General Next Generation Rumors and Discussions [Post GDC 2020]

Use them in PSNow servers with lowered clocks (PS4Pro) or lowered clocks and disabled CU's (PS4.) Or to power their nascent AI division.

Does sony have any plans to do this however ? For now it be a massive investment into PSNOW which seems to be struggling with sign ups and its only 720p so it seems like everything would have to be redone if they want to take advantage of ps5 level hardware. They also as others have pointed out signed a deal with ms for cloud stuff. I also don't know if the AI division could absorb enough defective dies.
 
Where's the 2.5GHz figure from?
Nowhere. What I'm saying is if Sony have bet on a new untested APU design, which this is for both PS5 and XSX, and targeted at frequencies that are at the very high-end of what the test yields have shown viable, then Sony have taken a huge, huge risk. I don't believe this is something Sony would do because they've never chased that high-performance crown, Microsoft merely yielded that for most of this generation.

Of course in yield terms, low yield could be getting 14 viable chips from a wafer rather than 15. "Poor" is relative to your expectations.

But simplistic data points thinking: we heard no rumour about XSX having yield or heat, in fact they started production a while back. Not much stepping according to Github. WRT PS5, lots of steppings, hearing a lot about heat and yield rumours, Sony still has not revealed a console,Tott brings up another rumour of yet another last minute revision. Simplistically without really knowing much and what is true or not, but based on this information I'd look at clockspeed differences for yield issues. A last minute revision to improve yields?

Rumours aren't data points: :nope: We didn't see PS4 until June 2013. Stepping, or conversely lack of stepping, doesn't indicate a whole lot. Traditionally Microsoft have made very few revisions to Xbox consoles whereas Sony endlessly make changes, some very minor. Just look all the the individual CUH model numbers of PS4/Pro where aren't due to region differences and it's kind of nuts. A lack of stepping could mean Microsoft decided not undertake last minute revision (there are cost implications) where Sony has traditionally invested in changes for long-term cost saving. More stepping also happens with more experimentation and refinements. Is that bad? I mean what can you you infer from that? That Sony revise things a lot? We know that, it's done this with four generations of PlayStation. ¯\_(ツ)_/¯

multiple people who work at sony that I know. I can ask about the revision. I was told yields are worse than they thought driving prices up. My sources in MS have said the opposite that their yields are better than what they thought they would be

A particular bad yield, as in outside the contracted supply, would drive prices up but not for Sony or Microsoft. AMD designed the chip and TSMC manufacture it. From the console manufacturer's perspective, this is SEP. If the problem were so severe to to limit console production sold the manufacturer would likely be due compensation which could drive the BOM cost down.

But this is all super unlikely. AMD are an experienced chip designer and TSMC are a very experienced fabricator. Bad runs happen, bad wafers, contamination etc, unexpected clashes of design against process. It would be extraordinary if AMD actually designed a chip that had really poor production yields on a TSMC process. :yep2:

Does sony have any plans to do this however ?
Everybody, including Microsoft, think streaming games is the future. Sony were convinced of this before Microsoft so why wouldn't they want to invest more in their future cloud technology?
 
So the only way PS5 could be having higher yield issues than XBSX is if the chips performance at speed is below expected. Wouldn't that result in fewer units, but no additional costs, because surely AMD are contracted to chips to spec? They'd have provided Sony with estimates that Sony agreed to, and failure of enough chips to be able to run at Sony's specified clock speeds would be a fault of AMD's estimates. I think.

Depends on the terms of the contract. So it's a matter of what AMD is comfortable rating the chips at, and how the rating is tested. It's then on the customer (Sony) if they want to use the chips beyond what AMD specced them at. What if the testing of the chip says its valid for the agreed upon spec, but then when running different code it doesn't perform up to Sony's expectations? Who eats that cost then?
 
But this is all super unlikely. AMD are an experienced chip designer and TSMC are a very experienced fabricator. Bad runs happen, bad wafers, contamination etc, unexpected clashes of design against process. It would be extraordinary if AMD actually designed a chip that had really poor production yields on a TSMC process. :yep2: If this happened at scale, the

I don't have much experience with chip manufacturing but It was my understanding that a company approaches TSMC and purchases a wafer for X amount of money and then you take the amount of chips made and divide it by the cost of the wafer. So why would TSMC be on the hook for anything ? Esp if it something like clock speeds being the issue ?

I dunno however because I"m not the primary source and for better or worse I am not the smartest on this board so some of you guys have asked questions that I wish I had the knowledge to ask.
 
pre or post tapeout?

pre-tapeout that's just an estimate.
I believe Sony have sone experience in semiconductor manufacture. They still operate fabs for some chip themselves, as well as CCDs, There are few companies in the CE who have as much experience in this as Sony so let's give them so credit to understand the perils of this process.

So the only way PS5 could be having higher yield issues than XBSX is if the chips performance at speed is below expected. Wouldn't that result in fewer units, but no additional costs, because surely AMD are contracted to chips to spec? They'd have provided Sony with estimates that Sony agreed to, and failure of enough chips to be able to run at Sony's specified clock speeds would be a fault of AMD's estimates. I think.

Yes, this is typically how it works. I mean if you want to engage AMD to design a chip, then independently contract a third party (like TSMC) to manufacture it with no monthly-supply contract and take all the risks yourself, I guess you could. I reckon both AMD and TSMC would bite your hand off! The reason you contract out is exploit expertise and spread risk. That expertise in itself minimises that risk.

I don't have much experience with chip manufacturing but It was my understanding that a company approaches TSMC and purchases a wafer for X amount of money and then you take the amount of chips made and divide it by the cost of the wafer. So why would TSMC be on the hook for anything ? Esp if it something like clock speeds being the issue ?

I've never heard anybody doing this. I couldn't say it's never happened but this is not the norm. Once you have a functioning chip design, TSMC will undertake several production runs to determine yields because that's really the only way to know how many chips you're likely going to get from any given wafer.

The fabricator will often, based on their experience of the process and other work, make recommendation for changes to the design that could improve yields but it's up to you whether you go through this process. If you're manufacturing tens of millions of chips, that costly process may be worth it depending on the potential improvement in yields. But generally you agree a contract where you pay a price according to the fabricators predicted yields. This is typically a curve and if production is particularly bad, the fabricator eats the cost. TSMC are very good at what they do, so this really isn't a big risk for them.

I dunno however because I"m not the primary source and for better or worse I am not the smartest on this board so some of you guys have asked questions that I wish I had the knowledge to ask.
It's not about being smart, you either have experience of something like this or you don't. There probably any that many that do.

Both consoles are likely to launch in the next 4-5 months, and just saying that still feels kind of unreal, so this won't be a mystery for very long! :yes:
 
Does sony have any plans to do this however?

Nothing official as far as I know. But if I had thousands of chips that had failed to meet the requisite 3.5+2.23GHz of the PS5, a service that allowed people to stream PS4 games, and hardware backwards compatibility in those thousands of defective chips... I'd probably find a use for them.

For now it be a massive investment into PSNOW which seems to be struggling with sign ups and its only 720p so it seems like everything would have to be redone if they want to take advantage of ps5 level hardware.

The PS3 is a 720p console. The base PS4 streams/records at 720p. The fact that PSNow streams at the resolution indicates that they've essentially plugged a PS4 into a server farm. I don't see why they wouldn't do the same with the PS5, which will at least be able to stream/record at 1080p60, because that's what the PS4Pro can do.

Personally, I expect 1440p60 and 4K30 as streaming/recording options on the PS5.

They also as others have pointed out signed a deal with ms for cloud stuff. I also don't know if the AI division could absorb enough defective dies.

Cloud stuff: that would require Microsoft to emulate the PS5 on a PC. I don't see that as being very likely.

AI division: I've no idea of the scale of their plans, but an AI division is going to need GPU's.
 
I believe Sony have sone experience in semiconductor manufacture. They still operate fabs for some chip themselves, as well as CCDs, There are few companies in the CE who have as much experience in this as Sony so let's give them so credit to understand the perils of this process.



Yes, this is typically how it works. I mean if you want to engage AMD to design a chip, then independently contract a third party (like TSMC) to manufacture it with no monthly-supply contract and take all the risks yourself, I guess you could. I reckon both AMD and TSMC would bite your hand off! The reason you contract out is exploit expertise and spread risk. That expertise in itself minimises that risk.



I've never heard anybody doing this. I couldn't say it's never happened but this is not the norm. Once you have a functioning chip design, TSMC will undertake several production runs to determine yields because that's really the only way to know how many chips you're likely going to get from any given wafer.

The fabricator will often, based on their experience of the process and other work, make recommendation for changes to the design that could improve yields but it's up to you whether you go through this process. If you're manufacturing tens of millions of chips, that costly process may be worth it depending on the potential improvement in yields. But generally you agree a contract where you pay a price according to the fabricators predicted yields. This is typically a curve and if production is particularly bad, the fabricator eats the cost. TSMC are very good at what they do, so this really isn't a big risk for them.


It's not about being smart, you either have experience of something like this or you don't. There probably any that many that do.

Both consoles are likely to launch in the next 4-5 months, and just saying that still feels kind of unreal, so this won't be a mystery for very long! :yes:

The issues could bring credence to the 9 tflop rumor. Perhaps Amd was contracted for a 9tflop gpu but when MS announced their specs sony reacted with higher clocks.

And yea I know , you guys brought up such great questions though so I was upset I didn't think of them ! to busy fawning over the baby on the other side of video chat lol
 
Why would that be a stipulation of their agreement?
double costs. They are signing with MS because it's cheaper to rent the hardware and maintenance upgrades etc, over trying to attempt it themselves. They could have both, sometimes it's important to have your own infrastructure, but I don't know how big a scale we're talking about here.
 
I believe Sony have sone experience in semiconductor manufacture. They still operate fabs for some chip themselves, as well as CCDs, There are few companies in the CE who have as much experience in this as Sony so let's give them so credit to understand the perils of this process.
just offering an explanation, I don't necessarily believe everything I write. But I also don't believe that makes experience infallible either. Sometimes you make a design for X, and someone realizes it needs changing because of Y and a whole design change is made. A lot of last minute changes can lead to problems.
 
Last edited:
The issues could bring credence to the 9 tflop rumor. Perhaps Amd was contracted for a 9tflop gpu but when MS announced their specs sony reacted with higher clocks.

And yea I know , you guys brought up such great questions though so I was upset I didn't think of them ! to busy fawning over the baby on the other side of video chat lol
We have heard about the 12tf target specs since about january 2019.
 
Wouldn't it have to be quite a lot more expensive to be more expensive than the XSX though?

Going by this wafer calculator, a wafer diameter of 450mm, and the following dimensions:
XSX = 360.45mm2
√360.45 = 18.99
Dies per wafer = 378

PS5 = 300mm2 (going from an old leak that seems more viable now that we know it only has 36CU's)
√300 = 17.32
Dies per wafer = 457

That's a difference of 79 that the PS5 would have to chew through. Sure, such high clockspeeds may have a detrimental effect on yield, but the same is true of a larger chip too.

These chips aren't manufactured on 450mm wafers, that size doesn't really exist. 300mm are being used, using your other metrics gives a difference of 34 chips per wafer(190 vs 156).
 
Sony absolutely had a design for a ~ 9.x TF SOC based on Ariel.
The idea of trying to crank up the speed in relation to MS's TF figure is the unlikely part. A 10% gain in speed is adding considerably to the cost - is it really worth it if you are pushing the silicon too far? It's more likely the design is what Cerny considers the sweet-spot for the hardware, selecting 36 CUs and then finding clocks to maximise that, and designing the entire system around that. It's very unlikely Cerny chose 36 CUs with a sweet-spot of 9 TFs, then saw MS's 12 TFs, over-volted the thing till it started melting, slapped a massive HSF on it and called it a day only to see chips failing by the bucket load now. ;)

If that is the case, it'll be a text-book case of poor engineering business. Sometimes these things happen. Seems unlikely with Cerny at the reins, though.
 
Back
Top