• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.

Bluntman

Member
Anyone interesting in the next-generation of discreet GPU's from AMD should watch this video, interesting stuff. Apparently "some" RDNA 2 cards will be hitting the same clock speeds as the PS5 GPU which is 2.23 Ghz, however the cards can't seem to go over that without having issues like logic breakdown, same reason to why the PS5 was not clocked higher than 2.23 Ghz as Mark Cerny pointed out, this seem's to be a constraint of the RDNA 2 architecture.

Another leak mentioned that the RDNA 2 GPU's will have an "Infinity Cache", this will be around 128 MB of cache memory and will apparently mitigate the slower memory bandwith. Interestingly enough RGT has mentioned a few times now that the PS5 will not have bandwidth issues and this is what he's heard from developers, this is due to its "robust caching system".



If you mean someting like a 128 MB of single SRAM cache then no, we would have seen that on the XSX die shot. It's possible that the PC RDNA 2 will have it, but unlikely.

If you mean something like 128 MB of caches all around the APU, that's more likely, but still sounds waaaay too much.
 

ArcaneNLSC

Member
Fake.

Needs to be blurry for legitimacy.

Maybe jump while taking the picture. Or run from side to side.

I wonder what the people inside looking at you doing so would be thinking.


HAHAHA sorry I even took the photo on my shitty phone camera not my brand new phone I just got. Didn't notice or give a shit if people thought I was doing odd shit
 

geordiemp

Member
Your only argument is that it comes from an unknown source towards a journalist. That's a lot of times how journalism works. They got anonymous sources that tell them things.

And my other argument is there are many customers for that fab, funny how it just affects 1 customer.

Funny the FUD, first it was yields are so good sony have 38 CU, now yields are so bad they cant make any.

Its FUD. Parametric Yields with a known normal distribution and 3 sigma well defined dont suddently go tits up day before a reveal.

Particulate yield will be size dependent of die.

There are no unknowns in well defined statistics, only possibility in fabs is if a tool beaks or goes down, PM, and that would affect all customers for the outage.

STOP TALKING CRAP. I work in this field, its fucking embarrasing, its clear the rumour has no idea what they are talking about.
 
Last edited:

Brudda26

Member
To be seen. It's not normal what is happening at the moment. Yields are pretty low.
The yields were initially 50% this would of been when production started, article already mentions yields have since improved and are stabilising. If you want comparison the PS3 had a usuable yield of between 10-20% initially for the CELL. It is normal for new tech on a not fully matured process. TSMC 7NM is good yes but even then the yield rate for its 7nm for graphics purposes is only 80% this is across the board including the high end and low end. The high ends yields would be lower and the low end yields higher. Nothing ever hits 100% yields and it's incredibly hard to hit 90% yields. Consoles have even tighter restrictions as they have customisations to the silicon meaning they have to be even more strict on what silicon gets through.

TLDR: yields were initially 50% have since improved shouldnt be of concern as this will not effect the quality of silicon that ends up in the consumer products because every bit that is classed as usable for PS5 will have gone through stress testing etc to make sure it meets the standards needed.
 
32e742eda80baca3298e2524d2c2a302c18a7046.gifv


Waiting patiently for tomorrow. I predict shitstorms.
 

geordiemp

Member
The yields were initially 50% this would of been when production started, article already mentions yields have since improved and are stabilising. If you want comparison the PS3 had a usuable yield of between 10-20% initially for the CELL. It is normal for new tech on a not fully matured process. TSMC 7NM is good yes but even then the yield rate for its 7nm for graphics purposes is only 80% this is across the board including the high end and low end. The high ends yields would be lower and the low end yields higher. Nothing ever hits 100% yields and it's incredibly hard to hit 90% yields. Consoles have even tighter restrictions as they have customisations to the silicon meaning they have to be even more strict on what silicon gets through.

TLDR: yields were initially 50% have since improved shouldnt be of concern as this will not effect the quality of silicon that ends up in the consumer products because every bit that is classed as usable for PS5 will have gone through stress testing etc to make sure it meets the standards needed.

If there was a yield issue out of the normal expected statistical distribution, TSMC give guidance to the market formerly like below


If rumour were true there would be a statement, TSMC are stating 70 to 80 % yields for 7nm, and thats for everyone, smaller chips have less particulate bins.

There is no gossip about billion dollar issues, they are given guidance.
 
Last edited:

DaGwaphics

Member
How would down clocking improve yields? Less active CUs I get it.... 🤦🏻‍♂️

I don't think PS5 clocks have changed at all, but clock speed can have a lot to do with yields. I've seen some here downplay that, but the reality is that each design will have a target for both clock speed and max power draw. I remember when Intel publicly released that less than 50% of their chips could hit both the clock and power draw target, hence why they bin to ensure no usable chips get left behind. Many binned chips will be defective in some way, but there will be a lot of perfect chips that just couldn't operate at the desired frequency without excessive power draw.

This isn't as much of an issue in the consoles because neither manufacture would push the limits too far and the systems will be designed to accommodate some variance in power draw (what MS calls their Hovis Method - they really will name everything).
 
If there was a yield issue out of the normal expected statistical distribution, TSMC give guidance to the market formerly like below


If rumour were true there would be a statement, TSMC are stating 70 to 80 % yields for 7nm, and thats for everyone, smaller chips have less particulate bins.

There is no gossip about billion dollar issues, they are given guidance.
One is an issue (caused by TSMC), the other is a single chip design issue that will not be of any financial impact for the TSMC top line. They get paid anyways.
 
You are talking rubbish.

No, that's how the business works.
And that's why you don't read such statements from TSMC just because one customer has some design issues. (and plenty had in the past and will continue in the future)
They don't matter for TSMC.

But if all their wavers suck, because they have bad materials, aka they fuck up their production, they have to issue a note for shareholders.
 
Last edited:

geordiemp

Member
No, that's how the business works.
And that's why you don't read such statements from TSMC just because one customer has some design issues. (and plenty had in the past and will continue in the future)
They don't matter for TSMC.

But if all their wavers suck, because they have bad materials, aka they fuck up their production, they have to issue a note for shareholders.

Your still talking crap, in the first 10,000 units the statistical parametric yeilds are known and thats it. There are no surprises in high volume 3 sigma statistics.

Particulate is well known and its a functyion of die size.

So any deviations from that first statistically significant yield will be a materials or tools issue and will be all customers including apple.

Facts are difficult.

In semiconductor, you dont make 10,000 wafer starts, get a yield, and then the next 10,000 it all falls apart just for that design - thats idiotic.
 
Last edited:

Darklor01

Might need to stop sniffing glue
It’s always from the same people, there’s always this pattern in them. You gotta give them credit for manipulating their audience into believing their fud

It is usually the same people. I find it harder when they are quoting a reporter from a news source, in this case, Bloomberg. There is an issue there because there are times info from Bloomberg about gaming is legitimate. How does the gross majority know what to believe.

Combine that with the fact that TSMC say that they're having better yields with 5nm than 7nm, and commentary by Intel, and suddenly there is this credibility to it.

The issue is, as a bunch of people point out... why is Playstation 5 singled out and XSS and XSX not mentioned? Was it because there was no original article claiming how many the XBOX team plan to have produced? No way to know. Additionally, as some point out, the timing on this "news" is awfully coincidental to when Sony is going to have a streamed event.

This is exactly how FUD is spread. It's textbook. Maybe it's legitimate, maybe it's not, but it is exactly how it's done. Practically a master class in it.
 
If there was a yield issue out of the normal expected statistical distribution, TSMC give guidance to the market formerly like below


If rumour were true there would be a statement, TSMC are stating 70 to 80 % yields for 7nm, and thats for everyone, smaller chips have less particulate bins.

There is no gossip about billion dollar issues, they are given guidance.
EnchantingExemplaryHoverfly-size_restricted.gif
 
just a clarification, this productions yields could only get 11 million and was cut short from 15 million by 4 million. Does this indicate anything about defect failure of 30% for the PS5
Dude, the yield has nothing g to do with failure of final product
zero relation . Chips than do not pass certification they r rejected and not put into the final product hence lowering the yield.
 
Your still talking crap, in the first 10,000 units the statistical parametric yeilds are known and thats it. There are no surprises in high volume 3 sigma statistics.

Particulate is well known and its a functyion of die size.

So any deviations from that first statistically significant yield will be a materials or tools issue and will be all customers including apple.

Facts are difficult.

So that's why every single design has the same yield and defect rates. Because they're all the same.
Sure that makes sense... I learn something new every day.

No there are no surprises, they know they have bad yield because of the design.
I never said that this is a surprise issue for Sony. It's just Bloomberg trying to sell it that way.
 
And my other argument is there are many customers for that fab, funny how it just affects 1 customer.

Funny the FUD, first it was yields are so good sony have 38 CU, now yields are so bad they cant make any.

Its FUD. Parametric Yields with a known normal distribution and 3 sigma well defined dont suddently go tits up day before a reveal.

Particulate yield will be size dependent of die.

There are no unknowns in well defined statistics, only possibility in fabs is if a tool beaks or goes down, PM, and that would affect all customers for the outage.

STOP TALKING CRAP. I work in this field, its fucking embarrasing, its clear the rumour has no idea what they are talking about.

Yep, the '50% yields' is fucking hilarious nonsense, that would be an absolute disaster for TSMC on their 7nm or 7nm enhanced node, no-one would use it and you don't just find this out 2 months before release day!
 

chilichote

Member
Especially as they went as far as branding them as a "Series". I would have expected more homogeneous design aesthetics. Odd indeed.
A halved Series X would have been absolutely right and I'm wondering why Microsoft decided to go the ugly way. Maybe to look more traditional with the Series S?
 

AeneaGames

Member
I don't believe the ps5 will play most of the 2860 ps4 games at least not for a while.

Such a waste of all that extra engineering Sony did to alter the PS5's GPU to include hardware BC modes for PS4 at the PS4's clock frequency, a Pro mode with the Pro's clock frequency and a PS5 clock speed boost mode when now according to you the games won't run, such a waste....

In other words: ALL PS4 games will at least work without any testing by Sony in the PS4 BC mode. If the game has Pro support it will run without needing to be tested in the Pro C
BC mode.

And Sony is testing all the games to see if they ALSO work in the PS5 boost mode, this is the only thing they need to test the games for.

Of course there will also be games that will get a PS5 patch which makes them PS5 games and don't need to run in a BC mode...

This was all made clear in Cerny's presentation but sadly people misunderstood what he said which caused wrong info to be spread.
 
Last edited:

NoviDon

Member
...Company announces a reveal date of their highly anticipated product->Scandalous rumor pops out of nowhere -> Gafers take it as gospel and freak out ->impending battle for the honor of your fandom breaks out->the truthfacts get revealed that completely shutdown the rumor ->Rumors continue to "leak" out, until the product is finally revealed-> Company announces reveal date for highly anticipated product...I'm think I'm seeing a pattern here..🤔
 
Status
Not open for further replies.
Top Bottom