It's kind of funny how it's always just this small part of the "Epsilon" document that's being posted. Probably because the full document doesn't look so good:
"gaming plataform"
"High Dinamic Range"
"Stably rotating games at 60 frames per second"
Idk, looks a bit fishy.
If these are the actual PS5 specs I'll eat my hat.It's kind of funny how it's always just this small part of the "Epsilon" document that's being posted. Probably because the full document doesn't look so good:
"gaming plataform"
"High Dinamic Range"
"Stably rotating games at 60 frames per second"
Idk, looks a bit fishy.
My first thought is the misspellings are for leak detection..
Exactly
60CU @ 1.85Ghz gives you 14.2 Tflops. ;pFunny thing about the 8TF rumor, the exact thing that created the rumor was then conveniently ignored when it changed to suggest a different power margin.
Anyways I still think its no going to be 14TF. As there is only one GPU from the GPU leak so far that will allow that and there is too much about it that means it can't go into a console. I think it could be 12TF though. If they use a Navi 10 CE clocked at around 1680Mhz,which by all indications will be them downclocking it. I can even see MS going with a 60CU variant of the card, basically the 64CU variant and disabling 4CU and landing at around 13TF.
So the PS5 is going to have the full Navi 20 GPU? This console is going to cost an arm and a leg if that's the case. Maybe that's why Microsoft are so dedicated to releasing a second cheaper console.
Even if its cut down from the full Navi 20 GPU slightly, just going off that AdortedTV source still shows it's going to be really expensive, for Sony anyway.
This is how companies figure out who is the leaker. They don't use identifiers because anyone could shop them out. No one is going to shop out common spelling errors. With spelling errors you can have essentially unlimited number of unique identifiers in your information without doing much work.
All the while I remain safe in my 12TF camp. Maybe at the end of the day we end up being the only right ones. Not too optimistic and not too pessimistic.Welp, we've fully returned to the 14TF part of the circle of life.
I'll be over here in the 10TF camp.
the only rumor that pointed to 8tflops was the first Gonzolo benchmark which was running at 1Ghz. It has since seen an upgrade to 1.8Ghz which could point to 14 tflops if they kept the CUs the same.
There is nothing else that points to 8 tflops. Even Jason Schrier said both consoles are aiming to be more powerful than 10.7 Tflops Google Stadia.
oh and Albert Panello thinks the PS5 might be 8 tflops, but he seems to be the only insider who believes this.
I can think of one platform for which this is not the case.I believe it was the case for at least the past generation, that the final hardware-equivalent devkits had double the RAM of the actual consoles for easier debugging.
You're right, Sony begged AMD for a clock speed increase after reading some video game forums. It couldn't possibly be because Gonzalo was still in heavy development and prototyping stage.Interesting.
Maybe it was originally 8TF but Sony changed it to 14TF because of the controversy that 8TF had circling it, I mean let's be real, ever since the original Gonzalo leak revealing the 8TF most fellow speculators all over the internet hated the idea of the next generation PlayStation console only having just double the TF of the Pro, hell even users on ERA have expressed their distaste of 8TF (like you and me) and with even google themselves boasting how their 10.7TF horsepower dwarfs the current gen consoles and even taunts the 8TF rumour, it's a marketing embarrassment, in fact when the PS5's TF are eventually revealed people will judge it harshly if it's 8TF after all.
Maybe Sony asked AMD to beaf up the Gonzalo's clock speeds to achieve a more desirable number that fans will be happy for, to prevent that marketing nightmare that could potentially ensue.
Or the first Gonzolo benchmark was for the 8 tflops Lockheart and not the PS5 like we all originally thought.Interesting.
Maybe it was originally 8TF but Sony changed it to 14TF because of the controversy that 8TF had circling it, I mean let's be real, ever since the original Gonzalo leak revealing the 8TF most fellow speculators all over the internet hated the idea of the next generation PlayStation console only having just double the TF of the Pro, hell even users on ERA have expressed their distaste of 8TF (like you and me) and with even google themselves boasting how their 10.7TF horsepower dwarfs the current gen consoles and even taunts the 8TF rumour, it's a marketing embarrassment, in fact when the PS5's TF are eventually revealed people will judge it harshly if it's 8TF after all.
Maybe Sony asked AMD to beaf up the Gonzalo's clock speeds to achieve a more desirable number that fans will be happy for, to prevent that marketing nightmare that could potentially ensue.
Hi all,
Something that has been playing in my mind regarding the various power targets, and thought I would get your perspective, as I am not as tech savvy as many here.
I understand the benefits the stronger CPU will bring, however from a developers point of you, what would be the better outcome, more Tflops, or more RAM?
E.g
10Tf console, with 32GB RAM with SSD
or
14Tf console with 16GB RAM with SSD
If I were to look at a city in an open world game, would it be better to have 32gb, have much higher res textures loaded and much more of the city and map loaded to eliminate LOD and streaming of assets, and the fake fog in the distance hiding low res textures.
With SSD's being as fast as Cerny claims in the next console, loading the highly detailed area could be quite fast to get the game started, and then in the background the rest of the game map, or as much as needs to be is loaded over the next 20-30 secs, or whatever that takes. Would this would get rid of load times entering/exiting buildings and many other streaming and pop-in issues we currently see?
This is probably not how game development works, but what would be the benefit for the extra teraflops versus the extra ram, outside of ray tracing and post processing?
Also means they are using a 64CU GPU which probably isn't much bigger in size than the 56CU GPU which will no doubt be a 60CU GPU with 4 CUs disabled. The issue with that though is the power draw, only way I see them getting it down to any acceptable levels by downclocking it from 1.8Ghz to around 1680Mhz. Which brings it down to around 13TF. The all round cheaper option will be going with the 56CU GPU and downclocking that to 1680Mhz and ending with 12TF.
It's kind of funny how it's always just this small part of the "Epsilon" document that's being posted. Probably because the full document doesn't look so good:
"gaming plataform"
"High Dinamic Range"
"Stably rotating games at 60 frames per second"
Idk, looks a bit fishy.
Also means they are using a 64CU GPU which probably isn't much bigger in size than the 56CU GPU which will no doubt be a 60CU GPU with 4 CUs disabled. The issue with that though is the power draw, only way I see them getting it down to any acceptable levels by downclocking it from 1.8Ghz to around 1680Mhz. Which brings it down to around 13TF. The all round cheaper option will be going with the 56CU GPU and downclocking that to 1680Mhz and ending with 12TF.60CU @ 1.85Ghz gives you 14.2 Tflops. ;p
Nope, not the full Navi 20 one. An RX 3080 is 15% faster that a Vega 64 which is a 12.6tflops GPU. That gives us a 14 tflops GPU.
If you dont want to treat the 15% extra performance as an increase to tflops, you can go to the next GPU on the list; The Navi 10 SE. With 8 additional CUs, you should get the 14+ tflops figure at 180W. The 6 Tflops RX 580 was also 180W and yet MS managed to put a 6 tflops APU in the X1X anyway.
The GPUs listed below are not APUs but Desktop GPUs. Their TDPs tend to be much higher than the APUs in console since they include the power usage by VRAM (for all we know these GPUs can have 12GB of GDDR6) and other parts on the card.
Welp, we've fully returned to the 14TF part of the circle of life.
I'll be over here in the 10TF camp.
Welp, we've fully returned to the 14TF part of the circle of life.
I'll be over here in the 10TF camp.
Or the first Gonzolo benchmark was for the 8 tflops Lockheart and not the PS5 like we all originally thought.
You're right, Sony begged AMD for a clock speed increase after reading some video game forums. It couldn't possibly be because Gonzalo was still in heavy development and prototyping stage.
Local titty flips are always better than remote.10TF is such a boring prediction though. It's bang in the middle of the Flat Earth <-> Jesus Planet spectrum. :P
In fact 10TF is less than Stadia. Sorry my dude but I'm not joining your team! :p
This was the quickest flop we did.few hours back we were at 8 and now we r back at 14 hahaWelp, we've fully returned to the 14TF part of the circle of life.
I'll be over here in the 10TF camp.
You forgot the CPU Zen 2 8C/16T @ 6GHz.
Ok so let's take a look at this. On the PS2 Sony took a $188 loss. On the PS3 they took a $400 loss. On the PS4 they took a $60 loss. This tells us historically when Sony is doing very well that they're willing to take large losses on their consoles. Even when Sony wasn't doing well they were still willing to take a loss on the PS4. Considering how well they're doing now I think just based on historical precident, how much more important PlayStation is to them now than ever, and the amount of ways they have to make back money on losses I think it's more than reasonable to expect them to take a large loss on the PS5. I'm now guessing up to a $200 loss per console.
So if we wind up getting a $399 PS5 then it might actually be up to a $599 console. If it's a $499 PS5 then it might be a $699 console. Based on the wide array of leaks we've gotten so far and based on the fact that we know they're going with an ultra fast SSD option and the high probability of hardware raytracing I am feeling more confident than ever we're looking at a heavily subsidized, and very powerful PS5.
Wtf is Xbox SNEK from the OP? Is that like the SNES and SNK being combined for some reason?
Let us pretend for a moment your interpretation of the performance column of the above table made by AdoredTV is right - just to make the case. You would need 2280 Mhz clock speed to reach those 14TF (or Vega 64 + 15%) with the 48CUs in the table. Is this really sounding realistic to you? I am talking just PC now not even console!
- The AdoredTv leaked Navi GPU performance chart from last week also gives us 14 tflops for RX 3080 which is something we can expect in a console APU. Vega 64 is a 12.5 tflops console. An Additional 15% would give us 14 tflops. The Special Edition Navi 10 gives us 14 tflops too if you dont want to include the +15% performance improvements directly to the Tflops number.
Joke post? I remember the last claim the PS5 could even launch in 2018 but 2019 latest. This type of well sourced information?SemiAccurate won't put this pdf behind a $1000 paywall if it wasn't well sourced. Why would they open themselves to litigation like that.
A Navi GPU with 10TF will be better than a Vega GPU with 10.7TF in game performance judging from the AdoredTV table. So yes, 10TF could be exactly that what Jason Schreier was told to expect. The only thing you need to do to change subject from a synthetical peak performance number to the expected performance profile in games.10TF is such a boring prediction though. It's bang in the middle of the Flat Earth <-> Jesus Planet spectrum. :P
In fact 10TF is less than Stadia. Sorry my dude but I'm not joining your team! :p
Now I have a question about this particular image.
Is it safe to rule out the first compute unit prediction due to the fact that the PS4 Pro was 36 CU's and that the PS5 will be an improvement to the PS4 Pro's CU?
I wouldn't rule that out. I'd say 36 or 44 cu are more likely than anything else.
I can share some PS6 specs. Mark my words.
CPU: ***
GPU: ***
Memory: ***