Plot twist: Sony actually making a PS5 for 1080p and a PS5 Pro for 4k.
Plot twist: 8TF chip is for portable version of PS5
Plot twist: Sony actually making a PS5 for 1080p and a PS5 Pro for 4k.
Looking at Sony's first party software pipeline, I doubt the decision to move PS5's release to 2020 was done in 2018. Also after the One X release, I doubt that Sony ever considered that Scarlett could be weak. I'm confident that any potential changes Sony made have little to do with what MS is doing.Or maybe Microsoft forced Sony's hand to re-think and build a beefier machine because of Lockheart ? Maybe Sony found out roughly in 2018 what Lockheart's specs were gonna be and holy shitted themselves thinking that they would be more close originally to Scarlett but got caught offguard because of Lockheart?
"We need to delay this thing a year to get our stuff straiten out and at the same time eventually learned roughly Scarlett's specs and doubled down on improving Ps5 to be closer to Scarlett than Lockheart?
Man oh man! thats just a theory so I apologize if it's outrageous loll!! lack of sleep and a blunt will do that.
But imagine if Oberon is PS Now server related to run PS4 games at Native 4K (hence the clock speeds) and is not PS5 related at all. Would give us Native 4K PS4 games, as Cerny pointed out.When did Reiner ever say target specs and not dev kits? Both Xbox and Sony had at the verily least PC-part dev kits in the wild at that point in time ( even end of 2018)
1.) lack of ray tracing tests does not indicate lack of ray tracing ability.
2.) Yes Cerny said 8TF for native 4K with PS4 generation graphics. That is fairly irrelevant for this discussion as that efficiency would be at GCN level.
When did Reiner ever say target specs and not dev kits? Both Xbox and Sony had at the verily least PC-part dev kits in the wild at that point in time ( even end of 2018)
1.) lack of ray tracing tests does not indicate lack of ray tracing ability.
2.) Yes Cerny said 8TF for native 4K with PS4 generation graphics. That is fairly irrelevant for this discussion as that efficiency would be at GCN level.
We will earn back your trust" but we will still have celebrities from NBA, movies at the show to make it better!!
I was talking about some 2nd imaginary PS5 APU other then Oberon.It doesn't exist,it's a fantasy.
PS5 APU is 36/40 CU design, we just don't know the final clock for retail console.It could be 2GHz as in a Prospero devkit but i expect a bit less then that.
1. ) but the lack of a higher CU count and clock speed indicates the architecture of the chip?
2. ) Ok, that sounds right.
well, i will wait for the official reveal of the specs.
The thing is, they're all meant to be hidden and secret. Now, to draw conclusively from what has leaked, you have to trust entirely the mechanisms that lead to their leaks, that they always occur the same way accurately.
Does every piece of semi-custom silicon from AMD leak onto these DBs? I don't believe the MacBook Pro Navi GPU was detailed anywhere ahead of the product being unveiled. In these GitHub leaks, do we know what the intern's objectives were? What the point of the documents were? The necessity of its accuracy? Do we know that when file was created in 2018 and last edited in July 2019 that it was updated in 2019, or just opened, and saved as in 2019?
If you know the answers to these questions, that there are mechanisms you trust enough, that will always lead to the same documents in the same places leading to all products, then by all means, draw your conclusions. Otherwise, I will keep an open mind about knowns and unknowns.
Wonder what? What could have been or something else?
Earn back your trust is a reference to the fact that attendees and devs and Others private information was embarrassingly made public accidentally outing many peoples private information.That tweet is laughable. "
Like do they not understand the cost for publishers? ANd understand the cringe that has been E3 the past couple years? There's a reason Nintendo got out, and there's a reason Microsoft uses their own theater. Same with what Sony use to do. Rarely past couple years they have suits from EA or NBA/FIFA celebs on stage.
I think Microsoft's last year they did this was 2 years ago. They even saw how dumb it was.
I am not knowledgeable about these things, just drawing upon what more knowledgeable people than me are saying and contextualizing with my own logical deductions.
Looking at Sony's first party software pipeline, I doubt the decision to move PS5's release to 2020 was done in 2018. Also after the One X release, I doubt that Sony ever considered that Scarlett could be weak. I'm confident that any potential changes Sony made have little to do with what MS is doing.
It's not a question of "hidden", just having an open mind. To rule it out in absolute terms you will need to have a solid proof that every single unannounced chip AMD is and has been working on recently has already leaked to public domain, and to be honest only AMD can tell you that (if they even care to track the leaks tbh).Exactly.
Chances for 2nd "hidden APU" are not exactly 0 but very close to that number :)
iirc he clarified in later tweets
Yup, I think that's the biggest question. And if we assume there are more CUs we don't know about, than the biggest WTF is how the hell Sony is going to keep an over 40 CU GPU running at 2Ghz cool?
I don't understand number 1. For certain the chip being tested was larger than 36 CU, most likely 40 with 2 WGP disabled to match PS4 Pro CU.
I believe it was discovered a week or 2 before launch, yes.
Wonder what? What could have been or something else?
Earn back your trust is a reference to the fact that attendees and devs and Others private information was embarrassingly made public accidentally outing many peoples private information.
I believe Albert already answered the first part and said it didn't make sense. If there are two distinct APUs, it would be a replan, or a frankenAPU and the actual custom APU. I don't see the point of developing two unless you plan on two (such as Series X and S).Wonderful post Albert. I have a question, is the development of two different APUs in parallel something that makes financial sense? Does it happen? For instance, developing an APU targeted for a 399$ machine and an APU for a 499$ and keeping both until a final strategy was formed? And if it does make sense, what is the timeline of such practice, in which point in time one will have to be discarded so developers can focus on the chosen APU?
PC GPUs get taped-out a long time before release, for instance the Pascal 1000 series was tapped-out 11 months before the first Pascal GPU was released. Consoles need even more time than that, they go into mass production in ~4 months if they intend to launch in November. Albert Penello had already said that both consoles won't get any changes done to their silicon a year from launch. It's probable that during 2019 all that was done on both APUs was tests, bug fixes, tweaks, etc. If at some point MS or Sony had upgraded their APU massively (and going from 36CU to 50+ is a massive change), it had probably happened before 2019.
The secrete APU theory, I guess anything is possible. But I would say two things regarding the secret APU theory (which seems like you don't really think is very probable either):
1) If Sony has a secrete APU that isn't Oberon, why do they keep spending millions on Oberon? It seems like we are on stepping E, if not higher by now, so why keep spending million on iteration after an iteration that they already knew in late 2018 that it will never see the light of day?
2) How come we haven't heard or seen a single leak regarding this secret silicon? We keep hearing about every one of the APUs again and again in different DB leaks, but this one has somehow avoided all of that.
So yeah, it's completely possible, but I think we both agree that chances are pretty slim.
The whole low-speed comment was in the context of the SSD, so maybe the low-speed comment was regarding the SSD which was some kind of stand-in at that time frame.
Even though the TF numbers are a really nice hit, I will say that the memory setup doesn't match the leaks, both Flute and Github suggest 256-bit interface with 16GB of GDDR6 while his post suggests 384-bit interface with a split 12GB GDDR6 and 4GB of DDR4. I'll also add that the 384-bit GDDR controller + DDR4 controller will make the die totally huge.
IMO the whole thing will stay within the CU. If I had to guess, the TMU will increase in size and the L0 cache will be added too. More L0 might make it too big and reach for their blood so it might not happen, but increasing the L0 will have great benefits to RT and will also contribute to performance outside of the RT realm. It's a win-win but it will make the die larger.
I don't understand number 1. For certain the chip being tested was larger than 36 CU, most likely 40 with 2 WGP disabled to match PS4 Pro CU.
....
Wouldn't that lead you to believe that AMD CAN keep Semi-Custom product details secret up until 2 weeks before a product lunch?
Yep, it makes you wonder which one of his sources thought it was using Jaguar when it was using Zen2+Navi, and how credible they are.
For me it is something that perspectivises the idea that a console at some time was replanned from an earlier release date.
Yep, it makes you wonder which one of his sources thought it was using Jaguar when it was using Zen2+Navi, and how credible they are.
Are you suggesting that a next gen console would have a last gen processor onboard to aid in BC?
That is my take also on the leaks.I am not knowledgeable about these things, just drawing upon what more knowledgeable people than me are saying and contextualizing with my own logical deductions.
I don't know if you HAVE to fit Oberon into some story about the PS5. There is a possibility that the PS5 is revealed and what is ends up being has no basis in the Oberon leaks.
If it HAD to guess and include the Oberon leaks, just because 36CUs @ 2Ghz doesn't make sense from a design perspective, is that the PS5 is likely in the 50CU range, same as the Xbox and runs around 1.7Ghz, same as the Xbox. The 36CU leak could be a PS4pro boost mode test, where they use the 36CUs to emulate PS4pro and run them at higher clocks to push performance because the PSU has the capacity and the cooling solution can cool it. It was an idea they had in the summer and they had AMD test it.
For me it is something that perspectivises the idea that a console at some time was replanned from an earlier release date.
<_<
Are you suggesting that a next gen console would have a last gen processor onboard to aid in BC?
Jaguar cores are probably <20mm^2 on 7nm. Just include them on the APU for BC.PS4 Pro 2 ;) probably one design when they were thinking the gen will continue much longer.
If the PS5 uses Jaguar cores...No, its pretty clear he is suggesting that this rumored revision of the PS5 seems to coincide with previous rumors that a system was going to be using Jaguar cores again.
IMO, whatever Sony and MS put in the cloud will be a modified home consumer console fitted to a server blade. I don't think Sony or MS want another SKU on their hands. There isn't any reason to build a custom APU just for the cloud unless you want it to have some other cloud-based purpose other than gaming. Home console design is already built to run cool in a tiny box, it will perform wonderfully on a server blade. IMO PS cloud will just have loads of PS5 blades and Xcloud will just have loads of XSS (for 4K streaming) and XSX blades (for 1080p streaming).Thats what I mean, I agree. Thats all we can deduce. I dont get how some people here jump on judging the type of GPU and CPU it may have had, maybe to push the narrative of the Github leak being accurate?
Like I said... we know nothing about PS5s specs outside of what Klee, Jason and Reiner has said.
Im starting to believe Oberon is for PS Now Servers. They cant keep crappy Jaguars running on there forever. Replacing it with stronger 8TFlop hardware and running PS4 games in emulation would even allow the games on PS Now to run PS4 games in Native 4K or better. And we know Mark Cerny already said 8TFlops is good for native 4K for PS4.
So imagine if Oberon is that. NOT PS5 hardware at all.
BUT
PS Now server hardware to run PS4 games in Native 4K.
Allowing them to make adjustments with advancements of Navi and Zen. FURTHER compounding why there is NO Hardware Raytracing in Oberon.
DrKeo , Jeffram What do you think?
Yup, building an APU "just in case" doesn't make much sense.I believe Albert already answered the first part and said it didn't make sense. If there are two distinct APUs, it would be a replan, or a frankenAPU and the actual custom APU. I don't see the point of developing two unless you plan on two (such as Series X and S).
If both platform holders are just using AMD's RT solution, IMO, according to their RT patent, the difference in performance will be based on clock speed and CU count. But there can be special customizations, like a data structure building silicon (like the powerVR patent) or adding more L0 cache (which is again, inside the CU and will scale with them). But if I had to guess, it will probably scale 1:1 with the TF count and clock speed based.DrKeo Thank you kindly for the response.
So, if I am understanding this correctly, under your speculation, the capability of RT HW will be directly tied to the no. of CUs within the GPU. Consequently, by the sheer virtue of having more more active CUs, XsX's RT capability will be comparatively better that PS5's (unless, in the hypothetical event where PS5's GPU is clocked to a frequency which results in it being as performant as XsX's GPU, despite having a lower active CU count).
We need a identifier you can click on for posts that are meant serious or not!Jaguar cores are probably <20mm^2 on 7nm. Just include them on the APU for BC.
(no, I am not serious)
Do you have a link to the tweet? Having trouble finding it pages back, could be missing it tho
Yea I have a feeling it's exactly like that . Some gddr6 and some ddr4 for os
I guess what I should've asked is, why 12? Why wouldn't they just put 16GB of GDDR6 on the devkits, and if it's a bus width issue, why not 24? And it's too much DDR4 for the amount of GDDR6, unless they're going for a PC-style split RAM design, which seems unlikely.Yes, I have been pushing this theory for a while, and there is precedent for it just looking at the PS4(256MB DDR3) and PS4pro (1GB DDR3/4 I am not sure)
I am expecting 16GB of GDDR6 and 3/4GB of LPDDR4.
Can someone explain to me how cu counts work and why would you need to deactivate that many to emulate ps4? Can't you just use all of them actively to run the ps4 games and that's it? How does it work? Would that make the ps4 games run too fast or render the emulation unstable?
Can't they just program limits to the emulation with specific settings without having to disable cu's?
I honestly don't know how the tech works but I'm a good learner.. Would someone please care to do a quick explanation?
Thnx
Agreed.I am not knowledgeable about these things, just drawing upon what more knowledgeable people than me are saying and contextualizing with my own logical deductions.
I don't know if you HAVE to fit Oberon into some story about the PS5. There is a possibility that the PS5 is revealed and what is ends up being has no basis in the Oberon leaks.
If it HAD to guess and include the Oberon leaks, just because 36CUs @ 2Ghz doesn't make sense from a design perspective, is that the PS5 is likely in the 50CU range, same as the Xbox and runs around 1.7Ghz, same as the Xbox. The 36CU leak could be a PS4pro boost mode test, where they use the 36CUs to emulate PS4pro and run them at higher clocks to push performance because the PSU has the capacity and the cooling solution can cool it. It was an idea they had in the summer and they had AMD test it.
I saw your post at the other place mentioning PSVR2 and the mindblown.gif. Is there anything you can tell us about it at all? Like what can we expect in terms of improvements over what we have now or possibilities? PSVR2 is my most anticipated thing next gen by far.
What I got from Albert Penello's yesterday explanation was that both Microsoft and Sony probably had the same available products from ATI to choose and tweak from. Bearing that in mind, it is not about downplaying the GitHub leak but trying to find a logic behind it.That is how I roll too. I also thought the downplaying would end after what Albert Penello posted just yesterday . But instead it went into the opposite direction . I really hope for a reveal soon to be able to move on.
me: oh a notification from the playstation account, probably something boring
me: oh shit its happening.
then keeps on reading
I guess what I should've asked is, why 12? Why wouldn't they just put 16GB of GDDR6 on the devkits, and if it's a bus width issue, why not 24? And it's too much DDR4 for the amount of GDDR6, unless they're going for a PC-style split RAM design, which seems unlikely.
The DDR on the PS4 systems is really for background garbage, the entirety of games and virtually all of the OS still sits in the GDDR5. From what I understand, the DDR is used more like a high-speed storage than something to execute code from, and the presence of a good SSD should reduce the need for it. If they're going by the same setup, the regular system RAM would be 12 GB according to that tweet.
Do we think they're going for another RAM doubling surprise at the last minute? I doubt anyone, especially MS, would fall for it (or even care, really). All it would do is annoy developers at this point.
its probably the emoji.Why does it say clear your calendar for March 21 and the game is not March 21 ?
It makes zero sense though. Jaguar was considered weak when this gem of consoles released and including one even as provisional wouldn't even make sense for a 2019 releaseFor me it is something that perspectivises the idea that a console at some time was replanned from an earlier release date.
<_<