Bloomberg on PS5 yields, orders, and price ranges [2020-09-14]

[Shortbread]

High clocks usually are part of the causes to lowered yields is that not the case?

There can be a whole host of factors on reduced or poor yields (i.e., core failures, transistor/logic issues, etc.), and yes, frequency targets can be one of them. And without knowing which ones are affecting Sony's PS5 chip yields, we're simply shooting arrows in the dark.

 

[gongo]

Clocks, size, design, thermal issues, process issues, non-optimal wafers (cost), contamination, packaging defects - and many more. It’s not just clocks.

Of these factors listed, PS5 high clock speeds stood out as the new and experimental for TSMC.

 

[Shortbread]

Of these factors listed, PS5 high clock speeds stood as new and experimental for TSMC.

This topic might be best served here: https://forum.beyond3d.com/threads/...al-spin-post-gdc-2020-xbsx-ps5.61641/page-188 Since there is a lot of technical speculation and facts covering PS5 architecture, chip design, and cooling solution.

 

[eastmen]

Are you saying they can cheat their way out in giving big numbers because nobody is able to spot it?

What do you mean by cheat ?

They already said powershift allows them to dynamically shift power between the cpu and gpu so they wouldn't be lying if yields aren't as good as they expect and the chip spends less time at max clock speeds and like i said who is going to figure out how often the chip hits max clocks and who is going to argue the semantics about what sony promised in terms of how often it hit max clocks ?

I mean it all depends on if the yield issue could be fixed by using chips that don't hit max clocks often enough. Who knows . I think the powershift was already being used to fix that issue

 

[HBRU]

powershift allows them to never have to announce a downclock. to the end user they wont notice if the ps5 hits max clock speed 10% of the time or 90% of the time (numbers out my ass). How would an end user even be able to check it.

Yes. That's why I'm gonna buy an XSX... this all powershift story sounds me 8 TF....

 

[eastmen]

Yes. That's why I'm gonna buy an XSX... this all powershift story sounds me 8 TF....

i wouldn't go that far.

I'm just buying a new video card. The majority of these games will be on my pc anyway

 

[Deleted member 11852]

Of these factors listed, PS5 high clock speeds stood out as the new and experimental for TSMC.

Can you explain your thinking for stating the PS5 APU clock is "new an experimental" for TSMC ?

 

[Deleted member 11852]

They already said powershift allows them to dynamically shift power between the cpu and gpu so they wouldn't be lying if yields aren't as good as they expect and the chip spends less time at max clock speeds and like i said who is going to figure out how often the chip hits max clocks and who is going to argue the semantics about what sony promised in terms of how often it hit max clocks ?

It's only diverting power unused by the CPU to the GPU. There is no way Sony could get away with this. Sure, they've not been specific about high often the max GPU clock is hit but that in part is the nature of dynamics clocks that it can vary in any one game and from one game to another. You see this on the PC all the time, your GPU is rarely hammering its max clock rate.

But even a mild tweaking of the performance profile will noticed by developers. Yields issues are short term, you don't change the performance profile of your whole console for the next 6-7 years, you man up and pay the cost and fix your yields issues.

 

[eastmen]

It's only diverting power unused by the CPU to the GPU. There is no way Sony could get away with this. Sure, they've not been specific about high often the max GPU clock is hit but that in part is the nature of dynamics clocks that it can vary in any one game and from one game to another. You see this on the PC all the time, your GPU is rarely hammering its max clock rate.

But even a mild tweaking of the performance profile will noticed by developers. Yields issues are short term, you don't change the performance profile of your whole console for the next 6-7 years, you man up and pay the cost and fix your yields issues.

I wouldn't develop something for best case scenarios . I would develop for worse case. So i would already expect the games to be developed with the lower clocks in mind esp launch games to ensure they perform properly and let dynamic scaling worry about best case

Also Sony already got away with it with smart shift. People already welcome their system running slower due to power delivery. Look at it this way. If a perfect ps5 die runs at max speed 98% of the time and sony needed to use dies that ran at max speed 96% of the time who would notice ? Will devs notice ? Would devs not already have been developing for worse case scenario ? The majority of titles next gen will have a dynamic resolution anyway and be designed with worse case scenarios.

 

[gongo]

Can you explain your thinking for stating the PS5 APU clock is "new an experimental" for TSMC ?

Do we have a working GPU running at 2.3ghz?

 

[Shortbread]

Console gamer's are so weird at times. Be it smart-shift, boost clocks, overclocking, or whatever terminology being used to describe performance balancing and/or scaling, PC gamer's welcome these techniques. But whatever...

 

[gongo]

What do you mean by cheat ?

They already said powershift allows them to dynamically shift power between the cpu and gpu so they wouldn't be lying if yields aren't as good as they expect and the chip spends less time at max clock speeds and like i said who is going to figure out how often the chip hits max clocks and who is going to argue the semantics about what sony promised in terms of how often it hit max clocks ?

I mean it all depends on if the yield issue could be fixed by using chips that don't hit max clocks often enough. Who knows . I think the powershift was already being used to fix that issue

I find it unusual to drop the news about yields days before the big event.
It opens the door for Sony to ethically announce new clocks, if old ones were the main cause?

 

[Inuhanyou]

Yes. That's why I'm gonna buy an XSX... this all powershift story sounds me 8 TF....

What is the basis for assuming the gpu would need to clock down to such a degree that it loses 20% of its stated gpu power? In a case that big neither heating or cooling issues should pose a problem. Nor should even the worst case scenario be anywhere near such a number

 

[eastmen]

I find it unusual to drop the news about yields days before the big event.
It opens the door for Sony to ethically announce new clocks, if old ones were the main cause?

first you have to assume that lowering clocks would increase yields. It could be another issue causing bad chips.

But if sony does have to announce downclocks why would they do it at a big show ? They would dump the news on a friday night or after a big news story breaks for something else. Anything to burry it. But again powershift already sets the stage for the apu not hitting its max clocks all the time. So why even draw attention to it. Anything that brings the ps5 closer to xss and further from xsx will be ran as a negative all over the place

 

[Xbat]

I find it unusual to drop the news about yields days before the big event.
It opens the door for Sony to ethically announce new clocks, if old ones were the main cause?

Good Lord, if anything them missing there target by 4 million means that they haven't lowered the clocks if anything.
That is if it's even the high clocks that's the reason for the worse than predicted yield.

 

[gongo]

first you have to assume that lowering clocks would increase yields. It could be another issue causing bad chips.

But if sony does have to announce downclocks why would they do it at a big show ? They would dump the news on a friday night or after a big news story breaks for something else. Anything to burry it. But again powershift already sets the stage for the apu not hitting its max clocks all the time. So why even draw attention to it. Anything that brings the ps5 closer to xss and further from xsx will be ran as a negative all over the place

Good Lord, if anything them missing there target by 4 million means that they haven't lowered the clocks if anything.
That is if it's even the high clocks that's the reason for the worse than predicted yield.

Both, what if the 11M 50% yield is AFTER the new clocks? Food for thoughts?

Sony can just sneak it into the specs sheet to be put up at Playstation.com with the price.

 

[HBRU]

Do we have a working GPU running at 2.3ghz?

Self explanatory....

 

[arandomguy]

Console gamer's are so weird at times. Be it smart-shift, boost clocks, overclocking, or whatever terminology being used to describe performance balancing and/or scaling, PC gamer's welcome these techniques. But whatever...

It's because the behaviors normalized now and used by all vendors on the PC (GPU or CPU). If it were introduced now with say only one graphics vendor doing so it'd have the same "controversy," which did exist back when Kepler launched on the PC.

 

[Shortbread]

first you have to assume that lowering clocks would increase yields. It could be another issue causing bad chips.

But if sony does have to announce downclocks why would they do it at a big show ? They would dump the news on a friday night or after a big news story breaks for something else. Anything to burry it. But again powershift already sets the stage for the apu not hitting its max clocks all the time. So why even draw attention to it. Anything that brings the ps5 closer to xss and further from xsx will be ran as a negative all over the place

Doesn't the power-shift profile only engage when the game isn't fully utilizing the GPU and/or CPU at a given instance. Thus, delivering the additional power to either the GPU or CPU when need be. Which seems like a smart power savings and thermal design decision.

 

[gongo]

first you have to assume that lowering clocks would increase yields. It could be another issue causing bad chips.

But if sony does have to announce downclocks why would they do it at a big show ? They would dump the news on a friday night or after a big news story breaks for something else. Anything to burry it. But again powershift already sets the stage for the apu not hitting its max clocks all the time. So why even draw attention to it. Anything that brings the ps5 closer to xss and further from xsx will be ran as a negative all over the place

You made it as though Sony has pulled a fast one on us!
This 50% yield and hefty drop in production, there may be some truths inside all the smokes and mirrors.
I don't remember TSMC 7nm having such poor output, in PC space.

How can one goes about testing PS5 game clock speeds to get the final say?