Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

[AbsoluteBeginner]

Alternatively, MS produce two units. The small unit is rejected by the market for being not powerful enough, and the large unit is rejected as being too expensive...

Alternatively, Sony considered the possibility of two SKUs and thought it could go badly and a single, mid-range common-ground was the best option for the platform.

No-one actually knows how a business decision is going to turn out until it's executed. Once upon a time, MS thought TVTVTV was a great way to sell a console, and Sony thought there was no upper limit to what their fans would be willing to pay. The choice of single or multiple performance options at launch isn't something revolutionary that one company failed to appreciate and consider an option. Both know what they're doing. Both chose different paths. Neither is inherently better or wrong at this point, and we'll have to see how the next-gen pans out to see who picked right, which is more luck than judgement as technically there's no superior option (otherwise both would be doing the same thing ).

That point stands, maybe market rejects it, but strategy is one that banks on it working, and not failing Simply put, if you have single SKU you have to be aware of lower and higher end, with 2SKUs (in theory), you are having your cake and eating it too.

Again, it might fail and we might look back and say "MS was stupid", but entire point of going 2 SKU strat is to have tailored console for lower end and premium one for top end, not worrying how to satisfy both.

 

[AbsoluteBeginner]

Still mid-range for a 2021 product.

Perhaps, but in 2013 Jaguar was bottom of a barrel type of CPU, while 7850 (let alone 7700) were almost 2 year old AMD mid range GPUs (that retailed for 150-200$, and not 400$ like XT).

2013 :

Notebook CPU
150$-200$ GPU (price in 2012)

2020

Top tier CPU
350$-500$ GPU (40CU and 60CU one)

 

[Deleted member 13524]

I'd not get banned. I'd keep the account and then trumpet my awesomeness when the specs were made official and I was proven right, linking back to all my previous posts where I was correct and naming and shaming all my detractors.

Which is exactly what he's doing, so...

Huh? What's a self-ban then? He requests to have his account banned so he doesn't get dragged into posting when he doesn't want to, and then requests the ban is lifted when he is ready to return?

AFAIK you can't PM anything (like idk racist insults) to banned accounts, nor can he be referenced in further posts, so there's that. He asks for the unban, his account gets reinstated and the inbox is clean.

There is no way semi custom chip info 6 months old is outdated.

So let me get this straight.
The XBOne got 1600MHz CPU / 800MHz GPU clocks announced at a pre-E3 event in May 2013. That was 6 (six) months before launch.
Then in August of the same year, clearly during the later part of their production cycle, they announced revised clocks for the GPU with a 6.6% boost. This was 3 (three) months before release.
Then in September of the same year, they announced revised clocks for the CPU with a 9.4% boost. This was 2 (two) months before release.

But now you're convinced that some tests that occurred in July 2019 for the PS5, 17 (seventeen) months before release, are showing the definitive clocks and WGP count for the consumer version.
The maximum clocks the GPU can ever widthstand will be 2GHz and not even a single extra CU could be turned on.

Yeah that makes total sense.







BTW, a 6.6% boost in clocks (the same percentage Microsoft got in late full production, 3 months before release) over Oberon's 2GHz test would put it at 2132MHz.
38 CUs at 2132MHz would result in 10.37 TFLOPs. 38 CUs could result from cutting either one WGP or two half WGPs. No redundancy on the CUs and same clocks would result in 10.92 TFLOPs.

Imagine that AMD/Sony managed to decrease the amount of redundancy after >18 months of revisions.
Super crazy right? Completely impossible. I must be in denial.

For PS5 to be more powerful, the only realistic option they have is to get an SOC from AMD that's bigger than the XSX one (even more CU's).

Or the XBSX needs very low clocks to fit a big die into a console's TDP, effectively running below 12TF. And/or the PS5 is using all 40 CUs and managed to achieve >2GHz. And/or the PS5 is offloading most RT processing to an external chip, effectively getting more available shader compute throughput when using raytracing. And/or there's a PS5 Pro that wasn't leaked.

There's a bunch of missing info but B3D's microsoft team® is just very quick to jump to conclusions and accuse everyone who's evaluating how all credible sources could be right at the same time as "being in denial reeeeee".
¯\_(ツ)_/¯

 

[AbsoluteBeginner]

Or the XBSX needs very low clocks to fit a big die into a console's TDP, effectively running below 12TF. And/or the PS5 is using all 40 CUs and managed to achieve >2GHz. And/or the PS5 is offloading most RT processing to an external chip, effectively getting more available shader compute throughput when using raytracing. And/or there's a PS5 Pro that wasn't leaked.

There's a bunch of missing info but B3D's microsoft team® is just very quick to jump to conclusions and accuse everyone who's evaluating how all credible sources could be right at the same time as "being in denial reeeeee".
¯\_(ツ)_/¯

There might be info missing, but at least some here are going by what is now confirmed as being AMDs testing documentation of upcoming APUs, instead of "Klee said they are close and going by my arbitrary math 30% is not close therefore this is false" (even though other things are much closer to 10%, then 30%).

Basically, we have a group of people digging through AMDs internal testing leak, and other that put their faith into vague insiders that haven't really broken anything at all (Matt, Klee).

The XBOne got 1600MHz CPU / 800MHz GPU clocks announced at a pre-E3 event in May 2013. That was 6 (six) months before launch.
Then in August of the same year, clearly during the later part of their production cycle, they announced revised clocks for the GPU with a 6.6% boost. This was 3 (three) months before release.
Then in September of the same year, they announced revised clocks for the CPU with a 9.4% boost. This was 2 (two) months before release.

But now you're convinced that some tests that occurred in July 2019 for the PS5, 17 (seventeen) months before release, are showing the definitive clocks and WGP count for the consumer version.
The maximum clocks the GPU can ever widthstand will be 2GHz and not even a single extra CU could be turned on.

I never mentioned upping frequency as impossible, although going from 2.0GHz higher does seem a bit far fetched.

MS, for example, went with 8% higher clocks while having 50% less CUs, so you can argue they always had headroom for a bit more.

As for adding of additional WGPs, therefore creating COMPLETELY different chip 6 months before final tapeout being equated to upping the clocks, I'll pass.

 

[RDGoodla]

Transition will depend on how people perceive the new PS versus the old one, and has nothing to do with XB. Historical evidence on sales versus price suggest a €500 console will suffer slower sales than previous €400 consoles. Why do you think a €500 PS5 will sell faster than the €400 PS4 did, especially when the performance increase is proportionally worse than from PS3 > PS4?

Historically consoles should be sold at $299 to have fast transition. At least before PS3
But PS4 changed the rules. PS4 price was $399 but it also has very good transition speed.

How will be PS5 have quicker transition speed at $499? There are 3 important keys:

1, A very good PS4 Pro Boost mode:
For example, we may have a PS5 to play PS4 Sekiro at locked 60fps with much faster loading

2. Zero loading on PS5 games:
It will be a game changer.

3. SONY has much stronger exclusives than 2013:
SONY has six IPs that can sell more than 10 millions compared to only 1 IP in 2013.
This is the most important factor to convince fans.


Besides transition speed will benefit from weak competitors because core users only have one choice for next-gen (for example, in Japan and Asia).

 

[RDGoodla]

1. Komachi said "E0" revision of a custom SOC. Not "fifth revision" (So it can't be Ariel A0 to OBR C0. It is OBR E0).

2. OBR A0/B0 in github are old chips imo (which should be manufactured in late 2018~early 2019).

 

[iroboto]

I'd not get banned. I'd keep the account and then trumpet my awesomeness when the specs were made official and I was proven right, linking back to all my previous posts where I was correct and naming and shaming all my detractors.

ha, well I don't think you'd put yourself in that position either. So I'm not sure if this is technically true .

Huh? What's a self-ban then? He requests to have his account banned so he doesn't get dragged into posting when he doesn't want to, and then requests the ban is lifted when he is ready to return?

Yes. As I understand it, members will ban themselves as a method of self control. I believe Stinkles would self ban before any major event; I guess the whole loose lips sinks ships bit.

One thing is for certain, MS trademarked "Power your Dreams" logo and put die shot on twitter right before Sony's CES press conference. That to me spells confidence, and since I can only go by what we officially have from AMD's side, then for me its pretty clear.

Yea that's why I continue to encourage everyone to stick with talking shop and not to waste their time picking apart PR or insider claims (also a big reason why I post here and not at other forums). You can't beat physics and even if you are talking up incorrect points, it doesn't matter. Scientists postulate all sorts of theories that get shot down; we all learn from it. Secret Sauce/Magic is really a synonym for, "I don't want to spend the mental energy to figure out how they did it", just call it magical and pretend it's going to beast shit up.

Series S will define the generation.

hmmm. Improbable. If timelines are being followed and regression test information is accurate, Lockhart will not be ready for launch in 2020, so understanding that, it cannot the baseline unit.

Then in August of the same year, clearly during the later part of their production cycle, they announced revised clocks for the GPU with a 6.6% boost. This was 3 (three) months before release.
Then in September of the same year, they announced revised clocks for the CPU with a 9.4% boost. This was 2 (two) months before release.

I can shed some light on this, and I've said it multiple times before, and this will be my last time (for a while).
a) they were super behind in nearly every aspect of Xbox One.
b) E3 demo was entirely faked basically (my understanding, could have been a poor paraphrase), the OS team had to work crazy hours to implement in the OS what they showed on E3 (Kinect gestures etc). OS was shipped literally 1-2 weeks before launch.
c) Hardware was behind, original goal was to have an internal PSU(like the 360 S etc), but due to time limitations and fear of RROD, they rushed the design of Xbox One to be overly cooled.
d) Fan was way over done, PSU moved outside.

Albert Penello then reveals at a much later time
a) you can only change things this close to reveal if you have additional cooling capacity available (check)
b) yields are better on the SOC than expected (check)

the combination of these two together is what allows for last minute upclocks.
he continues with paraphrase: But pricing and design of a console and all it's hardware is all known _well_ in advance. That means every single part, clock, chip, power, cooling, is all known is advance of any sort of production. So you _must_ have designed the console in mind with tolerances put into place to allow for this.

IMO, Xbox One was a Frankenstein of a project. TV, Kinect, Self Published Content, Set Top Device, etc. They were cramming in way too much into this device thinking they could beat out the market (in non gaming categories). When you have so many cooks in the kitchen, you're bound to have inefficiencies and slack of which you can tighten up before launch. Think of it like realizing you cannot have any chance to win the race with your sports sedan against a sports car, so you then tear out the interior that makes it luxurious to reduce weight to make it run faster. When the goals are much more straight forward with less requirements, you're already running super lean. You can't just take a tuned formula 1 engine and with a month before race time expect it to run 2000 RPM higher than redline suddenly without blowing the engine.

So all these things considered; if the price point is going to be very high, I'm not sure if we can expect either companies to build in tolerances. It's probably going to be cut as razor edge as it can be

 

[Deleted member 13524]

"Klee said they are close and going by my arbitrary math 30% is not close therefore this is false" (even though other things are much closer to 10%, then 30%).

30% isn't coming from arbitrary math.
12 / 9.2 = 1.30.
30% isn't close. I'm not convinced Jason Schreier is a math genius but not even he'd say 9 TFLOPs is close to 12 TFLOPs.
The difference between the XBOneS and the PS4 is smaller, at 28.5%, and no one considers them close.

Basically, we have a group of people digging through AMDs internal testing leak, and other that put their faith into vague insiders that haven't really broken anything at all (Matt, Klee).

You mean a group of people with little knowledge of spec changes that happened during regular console development cycles in the past who are convinced that clock values and number of pre-set redundancy units must be written in stone 18 months before the console's release, and the people who are trying to figure out how the target specs given to developers and leaked by trustable sources would fit within what was mentioned in the leak.

MS, for example, went with 8% higher clocks while having 50% less CUs, so you can argue they always had headroom for a bit more.

That's the wrong number but whatever.
Microsoft went with 9.4% higher clocks on the CPU while having the exact same number of cores.
Two months before release.
What's the excuse for that?

As for adding of additional WGPs, therefore creating COMPLETELY different chip 6 months before final tapeout being equated to upping the clocks, I'll pass.

Try reading it again. No one mentioned adding more WGPs.

 

[Shifty Geezer]

Historically consoles should be sold at $299 to have fast transition. At least before PS3
But PS4 changed the rules. PS4 price was $399 but it also has very good transition speed.

How will be PS5 have quicker transition speed at $499? There are 3 important keys:

It doesn't matter what features you have if people can't afford your product. For comparison, PS3 offered HDD as standard, network play as standard, a BRD player at the lowest p[rice you could get a BRD player, HD as standard...basically everything modern tech had to offer back then, and it sold poorly due to the high price. There's no way a $500 PS5 is going to sell faster than a $400 PS4 because of instant loading and PS4 games at 60 fps... Those must-have games won't be there at launch and potential buyers will be able to wait until PS5 is cheaper before buying one to play Grandad Of War etc.

 

[AbsoluteBeginner]

1. Komachi said "E0" revision of a custom SOC. Not "fifth revision" (So it can't be Ariel A0 to OBR C0. It is OBR E0).

2. OBR A0/B0 in github are old chips imo (which should be manufactured in late 2018~early 2019).

No, Komachi made a mistake which is easy to explain.

This is his own Excel spreadsheet :

ARL A0 - 00840F00
ARL E0 - 00840F40 (note ?)

So, he logically assumes Ariel went from A0 (F00) to E0 (F40) because code "jumped" from 00 > 10 > 20 > 30 > 40 (so, 5 steppings). First number signifies bigger change (aka letter, ex - A0 > B0), while last number signifies smaller change (aka number, ex - B0 > B1).

This all makes complete sense, because obviously, 00840F40 = 13F9 and 13F9 is part of 13FF range ie. Ariel (https://pci-ids.ucw.cz/read/PC/1022/13e9)

Well, kinda, because 13F9 also indicates Oberon, as they are SAME chip. Dont believe me? Look! This is just the other day when he said "SOC going to E0 stepping is unexpected".

https://twitter.com/i/web/status/1214190092316749829

And this is Oberon :

See 13F9 : OBR : A code? Its Oberon chip, likely from summer when Flute was tested. Its Oberon A0 stepping. We know there is B0 as well, and since they have same ID code (Ariel and Oberon), we can assume what Komachi thought to be E0 stepping of Ariel is actually C0 stepping of Oberon (Ariel A0/B0 and Oberon A0/B0/C0).

 

[Scott_Arm]

Lol, this is nuts.

 

[Shifty Geezer]

Again, it might fail and we might look back and say "MS was stupid".

That'd be wrong. It's not a stupid move. It may not be the right one, but it's not stupid, and people shouldn't be called out as being stupid on account of being wrong. Heck, you can be stupid and right, like Wii.

 

[AbsoluteBeginner]

30% isn't coming from arbitrary math.
12 / 9.2 = 1.30.
30% isn't close. I'm not convinced Jason Schreier is a math genius but not even he'd say 9 TFLOPs is close to 12 TFLOPs.
The difference between the XBOneS and the PS4 is smaller, at 28.5%, and no one considers them close.

You are trying to fit your arbitrary "close" or far apart definition to last gen, when TF difference between PS4 and Xbone after patch was 40%. Before it - 50%. Not to mention all other advantages PS4 had over XBone that were much bigger then Arden > Oberon.

You mean a group of people with little knowledge of spec changes that happened during regular console development cycles in the past who are convinced that clock values and number of pre-set redundancy units must be written in stone 18 months before the console's release, and the people who are trying to figure out how the target specs given to developers and leaked by trustable sources would fit within what was mentioned in the leak.

"Leaked trustable sources". Says who? Again, arbitrary definition. These sources haven't said much. They didn't put money where they mouth is. They haven't put specs black on white, so there is nothing to really talk about. Some people will tell you Klee went quite after the leak and asked for his ban shortly after, others will tell you he was "abused", even though his word was never to be questioned there.

That's the wrong number but whatever.
Microsoft went with 9.4% higher clocks on the CPU while having the exact same number of cores.
Two months before release.
What's the excuse for that?

Already told you, excuse is having 12CUs instead of 18CUs, as well as half the ACEs and DDR3 RAM instead of GDDR5. They most certainly had TDP headroom to push clocks further, especially on CPU side that uses much less electricity anyway. Even after these upclocks Xbone pulled less watts then PS4, therefore explanation is rather easy really.

Try reading it again. No one mentioned adding more WGPs.

You literally said above

"some tests that occurred in July 2019 for the PS5, 17 (seventeen) months before release, are showing the definitive clocks and WGP count for the consumer version.
The maximum clocks the GPU can ever widthstand will be 2GHz and not even a single extra CU could be turned on."

Yes...? Going by test which says "full chip used" in 18WGP mode and specifying formula for GT/s as "2*4*Sum_all_SX_channel" I will make a leap of faith and say, its 2 * 10WGP GPU based on Navi 10, like both Sony predecessors.

 

[PSman1700]

There's a bunch of missing info but B3D's microsoft team® is just very quick to jump to conclusions and accuse everyone who's evaluating how all credible sources could be right at the same time as "being in denial reeeeee".

And Klee is to be trusted, right?

 

[cheapchips]

It doesn't matter what features you have if people can't afford your product. For comparison, PS3 offered HDD as standard, network play as standard, a BRD player at the lowest p[rice you could get a BRD player, HD as standard...basically everything modern tech had to offer back then, and it sold poorly due to the high price. There's no way a $500 PS5 is going to sell faster than a $400 PS4 because of instant loading and PS4 games at 60 fps... Those must-have games won't be there at launch and potential buyers will be able to wait until PS5 is cheaper before buying one to play Grandad Of War etc.

I think Sony overestimated how much Bluray was worth to people. At launch, PS3 was like paying $100-$200 extra for what could have easily been the new betamax.

It's been Sony's only misstep on pricing so far. PS4 at £399 managed to look like a bargain next to XBone at $499. The original Playstation at $299 managed to look like a bargain next to the Saturn at $399. The $199 Dreamcast wasn't good competition to $299 PS2. That extra $100 for a DVD player was fantastic value in 2001.

 

[RDGoodla]

It doesn't matter what features you have if people can't afford your product. For comparison, PS3 offered HDD as standard, network play as standard, a BRD player at the lowest p[rice you could get a BRD player, HD as standard...basically everything modern tech had to offer back then, and it sold poorly due to the high price. There's no way a $500 PS5 is going to sell faster than a $400 PS4 because of instant loading and PS4 games at 60 fps... Those must-have games won't be there at launch and potential buyers will be able to wait until PS5 is cheaper before buying one to play Grandad Of War etc.

Um I don't think people can't afford another $100. If people are interested in those features they will pay another $100.

Imo we should investigate why PS4 can be sold very well at $399 while previous data showed $299 was most.
The most important reason for less sales of PS3 was the lack of big games at launch. In 2006 few games are cross-generation games. And xbox 360 had Gears of War in 2006. In 2013 AAA titles are almost all cross-gen so PS4 have a lot of big titles at launch.

If PS4 can be sold at a lot at $399 what is the highest price acceptable?

 

[mpg1]

Or the XBSX needs very low clocks to fit a big die into a console's TDP, effectively running below 12TF. And/or the PS5 is using all 40 CUs and managed to achieve >2GHz. And/or the PS5 is offloading most RT processing to an external chip, effectively getting more available shader compute throughput when using raytracing. And/or there's a PS5 Pro that wasn't leaked.

There's a bunch of missing info but B3D's microsoft team® is just very quick to jump to conclusions and accuse everyone who's evaluating how all credible sources could be right at the same time as "being in denial reeeeee".
¯\_(ツ)_/¯

"hey I'm not in denial but here's this highly convenient(unbelievable) scenario where PS5 ends up more powerful".

 

[Globalisateur]

No, Komachi made a mistake which is easy to explain.

This is his own Excel spreadsheet :



ARL A0 - 00840F00
ARL E0 - 00840F40 (note ?)

So, he logically assumes Ariel went from A0 (F00) to E0 (F40) because code "jumped" from 00 > 10 > 20 > 30 > 40 (so, 5 steppings). First number signifies bigger change (aka letter, ex - A0 > B0), while last number signifies smaller change (aka number, ex - B0 > B1).

This all makes complete sense, because obviously, 00840F40 = 13F9 and 13F9 is part of 13FF range ie. Ariel (https://pci-ids.ucw.cz/read/PC/1022/13e9)

Well, kinda, because 13F9 also indicates Oberon, as they are SAME chip. Dont believe me? Look! This is just the other day when he said "SOC going to E0 stepping is unexpected".

https://twitter.com/i/web/status/1214190092316749829

And this is Oberon :



See 13F9 : OBR : A code? Its Oberon chip, likely from summer when Flute was tested. Its Oberon A0 stepping. We know there is B0 as well, and since they have same ID code (Ariel and Oberon), we can assume what Komachi thought to be E0 stepping of Ariel is actually C0 stepping of Oberon (Ariel A0/B0 and Oberon A0/B0/C0).

Lol you really are bothered about that odd E0 chip. It has to be some kind of translation error from Komachi. I can't exist at all, it's a C, not a E !

Maybe you could ask yourself some interesting questions instead of doubting the source ? if it's the fifth chip, why didn't they call it C0 ? Why the odd E letter instead of C ? Either there are many others unknown previous revisions (how did Komachi miss them ?)... or they chose E after B because the letter 'E' means something. Maybe.

 

[BRiT]

Crazy speculation, so hang on ... if Microsoft has multiple SKU Xbox Series Console then Lockhart will not play a part in it. The multi-sku will come from binning of Anaconda (WGP/SE/CU counts and Speed).

 

[upnorthsox]

Ok so now that we've established that 30% doesn't mean anything, we can officially declare that a 5700XT and a 2080ti are the basically the same. I'm sure the other side of the forums are nodding in knowing agreement.