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

Maybe there are component logistics issues? :oops: (because Covid & supply lines)
yea I think some others have mentioned this, but with respect to this article, Bloomberg Japan is citing that the SoC is the one having yield issues. The article cites a 50% yield on SoC.
That's pretty bad. That's like big bulky expensive Intel CPU chips bad, definitely not the type of yield I expect for a game console.

So if we do some napkin math so that people can ground some discussion of what things could be in terms of pricing/BOM etc.

IIRC another member here cited 11K for a 300mm 7nm wafer.
So area = pi * 150^2
XSX is 360mm^2 == 196 chips per wafer
PS5 is 300mm^2 == 235 chips per wafer
XSS is 197mm^2 == 358 chips per wafer

If you multiply out yields, and lets say a low yield is 80%
196 (XSX) * 0.8 = 157 chips or 70 USD per chip
235 (PS5) * 0.5 = 117 chips or 93 USD per chip
358(XSS) * 0.8 = 287 chips or 38 USD per chip

So it's not double the cost just because it's 50% yield. But we can see how XSS got to 299 and we have an idea now that PS5 could potentially hover around XSX pricing as a result of that BOM increase.

But final price to the customers is really just about, whatever they want to price it at.
 
Last edited:
yea I think some others have mentioned this, but with respect to this article, Bloomberg Japan is citing that the SoC is the one having yield issues. The article cites a 50% yield on SoC.
That's pretty bad. That's like big bulky expensive Intel CPU chips bad, definitely not the type of yield I expect for a game console.
ah right.

hm....... other than the higher clocks, there shouldn't be anything exotic about the chip production that I can think of that would separate it from that of any other chip on the same node.
 
How does a thread supposedly about "lower than anticipated chip yields affect 15 million production target" devolve into another series of senseless FUD like "8TF console" and "Cerny is lying about clocks"?
And now "Sony is going to fire Cerny"?

Wow..
 
How does a thread supposedly about "lower than anticipated chip yields affect 15 million production target" devolve into another series of senseless FUD like "8TF console" and "Cerny is lying about clocks"?
And now "Sony is going to fire Cerny"?

Wow..


You're free to ignore what you think is FUD, otherwise, it could be useful to the forum to discuss why or why not on a technical basis as opposed to adding to the noise. :)
 
Seems unlikely they'd shoot themselves in the foot so blatently. :p Aside from wafer defects, enabling more CUs would increase power consumption that much more.

You know every generational leader tends to do that at the start of a new generation...:LOL::runaway:
 
ah right.

hm....... other than the higher clocks, there shouldn't be anything exotic about the chip production that I can think of that would separate it from that of any other chip on the same node.
hmm, it might be better to say that clockspeeds itself is not directly the issue with yield. The design of the chip is the cause of yield issues. As the clock rate goes up then things start failing as a result of the design.

There are many factors that effect yield. Physical problems that are highly layout dependent effect yield. Manufacturability issues such as defect density on the silicon, maturity of the process, and effectiveness of design rules also effect yield. Another factor, and the main topic of this paper, is how the design will react to process, temperature and supply voltage variations. The design's sensitivity to these variations is called parametric yield. Increasing parametric yield is a challenge that is within a designer's realm.

Last paragraph on the conclusion here:
Although there are a number of factors effecting yield, increasing parametric yield is within a designer's realm. Given that corners analysis remains the most popular method used to predict and increase yield, running as many relevant corners as possible increases the robustness of the design and therefore is likely to increase the parametric yield. An easy-to-use, automated, repeatable environment of tools would be the optimum solution.
Lowering the clockspeed is an effective way of increasing the yield but improving yield starts at the design.
 
PS4 was pretty perfect in execution. :eek:

What I meant was the switching of generational leaders, because of bad moves made at the beginning of new generation. PS2 > XB360 > PS4 You know, PS3/Cell being to expensive and difficult to use... while XB1 had poor messaging (at the beginning) and a dead weight accessory dragging it down (yes, Don Mattrick).
 
Last edited:
Sonys APU is most likely more expensive then XSX one at this point. One has to wonder how crazy is that each and every chip has to guarantee 2.2GHz clocks. That is insane (and people thought Oberon 2.0GHz was far too optimistic).
It's a little more nuanced than that: all PS5s must behave the exact same no matter what; it's not enough that the chip hit 2.2Ghz. They must sustain them in the same way that Sony sets up governance over clockspeeds for activity levels.
 
I wonder if Sony went full batshit and went with a fully enabled GPU core count (40CUs)!? Naw...

Would that even change anything? Simply put, if they're already targetting power envelops (roughly), so more CUs means eating more power means lower clocks?
 
The article cites a 50% yield on SoC.

So it's not double the cost just because it's 50% yield.
PS5 SOC with 50% yield will cost 2 times than the same SOC with 100% yield.


In fact I am wondering if the yield is right because Bloomberg still predicts PS5 can target $449.
 
You're free to ignore what you think is FUD, otherwise, it could be useful to the forum to discuss why or why not on a technical basis as opposed to adding to the noise. :)

How is "Cerny lies about clocks" a useful discussion to justify lower than anticipated yields?
What is the connection here, in your opinion?
 
PS5 SOC with 50% yield will cost 2 times than the same SOC with 100% yield.


In fact I am wondering if the yield is right because Bloomberg still predicts PS5 can target $449.
Right, but we know you can never achieve 100% yield. So we need to be realistic here. A lower yield may be 80% for instance (perhaps a little higher) but not like 95% for instance.
I tried to be conservative on the numbers.

Prices don't need to reflect yield costs. Sony may be more than willing to bite the bullet knowing that they have a better design coming. It may be important to them to ensure that they continue to hold the lead at all costs. They must hold in the face of aggressive pricing from MS. In the end, it shouldn't matter to us consumers, unless they cannot keep stores filled with stock.
 
Wouldn't that jeopardize backward compatibily because of CU count?

Who knows. I think Sony's decisions surrounding their GPU design is quite questionable. IMHO, I would have forgone BC in favor of more raw GPU compute.
 
Back
Top