see on a typical dual-channel CPU) as it only connects to one chip and thus very slow. iFixit has the chip ID as H5CG46AGBDX017N, which is a 5600MT/s 16-bit DDR5 chip. So we even know the bandwidth from that 5600 * 16/8 = ~11GB/s. Take one of the 32-bit PHYs from Strix Point and cut it in half and it would be about that size in terms of die area.

That "16bit" part is prolly talking about 16Mbit DDR5 module, which would make it a 2GB module. 16bit bus dont make no damn sense.

 

[Xyphie]

That "16bit" part is prolly talking about 16Mbit DDR5 module, which would make it a 2GB module. 16bit bus dont make no damn sense.

No, it's the width of the chip.

A typical non-ECC DDR5 stick has either 4 or 8 DRAM chips.

If you have 8 chips they are 8x chips = 8 x 8 = 64-bit. Two sticks = 128-bit.

On a 8GB DDR5 stick since there are no 1GB DDR5 chips they run 4x 2GB chips with 16-bit width. 4 x 16 = 64-bit. Two sticks = 128-bit.

 

[MCH2024]

I know you are going by that initial MLiD RDNA 5 leak, but if AT2 is binned down to 48/44 CUs from 70 in enough quantities to drive whole segments, that means TSMC 3nm has gone extremely wrong. LOL

And that chart is clearly outdated since Magnus has 68 CUs from 70, yet it lists only 64 CUs for AT2. I seriously doubt AT2 in Magnus will have more CUs than the AT2 in discrete cards.

It's not outdated. You are coping.

Even with 70 CUs, they can still release the top DT chip at 64CU, and if they're front-end limited like 9070/9070XT, it barely matters anyway. Most volume is Magnus regardless. Given putrid RDNA4 dGPU sales, betting on RDNA5 discrete driving good numbers is optimistic.

The binning also makes sense when you look at it properly. 9070 only had tolerance for CU defects, any SE/MEM/L3 defect made it a 9070 GRE. For AT2, 64 CU tolerates CU defects; 48CU tolerates CU/SE/MEM/L2 defects; 44CU covers rare cases where two memory controllers or two different shader engines are defective, which tracks with it being a China-only SKU.

And just compare the 48CU configs across dies:
- AT2: 48CU, 96 ROPs, 20MB L2, 15GB GDDR7 (720 GB/s)
- AT3: 48CU, 64 ROPs, 10MB L2, 32GB LPDDR6 (455 GB/s)

AT3 SKU is more expensive and way worse. AT2 strategy is similar to RDNA4, just better optimized for cost since the 18/15/12GB VRAM spread meaningfully cuts BoM on lower SKUs too.

And as for AT4 dGPU, the idea is like the dumb AIBs that thought there was a market for 6500XT 8GB. The only market where low performance high memory is good is AI.

Lastly, N3P is sitting at <0.04 D0 with god-tier device variance. None of this is yield-driven. It's pure artificial segmentation where pricing is to value and not to cost. AMD is ruthlessly optimizing the lineup for ASP, Marketshare and COGs. If that requires binning perfectly fine chips? Well tough shit.

It's also just the norm now. Qualcomm's doing the same thing with X2 Elite where they are shipping SKUs with 1/3rd the CPU Cores and 1/2 GPU Cores running at half clocks.

Or AMD shipping 6 Core Zen6 CCDs….

 

[Shogmaster]

It's not outdated. You are coping.

Even with 70 CUs, they can still release the top DT chip at 64CU, and if they're front-end limited like 9070/9070XT, it barely matters anyway. Most volume is Magnus regardless. Given putrid RDNA4 dGPU sales, betting on RDNA5 discrete driving good numbers is optimistic.

The binning also makes sense when you look at it properly. 9070 only had tolerance for CU defects, any SE/MEM/L3 defect made it a 9070 GRE. For AT2, 64 CU tolerates CU defects; 48CU tolerates CU/SE/MEM/L2 defects; 44CU covers rare cases where two memory controllers or two different shader engines are defective, which tracks with it being a China-only SKU.

And just compare the 48CU configs across dies:
- AT2: 48CU, 96 ROPs, 20MB L2, 15GB GDDR7 (720 GB/s)
- AT3: 48CU, 64 ROPs, 10MB L2, 32GB LPDDR6 (455 GB/s)

AT3 SKU is more expensive and way worse. AT2 strategy is similar to RDNA4, just better optimized for cost since the 18/15/12GB VRAM spread meaningfully cuts BoM on lower SKUs too.

And as for AT4 dGPU, the idea is like the dumb AIBs that thought there was a market for 6500XT 8GB. The only market where low performance high memory is good is AI.

Lastly, N3P is sitting at <0.04 D0 with god-tier device variance. None of this is yield-driven. It's pure artificial segmentation where pricing is to value and not to cost. AMD is ruthlessly optimizing the lineup for ASP, Marketshare and COGs. If that requires binning perfectly fine chips? Well tough shit.

It's also just the norm now. Qualcomm's doing the same thing with X2 Elite where they are shipping SKUs with 1/3rd the CPU Cores and 1/2 GPU Cores running at half clocks.

Or AMD shipping 6 Core Zen6 CCDs….

Brah, no one's reading all that shit. Summarize!

 

[Xyphie]

Presumably those AI/ML SKUs would have the full bin, why would you bin your profitmaxxed SKU? So either it's an outdated/fake chart or AT2=64CU.

 

[MCH2024]

Brah, no one's reading all that shit. Summarize!

It takes a lot more effort to refute things than to spew them.

 

[reinking]

To those who don't want a Canis TV box. Imagine this scenario.

PS6 Portable for $549, PS6 Orion for $899. Multiplayer paywall on both.

Vs

Xbox Ally For $599 and $999 SKUs, The $999 SKU having the 24 CU chip. Then a $499 Xbox PC/Console for the same 24 CU chip. And all of that without online multiplayer paywall.

What will the masses go for?

Sony's options are either do a Canis TV box, or keep selling PS5s at $550. I think the PS5 option would be worse long term.

A lot of ass-umptions in there.

 

[memoryman3]

The only market where low performance high memory is good is AI.

Driving 4K graphics at 30fps? 8GB VRAM GPUs just choke at 4K even if the compute is there to deliver it.

 

[Mr.Phoenix]

They are a thing. And you can have different memory controllers within the same die. 160b GDDR7 bus + 32b LPDDR5X bus was the right thing to do.

Of course you can have different memory controllers within the same die. And of course, a dual-purpose memory controller is a "thing".... but there is a reason they arent done.

Different mem PHY controllers on the same die is the first thing we are talking about here. The easiest thing to do is to have your mem PHY for the GDDR 32bit x 5 of them for 160bit in the PS6, and then have a separate LPDDR (most likely 64bit, so that's 2x32bit) mem phy controllers. But here is the thing, each mem phy takes up around 1.4mm2 and 1.1mm2 for GDDR and DDR controllers, respectively, in a 3nm die. So for 5 GDDR controllers, that's 7nm used up on the chip. You throw in 2 more controllers that take that number up to around 10mm2 used up.

Then there is in trying to use a combined phy controller; while they exist, they are super rare when dealing with controllers across memory types, eg, DDR + GDDR controller. But the worst part of that is that doing so would at best half the bandwidth of your whatever memory you are using on that bus, or at worst lock your max bandwidth to the max speed of the slowest RAM on that controller. Which is why no one does that nonsense.

 

[MCH2024]

Driving 4K graphics at 30fps? 8GB VRAM GPUs just choke at 4K even if the compute is there to deliver it

Not really. Most of the stuff that really drives VRAM consumption also has a big effect on compute usage.

Who would want a 5060 Ti OC with 32GB when 5060 Ti 16GB was already a meme for too much VRAM?

 

[Shadowstar39]

It doesn't make sense to have 2 SKUs with the same performance but different RAM amounts. If they need to cut RAM on one SKU to reduce costs they should also reduce performance.

And having only 20GB is not enough imo, remember PS6 will be sold until 2035 at least, and have games made for it until 2040. By then even low end PCs will have 64GB RAM + 32 GB VRAM.

Who is going to be able to afford 64gb ddr5? Shit 32gb is unaffordable today $400 and (64gb)$800. unless the prices go down, but I am not seeing it. There needs to be more companies making ram and selling it to normal people, not data centers. The only reason I can think it is not is that it takes too long to setup production.

 

[KeplerL2]

Who is going to be able to afford 64gb ddr5? Shit 32gb is unaffordable today $400 and (64gb)$800. unless the prices go down, but I am not seeing it. There needs to be more companies making ram and selling it to normal people, not data centers. The only reason I can think it is not is that it takes too long to setup production.

DDR5? I'm talking about 2040 dude, we'll be on DDR7 by then.

 

[mckmas8808]

Because people in the real world bought a Series S as a cheap way to play their games. My son bought one as a cheaper way to play live service games with his buddies. I bought one because it was cheap.

Having a cheaper hardware option is a good idea. There are cheaper GPUs than the 5090, so why can't there be a cheaper entry level console? It makes sense, especially in this economic climate.

You realize that you and your son were in the super minority though, right? Ultimately it didn't work out for Microsoft.

 

[Mr.Phoenix]

I already saw that imagebwhen they did the teardown. But that Pic dont explain crap.

The SoC has 256bit bus. Thats 8 GDDR6 RAM modules with 32bit bus each. WHERE DA FUG DOES THAT 2GB DDR5 GO?!? Its waaaay off to the side! Which part of bus is it on? Is it sharing a 32bit channel with one of the GDDR6 modules? How the fuck does that work?

SOME BIG BRAIN EXPLAIN IT TO MEEEEEEEE~

I already did... There will also be a DDR mem phy controller on that chip. There is just no other way around it. You cannot have a DDR chip share the same mem phy controller with a GDDR chip. Different voltages, timings, signaling... etc.

 

[mckmas8808]

99% of console buyers aren't NeoGAF forum users. They never even read the acronym PSSR before in their entire lives, let alone know what upscaling, framerate, or VRR are. These are fundamentally the same people who buy Switches. They love small, portable and cute. They don't care for RAM amounts or whatever. And these people will determine which SKU will be more successful.

These buyers that you speak of care more about their current ecosystem, than they care about paying the lowest possible price. Have yall already forgotten that the Xbox Series S was on sale for $300 a launch! How have you forgotten that already?!

What do you mean "flopped"? Sales wise it did very well IIRC. I'm pretty sure it out sold the Series X, but correct me if I'm wrong.

No you're wrong. Most analysis have them at around 50-50 worldwide. The Series X shortages have been over for years.

Period	Series S Share	Series X Share
Early 2022 (Global)	~75%	~25%
2023 (US)	~50%+	~50%-
Sept 2024 LTD (US)	~49%	~51%
Sept 2024 Monthly (US)	~42%	~58%

September 2024 Sales Charts: Xbox Series X Is Finally Outselling the Series S

September 2024 Sales Charts: Xbox Series X Is Finally Outselling the Series S - IGN

It took four years, but the Xbox Series X is finally outselling the Xbox Series S in the United States by a slim margin, according to this month's report from Circana.

www.ign.com

Vita TV and Series S weren't bad ideas, they were just improperly executed. Vita TV died because Vita died before it had even launched (there was no larger ecosystem to support it, plus Sony seemingly went out of their way to fuck up both backward- and cross-compatibility), and Series S was excessively cost-optimized (specs too low). I'm pretty sure the Series S to Series X ratio is something like 3:1, we know it kept the lights on at the Xbox division for the first couple of years.

Sony just needs to avoid the aforementioned pitfalls.

Can yall stop believing this lie?! When given the choice, people will choose the "real" PS6 as long as supply is available. The Series S lost vs. the Series X, even in it's own Xbox family. When given the choice, more people picked the Series X bro. Think about that for a second.

We spoke with Piscatella about why the Series X has passed the S, and he had this to say: Few things happening here as opposed to earlier in the cycle when S was leading. Of course, we have to start with supply. Series S was more widely available during 2020-2022, with X only getting fully in stock a few years after launch due to looks at everything 2020-2022. Then we also have console buyers in general shifting to a slightly older audience with higher average household income than we saw early in this generation, which would suggest that perhaps price sensitivity may be a bit lower now than it was previously. There are likely several other contributing factors, but these are the two I'd put on the top of my list.

The Series S was the worst selling current gen launch hardware. It's sold worst than the current Nintendo Switch 2. Again think about that. The Switch 2 has already passed the Series S and it's only been out for less than a year.

 

[Proelite]

Yeah if they really feel the need for a lower end SKU it makes sense to use Orion.

Though I would do a little bit different than what you suggested:

6-core CPU cluster
16 WGP GPU
-10% clock speed on CPU/GPU
128-bit bus with 24GB RAM

This basically guarantees near 100% yield rate for the SoC, reduces the BOM by $60 for the RAM, plus another $20-30 for the reduced board/cooling requirements. Shrinking the SSD to 512GB would be another big BOM reduction but I think they could only get away with that if they forced devs to use NTC.

How much of the GPU are they disabling here?

Disabling 1 out of 3 Shader engines should get you to 18 WGP GPU.

 

[Physiognomonics]

How much of the GPU are they disabling here?

Disabling 1 out of 3 Shader engines should get you to 18 WGP GPU.

That would have a bad balance. Like say 30% less expensive (still 24GB of the most expensive component) and with 40%-50% less performance. a very bad deal

 

[asdasdasdbb]

That would have a bad balance. Like say 30% less expensive (still 24GB of the most expensive component) and with 40%-50% less performance. a very bad deal

I expect the clock speed of the PS6 to be not much better than the PS5 because there's no point due to the lack of memory bandwidth.

But I think 3+ Ghz should be possible.

 

[Proelite]

That would have a bad balance. Like say 30% less expensive (still 24GB of the most expensive component) and with 40%-50% less performance. a very bad deal

This is better deal than PS4, XB1 and PS5 after their pro counterparts released.

 

[IDKFA]

You realize that you and your son were in the super minority though, right? Ultimately it didn't work out for Microsoft.

A super minority you say........

No you're wrong. Most analysis have them at around 50-50 worldwide. The Series X shortages have been over for years.

Period	Series S Share	Series X Share
Early 2022 (Global)	~75%	~25%
2023 (US)	~50%+	~50%-
Sept 2024 LTD (US)	~49%	~51%
Sept 2024 Monthly (US)	~42%	~58%

September 2024 Sales Charts: Xbox Series X Is Finally Outselling the Series S

September 2024 Sales Charts: Xbox Series X Is Finally Outselling the Series S - IGN

It took four years, but the Xbox Series X is finally outselling the Xbox Series S in the United States by a slim margin, according to this month's report from Circana.

www.ign.com

That doesn't looks like a super minority to me. A super minority would be globally 10% or less.

Looks like the S was very popular. Imagine how popular a budget PS6 could be!

Can yall stop believing this lie?! When given the choice, people will choose the "real" PS6 as long as supply is available. The Series S lost vs. the Series X, even in it's own Xbox family. When given the choice, more people picked the Series X bro. Think about that for a second.

People will choose the real PS6, but a cheaper option would be smart in the current economic climate. The Series S did sell too badly looking at the data you provided, so I see no reason why a light PS6 would be a success.

The Series S was the worst selling current gen launch hardware. It's sold worst than the current Nintendo Switch 2. Again think about that. The Switch 2 has already passed the Series S and it's only been out for less than a year.

I'm not sure this is a valid argument because in a few months the Switch 2 will outsell the Series X.

 

[HerjansEagleFeeder]

These buyers that you speak of care more about their current ecosystem, than they care about paying the lowest possible price. Have yall already forgotten that the Xbox Series S was on sale for $300 a launch! How have you forgotten that already?!

Here in Europe, PlayStation is part of pop culture. Everyone knows the brand. Everyone wants to get on the new gen as fast as possible, as long as prices are sane. This is not a Series S situation. When Sony launches any next gen device in the ballpark of 500 Euros, people will pay. The handheld will be sold out at launch.

 

[mckmas8808]

A super minority you say........


That doesn't looks like a super minority to me. A super minority would be globally 10% or less.

Looks like the S was very popular. Imagine how popular a budget PS6 could be!


I'm not sure this is a valid argument because in a few months the Switch 2 will outsell the Series X.

When I said "super minority" I was comparing the Xbox Series S to the whole current generation numbers. Not just Microsoft Xbox numbers. The Xbox Series S has sold around 15 million total consoles in 5+ years. That's not good, by any measure.

And the to your Switch 2 point, that's the point I'm saying also. Microsoft's whole plan was screwed up for the Series console. They focused on the wrong things, because they don't understand gamers world wide.

Here in Europe, PlayStation is part of pop culture. Everyone knows the brand. Everyone wants to get on the new gen as fast as possible, as long as prices are sane. This is not a Series S situation. When Sony launches any next gen device in the ballpark of 500 Euros, people will pay. The handheld will be sold out at launch.

Which is why Sony should focus on LOWERING the cost of the PS5 to reach more people from the PS4 gen in 2027. Imagine if people in Europe could play GTA6 by buying a PS5 for 350 Euros in 2027!

 

[IDKFA]

When I said "super minority" I was comparing the Xbox Series S to the whole current generation numbers. Not just Microsoft Xbox numbers. The Xbox Series S has sold around 15 million total consoles in 5+ years. That's not good, by any measure.

And the to your Switch 2 point, that's the point I'm saying also. Microsoft's whole plan was screwed up for the Series console. They focused on the wrong things, because they don't understand gamers world wide.

If we just look at raw numbers then the Series S doesn't look like much of a success. However, when you consider it against low overall Xbox sales then it isn't half bad.

We could compare to another under powered system in the Switch. Was less powerful than a PS4 but sold over 100 million. The Switch 2 is underpowered, yet selling like gangbusters.

I think a budget PS6 would be a success for Sony. They need as many people as possible to buy their hardware so they can make money with software/services. Releasing a traditional powerhouse might not cut it in 2027 if it's priced close to a four figure sum because it's unaffordable for the most people. However, the handheld/hybrid might turn out to be the "budget" option anyway.

 

[slrbomber]

What do you mean "flopped"? Sales wise it did very well IIRC. I'm pretty sure it out sold the Series X, but correct me if I'm wrong.

Series X is marginally higher in sales its 51% X and 49% S. Thats because X had more stockouts earlier on in the gen while Series S was produced more. So the X has had higher demand and now has sold more. When I said flopped I meant it was supposed to be a mass appeal product that sold significantly more and didnt create significant developer issues. On those metrics it failed. I dont think Helix is going to have a Series S type device for example.

 

[Shogmaster]

I already did... There will also be a DDR mem phy controller on that chip. There is just no other way around it. You cannot have a DDR chip share the same mem phy controller with a GDDR chip. Different voltages, timings, signaling... etc.

Makes sense, and I suppose a small bus just for DDR5 would be relatively small space on the die. So its 256bit + 64bit bus? Or did they make super slow with 256+32?

 

[Ebrietas]

Maybe PS6 "S" is meant to slot in where PS5 would otherwise have in the upcoming PS lineup as the cheap entry level console (compared to the PS6, PS6 handheld, PS6 Pro). Similar or slightly better overall performance than a PS5 but a lot cheaper to make. Kind of like the M1 MacBook Air vs the MacBook Neo.

 

[yogaflame]

Its good news Sony back out from a cheaper version of ps6.We dont want another xbox series s scenario that hinders and held back game development and maximizing the full potential of the main console.

 

[Mr.Phoenix]

Makes sense, and I suppose a small bus just for DDR5 would be relatively small space on the die. So its 256bit + 64bit bus? Or did they make super slow with 256+32?

Yup, most definitely a 32bit bus.

Just keep in mind tho that every 32bit phy controller takes up space. And things like controllers and SRAM don't scale as well as chip logic units. It was, hence, a no-brainer why Sony would cut the bus size down to 160bit on the PS6 the second they were using RAM chips fast enough to give them reasonable bandwidth on a smaller bus.

 

[Shogmaster]

Yup, most definitely a 32bit bus.

Just keep in mind tho that every 32bit phy controller takes up space. And things like controllers and SRAM don't scale as well as chip logic units. It was, hence, a no-brainer why Sony would cut the bus size down to 160bit on the PS6 the second they were using RAM chips fast enough to give them reasonable bandwidth on a smaller bus.

In normal economy, I can see PS6 being $600 easy.

 

[Merline]

Its good news Sony back out from a cheaper version of ps6.We dont want another xbox series s scenario that hinders and held back game development and maximizing the full potential of the main console.

Do you really think you'll get true next gen exclusives on a $799 PS6? At that price point, it stops being a mass market product and becomes a niche device. This thing will sell at most 40 million units, though I'm not even sure about that.

So if there is no cheaper model like a PS6S, don't expect exclusive AAA games.

 

[mckmas8808]

Somewhat off-topic. Put this latest page in Google Gemini Pro and it's able to read all 17 pages of this thread and give me a summary.

There's many things that A.I. makes better. Don't let the AI haters confuse you.

 

[Mr.Phoenix]

Do you really think you'll get true next gen exclusives on a $799 PS6? At that price point, it stops being a mass market product and becomes a niche device. This thing will sell at most 40 million units, though I'm not even sure about that.

So if there is no cheaper model like a PS6S, don't expect exclusive AAA games.

Even if there were cheaper versions... I wouldn't expect exclusives.

Well, we can expect PS6 exclusive "features", but I am not expecting any PS6 exclusive games. At least not for the first 4 years.

 

[saintjules]

There's many things that A.I. makes better. Don't let the AI haters confuse you.

I use it as an aide and not to replace anything. So all good there!

 

[mckmas8808]

If we just look at raw numbers then the Series S doesn't look like much of a success. However, when you consider it against low overall Xbox sales then it isn't half bad.

We could compare to another under powered system in the Switch. Was less powerful than a PS4 but sold over 100 million. The Switch 2 is underpowered, yet selling like gangbusters.

I think a budget PS6 would be a success for Sony. They need as many people as possible to buy their hardware so they can make money with software/services. Releasing a traditional powerhouse might not cut it in 2027 if it's priced close to a four figure sum because it's unaffordable for the most people. However, the handheld/hybrid might turn out to be the "budget" option anyway.

Then why not just double down on the PS5 and sell it for cheaper than what you sold it for at launch? Wouldn't that achieve the same thing, without screwing up a whole next-gen? You don't believe a $400 PS5 with GTA6 bundled-in wouldn't sell like hotcakes?

 

[Bitstream]

Even if there were cheaper versions... I wouldn't expect exclusives.

Well, we can expect PS6 exclusive "features", but I am not expecting any PS6 exclusive games. At least not for the first 4 years.

Has any next gen console launched without a single exclusive game? It would be a mistake to put out a $700 console without a single exclusive first party title that really shows off things that aren't possible on the last gen.

 

[Omnipunctual Godot]

$350 would be a very fair price for a budget model.