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

Nvidia or Intel don't use the adaptive clocking system like AMD uses since around 2013 starting with steam roller. It's a very common technique for PC gamers if they especially use Ryzen CPUs.


IIRC, Voltage for ideal processors wouldn't fluctuate when activity occurs. However, due to ohmic laws it fluctuates when current flows. Voltage drops when current is applied and it rises when the current is stopped. The amount of the drops and rises are bigger when the events occur faster and they are smaller when they occur slower (ohm's law for inductors).
Viewed over a long period of time the overall voltage is pretty much constant, but it's crucial for the processor that the voltage doesn't drop below a minimal threshold to function properly.
To guarantee that the voltage never drops below a threshold the overall voltage for the whole system is in most cases (other techniques or workarounds also exist) set higher than needed for the majority of the time and thus producing more heat than needed (for the majority of the time).
AMD uses a system of adaptive clocking which dynamically adjusts the cycle time (e.g., decreasing the frequency) to tolerate the fluctuation, without increasing voltage or decreasing the threshold barrier significantly. Response latency is critical of an adaptive clocking system and the faster the system can respond, the greater the reduction in voltage and therefore the greater the power savings. The latency for current Ryzen processors are as low as 2 or 3 cycles, so it's really really fast.
Yes but this works on a Windows OS that is designed to work with different HW, different clock and such to give out steady results (more or less, pc gamers are used to have different fps and such frome theyr games). It's the first time this happen into the Console world and also with an OS that is not known to be so much flexible. So the silicon variability is abstracted trough a model that leads to predicted and predictable response.... and that to a degree (also into a long period... years of console life) that I find quite worrying. I mean: all thoose sensors that measure calc activity and lead to the modelled variability in frequency and voltage for how long are reliable ? I find this system too complicated... i trust more in simpler systems. I would never have built into my car such a system....
 
So we're suddenly spitting out FUD in technical discussions using powerpoint slides as sources for die size?

Keep in mind that 11 million is actually roughly what the PS4 managed in the same time frame.
as in end of March they sold 7.5 million units to retail that respective year. Add in WIP you'll probably get a maximum of 10 million chips or so done.
just give it a moment, mods are still sleeping ;)
I think normally shifty would handle this if he were here.
 
That’s not the only reason chips yield poorly. It’s also likely 300 mm^2 or thereabouts. It’s not a 70nm chiplet like AMD is known for.
These are what we heard:

1. Sony can still ship 11 million PS5 before the end of March 2021.

(highest number for any consoles in the first 4~5 months)

2. Yield of SOC is 50% but retail price is still possible to be $399 $449.

(Why is retail price still very good?)
 
I think MS somehow helped Sony with the XSS launch. Games will be designed with built-in scalable design that fits with the weird power/frequency management of ps5... The funny will be to compare XSS to PS5 performances... [emoji1]. That's why I think the ps5de will be at 399 (but almost impossible to find at the beginning) and the regular ps5 will be 549.
Quite the contrary. I'd say PS5 is designed to display plenty of polygons and effects at a slightly lower resolution. For XSS devs will have to make games will less details (think Halo Infinite) running at a higher resolution on XSX. And we know (from Cerny) that games with fewer polygons actually stress more the system then games with more polygons to process.

So designing a game for the new baseline (XSS) will do no favors for PS5.
 
Quite the contrary. I'd say PS5 is designed to display plenty of polygons and effects at a slightly lower resolution. For XSS devs will have to make games will less details (think Halo Infinite) running at a higher resolution on XSX. And we know (from Cerny) that games with fewer polygons actually stress more the system then games with more polygons to process.

So designing a game for the new baseline (XSS) will do no favors for PS5.
Thanks for explaining. Yes it's understandable MS did not want to help Sony with XSS.... your explanation fits more with real life.
 
Bloomberg isn't typically in the rumor business though. They have well-sourced contacts.

I meant more the noise that's been generated from the news. The constant spewing of misinformation form the usual suspects.

I come here for PS5 updates and tech talk, not people banging on about 'Sony dun fkuked up and went OC to close the gap on XSeX - advertised as 10.2TF really a 9TF machine' etc. This is not the thread for that low brow BS which has been discussed to death.
 
I've heard the opposite. Since neither of us can cite a source, we're equally credible. Let's avoid such non sense unless you can corroborate.

hmm...I don't know if AquariusZi is a trustworthy source in this forum?
He said PS5's SOC fixed problems after C0 and will start mass production in May.
 
hmm...I don't know if AquariusZi is a trustworthy source in this forum?
He said PS5's SOC fixed problems after C0 and will start mass production in May.

Nobody is trusthworty on here, we dont even belive Cerny or Spencer, just because (marketing speak) we know better :D
 
A story about yield issues when we're presumably so close to a price announcement? I just see this a clever attempt to raise price expectations.

And it's clever because if the price ends on the higher end of expectations it's less of surprise, but if the price comes in on the lower end of expectations it's more of a surprise. So it's win win really.
 
Full respin for 7nm chip takes a long time. Probably close to 6months. If there was a yield issue this spring it's conceivable the respinned fixed chip would come too late to significantly help the launch considering time is also needed to manufacture millions of chips and build millions of consoles and then deliver those consoles. Perhaps this is why there is talk about air deliveries. Buy some time back as the respinned chips come in late.

Even in worst case it's still more availability than what ps4 had and the problem seems to be already solved(if it really existed).
 
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...

In some semblance of fairness.. If you could only get a CPU and new graphics card every 6-8 years.. how much would you be drilling into the details?

yea and wait to you see what a crazy vr year we are entering

This offers me some hope. Have to wonder if it is going to be dashed by reality but..... Still evil to say that! I was nice and calm before now!
 
Spawn wave mentioned that MS also said that their dies were getting more expensive because of bad yields. Could it be an amd thing?
 
http://www.bloomberg.com/amp/news/a...an-limited-playstation-5-output-in-first-year

In April Sony only targeted 5~6 million PS5 by the end of 2021.

Then we hear Sony plans to raise the quantity to 15 million, but Sony can only produce 11 million.

So Sony has doubled the PS5 production without increasing the retail price. I don’t see what the yield issue is.

In fact I am wondering if TSMC can make another 10 million SOC when Sony suddenly wants to triple the production. It seems much more reasonable TSMC can only produce 5 million more SOCs when Sony asks 10 million.
 
Spawn wave mentioned that MS also said that their dies were getting more expensive because of bad yields. Could it be an amd thing?
Lower yields and bad yields are generally 2 different things.
Yields shouldn't be confused with defects. Defects are silicon defects, and at most make up 5% of the wafer (IIRC or if this number is consistent through every node). The redundant CUs should remove the 5% defect issue. That's why they are there. So if you remove the 5% defect problem, by removing 2 CUs, you've already given up quite a lot.

Remaining yield issues are categorized under parametric yield. This is where the chip is 100% fully functional and performs everything but not at the performance level you want it to. So this is generally where chip makers use binning techniques to release higher and lower spec parts. So the 5700, 5700XT, and 5700XT anniversary for instance are all the same chip. Parametric yield basically stops the chips from moving up the ladder in terms of performance. Effectively those chips become increasingly more unstable as they move up the clocks or just run too hot for the stock cooler provided.

The greatest factors for quickly improving parametric yield is to just run lower clocks otherwise you can improve parametric yield on a respin or something similar.
Typically speaking running lower clocks means lower wattage, lower TDP, lower cooling requirements, and of course, much better yield.

To meet consumer pricing, and thus reduce loss, they should be trying to get yield as close to 100% as they can. A lower yield just means that the expectation could be between 80-90% yield. As long as that is the expectation, then ti's just a lower yield and as long as you are getting that yield as expected nothing needs to change. You proceed as normal.

A bad yield is when it's no where close to expectation.

Hope that helps.
 
Last edited:
http://www.bloomberg.com/amp/news/a...an-limited-playstation-5-output-in-first-year

In April Sony only targeted 5~6 million PS5 by the end of 2021.

Then we hear Sony plans to raise the quantity to 15 million, but Sony can only produce 11 million.

So Sony has doubled the PS5 production without increasing the retail price. I don’t see what the yield issue is.

In fact I am wondering if TSMC can make another 10 million SOC when Sony suddenly wants to triple the production. It seems much more reasonable TSMC can only produce 5 million more SOCs when Sony asks 10 million.
Or - because of covid or yield issues you want to produce as much as possible now and ship out the largest shipment you can. Once that order is out and done, it's very difficult to keep up with demand afterwards. PS4 was very profitable for them because they could produce as much as they sold (as they shipped them). A well oiled machine. But if you have all these weird kinks in the process, it may be better to overbuild now, store them in warehouses (eat additional costs) but ensure that stock is there. Sony doesn't want to be in a position where stores are out of stock for a while. That just gives their competitor a free opportunity at a sale.

If they price the PS5 low, they'll be out of stock quicker than later.
 
Back
Top