Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
More
1. 6nm Oberon+
2. Not because cost cut like 28nm to 16nm
3. Just save few % power consumption
4. so buy when you want it, don't wait
5. the world have shortage on silicon and wafer now and near future
6. die shrink = can cut more
 
Last edited:
More
1. 6nm Oberon+
2. Not because cost cut like 28nm to 16nm
3. Just save few % power consumption
4. so buy when you want it, don't wait
5. the world have shortage on silicon and wafer now and near future
6. die shrink = can cut more

Very interesting. So is Oberon+ a smaller Oberon (6nm? Not 5nm EUVL?) for power consumption savings? They could still ecke out a slight power boost, especially if going 5nm EUVL, and also give more of a die shrink than 6nm (actually this is my first time hearing of 6nm in any capacity. Maybe it's a rogue TSMC node, I might've seen a single roadmap with 6nm on it xD).

Also it coming 2022 is very interesting too; this sounds like it'll be moreso a PS5 Slim than a PS5 Pro. But it doesn't rule out chance of a PS5 Pro a year or two later, either.
 
Given the current form factor of PS5 I don't even see how they do a Pro model.

Lol yeah that is a very good point xD. I have a few ideas; for starters they'd need 5nm EUVL for as much power consumption reduction and die shrink as possible. They'd also probably need to go chiplet for the GPU, but I don't think doubling i.e 2x 36 CUs is going to work.

Advantage here though is that they could simply leverage somewhat higher clocks with savings on GPU TDP and power savings on other things (CPU, maybe more energy-efficient NAND, lower-power GDDR memory (provided that could come about), etc.) to bring out a bit more performance. Nothing like 2x base PS5, but a neat bump.

PS5 Pro could better serve as a testing ground for some technologies that can be refined and perfected on a PS6, mainly.


Looking forward to seeing this.
 
Lol yeah that is a very good point xD. I have a few ideas; for starters they'd need 5nm EUVL for as much power consumption reduction and die shrink as possible. They'd also probably need to go chiplet for the GPU, but I don't think doubling i.e 2x 36 CUs is going to work.

Wouldn't really be much a pro model if you are not at least doubling the Teraflops...at least imo.

Not to mention there may be BC reasons for choosing 72CU's...

I'd think they'd have to adopt a console design like the XSX. Split motherboard, vapour chamber, in a tower-like form factor.
 
Very interesting. So is Oberon+ a smaller Oberon (6nm? Not 5nm EUVL?) for power consumption savings? They could still ecke out a slight power boost, especially if going 5nm EUVL, and also give more of a die shrink than 6nm (actually this is my first time hearing of 6nm in any capacity. Maybe it's a rogue TSMC node, I might've seen a single roadmap with 6nm on it xD).

Also it coming 2022 is very interesting too; this sounds like it'll be moreso a PS5 Slim than a PS5 Pro. But it doesn't rule out chance of a PS5 Pro a year or two later, either.

Leaker explained:
The capacity of TSMC already full.
And Apple is the 1st priority cilent, Apple already took those reserved for Huawei and still not enough.
Others like AMD/NVIDIA/SONY/MS ate the remains.
So if you want more chips you can only improve yield or die shrinking.
6nm have same design rule as 7nm, so it can deploy relatively fast.
But cost much more than just improve yield on 7nm.
 
Maybe Sony will use 6nm as additional waffer allocation given that 7nm will still be quite full in 2022.

So I would not expect Pro versions to come anytime soon (if at all this gen) and also Sony to be in a bit of a pickle duo to waffer shortage, they would probably be able to sell 25M per year for next 5yrs if there wasnt any.

MS will be interesting with XSS being 195m², with that size and those clocks yields must be pretty damn good, and there are tons of them per waffer.
 
Yeah I don't expect pro versions at all this gen either. A pro model would probably cost $600.

They'll use any savings for better efficiency in size and power and increased storage.
 
.
Yeah, smaller die but I supposed that is predicated on them getting the same or better yields.

However, yield is likely lower and cost will be higher as well. Right now, you only go to a new node if you need either more speed or less power consumption because each chip is going to cost more on a new smaller node than it would on a more mature larger node.

Apple is taking up all the wafer starts they can for 5 nm because they need the power savings for their mobile chip, but they're also paying quite a bit per chip.

But while the cost per chip on 6 nm might be higher than it would be on 7 nm, it could still be beneficial for Sony if it reduces the power consumed by the chip and thus allows them to reduce the cost of their cooling solution.

Regards,
SB
 
I would be surprised if it would allow any meaningful reduction in cooler size,
a bare die from TSMC is ~$85-95 for the ps5 SOC, say 5 nm is roughly twice the price of 7 nm at TSMC if the link below is anything to go by, lets be generous and say a 6 nm chip is only 20% more than a 7nm one, so that would be $102-114 per PS5 SOC, even if the 6nm soc produced a third of the heat of the 7nm version there is no way you could make a saving of $15-19 in the cooler that they have.
The only way I can see it making sense is for a PS5 slim, just so it can have a smaller footprint, allowing smaller boxes/cheaper shipping and all the rest of it. The problem there though is that unless they start selling a PS5 slim in the next 6 months it doesn't matter, because once ps5s are being shipped via container ship the marginal savings on shipping will be immaterial.

The only reason it makes sense is for capacity.

It is interesting that microsoft seems to have more SOCs coming off the line, they have less retail units out there than sony it seems, iirc phil said that they started producing systems 2-3 months ago, and anecdotally a bunch of people were comparing series x build dates on resetera, the earliest build date anyone found was mid july.
I wonder how many Xcloud server blades they are building? surely it must be in the millions if they are dual use for AI stuff in the data centre. bdsams, the guy who leaked the Series S, said the first xcloud server blades rolled off the line second week of september.
I really want a blow by blow behind the scenes book about this launch one day, from microsofts and sonys perspective, Im not holding my breath though


https://www.techpowerup.com/272267/alleged-prices-of-tsmc-silicon-wafers-appear
 
However, yield is likely lower and cost will be higher as well. Right now, you only go to a new node if you need either more speed or less power consumption because each chip is going to cost more on a new smaller node than it would on a more mature larger node.

Apple is taking up all the wafer starts they can for 5 nm because they need the power savings for their mobile chip, but they're also paying quite a bit per chip.

But while the cost per chip on 6 nm might be higher than it would be on 7 nm, it could still be beneficial for Sony if it reduces the power consumed by the chip and thus allows them to reduce the cost of their cooling solution.

Regards,
SB

I was talking about now currently on the 7nm process.
 
Status
Not open for further replies.
Back
Top