• 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.

PlayStation 5 9.2 Teraflops Leaked Figure May Not Be Accurate, Insider Suggests – Rumor

Status
Not open for further replies.

R600

Banned
Still waiting for that GitHub link everyone keeps talking about and basing on.
What's the story about that anyway?
Some poor souls at AMD and their outsourced ASIC engineer firm put out bunch of regression tests on upcoming APUs and GPUs from AMD on Github.

Unfortunately for them, seems like Xbox and PS5 chis were there so we got alot of info on chips dating as late as June. So people are speculating if this is too old and if things have changed...
 

Gavin Stevens

Formerly 'o'dium'
I swear people exaggerate about how loud the Pro is. I have a heavy overclock on my gpu and have my fans set to 90 to cool it. It sounds like someone turned on a vacuum when i start the overclock, and 5mins into gaming, i dont even hear it anymore.

Consoles are usually quiet little machines under your tv. My GPU/CPU kicks out some heat and so it’s louder, but you would expect that. It cost 4x the cost of these consoles and runs like a dream compared to them.

But the Pro? That’s something else, it was loud as hell, easily the loudest and most annoying console I’ve ever heard. And what makes it worse is when you stick it next to an X, a console that’s more powerful but utterly silent... You start to say “well why should I have to suffer with this?”
 
I honestly think the ones who will be disappointed are those who will be expecting to get a 2080TI machine, with 24GB RAM, and 1TB SSD for under £500.

I fully expect PS5 to be the equivalent of a 5700XT, coupled with 16GB RAM.
yep.
5700XT is 7nm DUV. EUV should give 20% savings out of the box.
We still don't know about EUV.
Yeah. This didn’t need its own thread. A single post in the ongoing next gen thread at best. Since nothing was actually said.
This thread is a summary for people that doesn't follow the other thread since it's crowded with console war and people don't wanna waste their time reading that nonsense.
 

CeeJay

Member
That Wikipedia page on stepping must have got a shit ton of hits recently, I had never heard the word "stepping" mentioned on this forum until the last couple of days and now everyone's an expert!
 
yep.

We still don't know about EUV.

This thread is a summary for people that doesn't follow the other thread since it's crowded with console war and people don't wanna waste their time reading that nonsense.
All of it is nonsense at this point. No need for summarizing when nothing was said.

But carry on. Just my opinion. People can do as they like.
 

Nikana

Go Go Neo Rangers!
I swear people exaggerate about how loud the Pro is. I have a heavy overclock on my gpu and have my fans set to 90 to cool it. It sounds like someone turned on a vacuum when i start the overclock, and 5mins into gaming, i dont even hear it anymore.
I don't know about that. The PS4 in general seems to vary vastly depending on a lot of factors. I have a pro but my entertainment center is very open. Adequate space on all 4 sides and the back is open. The fan gets going but its not that loud.

My buddy has a ps4 slim and his has good ventilation on the sides but not the back and his slim get louder than my pro.

Those things are just a crapshoot.
 

Grinchy

Banned
I swear people exaggerate about how loud the Pro is. I have a heavy overclock on my gpu and have my fans set to 90 to cool it. It sounds like someone turned on a vacuum when i start the overclock, and 5mins into gaming, i dont even hear it anymore.
I have one of those newer models that supposedly just allow themselves to get hot instead of getting loud, but yeah, I never hear the damn thing unless it's spinning a disc.

I see people talking about the X being loud too, and I am always surprised. I have one of those Project Scorpio ones from the launch here and if it has ever made a peep, I'd need a dog's ears to hear it. It's the quietest piece of high tech I've ever encountered.
 

R600

Banned
You don't need to thing so it is on AMD documentation... it is a fact.

It is the same chip with different steeping and E0 added 2 cores.

All C2/C4 stepping are 4 cores.
All E0 steeping are 6 cores.

You said you never saw that so I give you one example... you can find more if you look on google.
No, I dont think thats the case. Your theory would make every single chip from AMD document essentially the same, as there are Athlons and Opterons with E0/C2 stepping.

What is likely the case here is, steppings clearly define (as per AMD table) slecific fixes. There are no Deneb E0 steppings, as the last one is C2, but there is E0 stepping for Thuban (6 core) and Zosma (Thuban with 2 cores deactivated).

They are new chips in Phenom family with specific codenames and changes to hardware level. They were released 1 year later then Deneb, and as such have fixes beyond C2/C3 steppings - hence E0 and new chips.

Basically,

Deneb 2009 C2 4 core chip
Thuban 2010 E0 6 core chip

Oberon 2019/06 B0 36CU chip
Oberon 2019/12 E0 60CU chip

Dont think thats the case, especially with info coming from the guy that leaked this first, AquariusZi, who said Oberon is going through constant steppings and is one size smaller then Arden.

If upgrade was additional 50-60mm2 I think he would not say its smaller chip. But lets wait and see :)
 

Clear

CliffyB's Cock Holster
People are inevitably going to be disappointed once the actual figures are revealed because there is no fucking way on Earth a console APU is going to outmatch a high-end desktop GPU from the same supplier that retails for a far higher price.
 

ethomaz

Banned
No, I dont think thats the case. Your theory would make every single chip from AMD document essentially the same, as there are Athlons and Opterons with E0/C2 stepping.

What is likely the case here is, steppings clearly define (as per AMD table) slecific fixes. There are no Deneb E0 steppings, as the last one is C2, but there is E0 stepping for Thuban (6 core) and Zosma (Thuban with 2 cores deactivated).

They are new chips in Phenom family with specific codenames and changes to hardware level. They were released 1 year later then Deneb, and as such have fixes beyond C2/C3 steppings - hence E0 and new chips.

Basically,

Deneb 2009 C2 4 core chip
Thuban 2010 E0 6 core chip

Oberon 2019/06 B0 36CU chip
Oberon 2019/12 E0 60CU chip

Dont think thats the case, especially with info coming from the guy that leaked this first, AquariusZi, who said Oberon is going through constant steppings and is one size smaller then Arden.

If upgrade was additional 50-60mm2 I think he would not say its smaller chip. But lets wait and see :)
These I listed are the same with new stepping... there are different tables for different chips in the AMD doc.

Oberon I don't know because I don't even know what it is.... there is rumors it is the PS5 chip.
 
Last edited:

joe_zazen

Member
Github Oberon is stepping B0 updated in June.

Only credible source of cancellation for 2019 is from Jason, who said it was scrapped at beginning of 2017. They had no chip then so makes 0 sense to have a 36CU one in June, wgich was sent in V dev kits, just to have totally different design 6 months later in time of tapeout (which, coincidentaly, has a same codename but different stepping).

If there is Oberon E0 waiting for tapeout now, and we know what Oberon B0 looked like 6 months ago, there is only one result IMO and that is 36CU one. But lets wait and see...

They did get chips in 2018 according to semiaccurate, and in quantities indicating kit creation. How this fits into the story I am not sure now.

I believe the github data, and if the tests are data from june, then Sony is using a 36 CU part. That means either 2x36 for competitive machine and better VR or max 8 TF machine. DrKeo explained why the 5700xt heat/power/energy info we have is not applicable to console apu:

Summary:
That's why these tables and graphs are irrelevant to consoles. There are no different SKUs to utilize the lower quality chips and the graphs are made based on high-quality chips that provide figures that most of the chips on the wafer just can't reach, thus irrelevant to console makers

full explanation: https://www.neogaf.com/threads/next...-leaks-thread.1480978/page-507#post-256433710

There is no way people would be saying an 8TF machine is close or above the xsx. A 36CU apu is 100% underpowered and max 8tf.



That's possible, but something tells me Sony hasn't done that. It would be similar to what SEGA did back in '93 when they got word of the PS1's specs and panicked, throwing in the 2nd CPU and such. Which screwed up the balanced architecture they already had, incurred additional costs they had to eat up on R&D and systems produced, and forced them to run behind on SDKs for devs (among other issues).

Sure Sony has the money to throw out all the progress on Oberon and start again, but that would be them throwing out 3 years of development, needing to re-architect everything from scratch, throw billions of dollars down the drain (and invest yet more billions for new R&D), re-sort contracts for components, and squeeze all of this and testing in for the new chip in just under two years.

I can't imagine what a resource and manpower drain that would be the PS division. Like I said, it's theoretically possible Sony may've done that, but at what cost? I definitely think they made some other types of adjustments though; again, guys like Shawn Layden left suddenly, there were tons of rumors for a 2019 release, etc.

Most realistic possibility is that they've been fine-tuning the hell out of Oberon and trying to squeeze as much performance out of it as possible. They still have roughly up to E3 before things need to be absolutely finalized (I guess) though.

well, this is the conundrum: can we make 36CUs from rdna first gen on 7nm into a powerful console? Simply upping the clocks to get 9or 10 TF means a weaker machine that makes more heat and uses more energy than the competition and throwing out a hell of alot of dies from the wafer. Expensive apu, noisy, underpowered, hot, and energy guzzling....

I prefer either secret chip theory or 2x36, otherwise ps5 is garbo.
 
Last edited:
They did get chips in 2018 according to semiaccurate, and in quantities indicating kit creation. How this fits into the story I am not sure now.

I believe the github data, and if the tests are data from june, then Sony is using a 36 CU part. That means either 2x36 for competitive machine and better VR or max 8 TF machine. DrKeo explained why the 5700xt heat/power/energy info we have is not applicable to console apu:

Summary:

full explanation: https://www.neogaf.com/threads/next...-leaks-thread.1480978/page-507#post-256433710

There is no way people would be saying an 8TF machine is close or above the xsx. A 36CU apu is 100% underpowered and max 8tf.





well, this is the conundrum: can we make 36CUs from rdna first gen on 7nm into a powerful console? Simply upping the clocks means a weaker machine that makes more heat and uses more energy than the competition and throwing out a hell of alot of dies from the wafer. Expensive apu, noisy, underpowered, hot, and energy guzzling....

I prefer either secret chip theory or 2x36, otherwise ps5 is garbo.
Having 2 chips is unlikely, 8TF is not happening, we still don't know how RDNA2 performs against RDNA1, 9.7TF is the minimum guess.
 

R600

Banned
What is happening with Klee? Reading Ree seems like he has had enough.

For folks here, 5700 (7.9TF) > Vega 64 by 10% (12.6TF card). Keep your expectations in check.
 

joe_zazen

Member
Having 2 chips is unlikely, 8TF is not happening, we still don't know how RDNA2 performs against RDNA1, 9.7TF is the minimum guess.

i guess that is another option: ps5 is rdna2 on euv (7nm+) while MS is stuck with rdna1 on 7nm. If so, Sony has out manoeuvred MS big time.
 

GymWolf

Member
I swear people exaggerate about how loud the Pro is. I have a heavy overclock on my gpu and have my fans set to 90 to cool it. It sounds like someone turned on a vacuum when i start the overclock, and 5mins into gaming, i dont even hear it anymore.
Same (except the heavy overclock), never had a problem with loud consolles or pc fans...
The volume of the game is always louder and when i play at night i use headphones...
 

ANIMAL1975

Member
Why are people still holding on to the github leak that was immediately discredit for its flaws? _ it wasn't a datamine, it had stupid things like a shopping list,... and even you don't trust the 'so called insiders' like Klee (and Matt by the way!) that said that nothing has changed by their sources... We got the Komachi tweet, quickly deleted, about the finding of a abnormal Oberon E0 by is words.
 

R600

Banned
Why are people still holding on to the github leak that was immediately discredit for its flaws? _ it wasn't a datamine, it had stupid things like a shopping list,... and even you don't trust the 'so called insiders' like Klee (and Matt by the way!) that said that nothing has changed by their sources... We got the Komachi tweet, quickly deleted, about the finding of a abnormal Oberon E0 by is words.
Abnormal as in - how the fuck they havent fixed everything yet by E0.

In any case, not much to believe now...Klee has another melodramatic post asking for ban and he wont be coming to Ree anymore. Matt said one unambiguous sentace regarding the leak and bounced of.
 

Ogbert

Member
I swear people exaggerate about how loud the Pro is.

I think a lot of it depends on your personal situation. If you have a nagging wife and kids asleep in the next room, you can't drown out the noise with the volume.

I think the Pro is loud - and getting louder. But it's perfectly understandable.
 
They did get chips in 2018 according to semiaccurate, and in quantities indicating kit creation. How this fits into the story I am not sure now.

I believe the github data, and if the tests are data from june, then Sony is using a 36 CU part. That means either 2x36 for competitive machine and better VR or max 8 TF machine. DrKeo explained why the 5700xt heat/power/energy info we have is not applicable to console apu:

Summary:

full explanation: https://www.neogaf.com/threads/next...-leaks-thread.1480978/page-507#post-256433710

There is no way people would be saying an 8TF machine is close or above the xsx. A 36CU apu is 100% underpowered and max 8tf.





well, this is the conundrum: can we make 36CUs from rdna first gen on 7nm into a powerful console? Simply upping the clocks to get 9or 10 TF means a weaker machine that makes more heat and uses more energy than the competition and throwing out a hell of alot of dies from the wafer. Expensive apu, noisy, underpowered, hot, and energy guzzling....

I prefer either secret chip theory or 2x36, otherwise ps5 is garbo.

I still think a 36/40 CU chip is going to be somewhat cheaper than a 56/60 CU chip no matter how much they push the clocks. And again, going by the design of the dev kit, that thing seems built for plentiful cooling so I don't think it would be very loud. As for the energy part, it really depends on how much power the other system components draw.

One of the other consistent things in the rumors is PS5 having a lower CPU clock which would help (somewhat) in the power draw. Same with what specific GDDR6 modules they choose, the type of SSD they go with, the number of ports on the machine, etc.

My own "secret chip theory" is just in regards to RT; Sony might've just gone with a custom RT architecture and implementation, collab with AMD to spec and build it out into the APU. But 2x36 will probably bump whatever price they had in mind for PS5 originally by $150 or so (the additional cooling, larger PCB, more complex traces on the motherboard, the 2nd GPU itself etc.), that doesn't seem like something Sony would revert back to imo; it doesn't fit their historical console design philosophy.

Why are people still holding on to the github leak that was immediately discredit for its flaws? _ it wasn't a datamine, it had stupid things like a shopping list,... and even you don't trust the 'so called insiders' like Klee (and Matt by the way!) that said that nothing has changed by their sources... We got the Komachi tweet, quickly deleted, about the finding of a abnormal Oberon E0 by is words.

It wasn't discredited; in fact it being pulled down so quickly is what helped lend it some credibility. Klee and Matt are free to go with their sources but people are free to take it all with a grain of salt and have justified reasons to do so.

The Komachi tweet you're talking about used "abnormal" in the context of there being so many steppings for the chip, not that the chip mutated into some other creature.
 
Last edited:

Mod of War

Ω
Staff Member
Please use this thread. We don't need a new thread every time an insider squeaks.

 
Status
Not open for further replies.
Top Bottom