Sony PlayStation 5 Pro

btw one interesting factoid from Mark Cerny's presentation is that around 75% of PS5 gamers are choosing performance mode. I always appreciate options and I don't want to see 30/40fps Fidelity modes go away mind you, but over the years I've seen numerous arguments that your typical gamer doesn't care about performance, they're fine with 30fps - it's only us technical dweebs who think this matters.

Well, it matters quite a bit to the general public it seems. They may not be joining in the #shaderstruggle anytime soon, but it seems even your more casual gamer can at least recognize the benefit of a high frame rate.
It depends if they are counting people not choosing any graphical mode at all. A lot of games start with a performance mode. I can see the people actually going in to the settings choosing the performance mode.
 
Last edited:
It depends if they are counting people not choosing any graphical mode at all. A lot of games start with a performance mode. I can see the people actually going in to the settings choosing the performance mode.

Depends on what you've set the default as on the PS5 (or if it's been set at all). Also we don't know what 'chooses' really means in Cerny's presentation - do they have data on actual hours vs. mode chosen to see if Fidelity mode was chosen and rejected at any point? Dunno.

RIP 40fps mode in games i guess, shame.

Not sure how that computes from this data. 25% is a lot of people who may want that option.
 
That doesn't really matter if the so called "fidelity mode @ 60hz w/ PSSR" is actually running at a internal lower resolution on the upgraded SKU in comparison to the "fidelity mode" on the original HW ...
You are approaching this from the wrong angle, the angle of performance alone. You should consider both the angle of performance and the angle of image quality, because that's what PS5 Pro is offering.

PSSR needs to both look better and deliver higher fps than FSR.

If PSSR is a reuglar upscaler, then it won't look better and deliver faster fps at the same time, it will do one or the other not both.

Starting from a lower res than FSR, then the best it could do in that case is to look better at the same performance or look the same but deliver higher performance.

The only possibility for PSSR to look better than FSR at a lower resolution and deliver better fps is if PSSR is hardware accelerated.
 
That doesn't really matter if the so called "fidelity mode @ 60hz w/ PSSR" is actually running at a internal lower resolution on the upgraded SKU in comparison to the "fidelity mode" on the original HW ...

There's more than enough headroom (lower rendering res/more compute resources w/ maybe even VOPD too ?) to reasonably implement a software upscaler at that point ...

Well I guess the raison d'etre for game consoles is effectively coming to an end if developers won't embrace custom programming with custom hardware ...
The specs on the developer portal have been 100% accurate so far, including the 45% performance figure. So I see no reason to disbelieve the 300 TOPS figure, which is far in excess of current AMD hardware.

Presumably your claim would be that Sony invested in a so-called A.I accelerator, only not to use it for anything?
 
I wonder if Sony will borrow a page from DirectSR and allow the PSSR API to be used on both base PS5 and the Pro, with FSR serving as a fallback on base models.
 
You are approaching this from the wrong angle, the angle of performance alone. You should consider both the angle of performance and the angle of image quality, because that's what PS5 Pro is offering.
Am I when the real world falls very short of the rosiest interpretation of these rumors that you so devoutly believed ?
PSSR needs to both look better and deliver higher fps than FSR.

If PSSR is a reuglar upscaler, then it won't look better and deliver faster fps at the same time, it will do one or the other not both.
That's not as hard as you think it is even with something like XeSS DP4a when you have a bigger GPU and use lower internal resolution ...
Starting from a lower res than FSR, then the best it could do in that case is to look better at the same performance or look the same but deliver higher performance.

The only possibility for PSSR to look better than FSR at a lower resolution and deliver better fps is if PSSR is hardware accelerated.
It's not hard to beat FSR at "fidelity mode" when we have R&C Rift Apart running at an internal resolution of a dynamic 1800p-4K range on original HW so there's plenty of pixels to spare to give the upgraded SKU more reaching distance ...
The specs on the developer Portal have been 100% accurate so far, including the 45% performance figure. So I see no reason to disbelieve the 300 TOPS figure, which is far in excess of current AMD hardware.

Presumably your claim would be that Sony invested in a so-called A.I accelerator, only not to use it for anything?
So then why is it not meeting the 2-3x RT performance target figure in real applications then ?
 
Doesn't really bode well for PS6 does it really?
According to Shifty Geezer, these are just how these companies have to price these products now. They have no choice, cuz they're struggling so badly!

PS6 will be like $1000 and if you complain about it, you're simply not understanding the importance of catering to shareholders above all else! It clearly has nothing to do with Playstation being greedy and trying to take advantage of thinking they own the console market completely now as Xbox slowly exits. No, it's just the necessity of the market, and as consumers, we need to understand and accept that making Playstation as much money as possible is what is most important!
 
Am I when the real world falls very short of the rosiest interpretation of these rumors that you so devoutly believed ?
No one believed a 2x overall improvements, it's widely understandable that this refers to isolated ray tracing tests. The overall performance is expected to be in the range of 50%.

That's not as hard as you think it is even with something like XeSS DP4a when you have a bigger GPU and use lower internal resolution ...
DP4a XeSS is slower than FSR at the same internal resolution. If you lower the resolution of XeSS it would run faster and potentially become as fast as FSR but will look worse or the same as FSR.
 
Welcome to capitalism, where a healthy dose of inflation is inevitable.
You want deflation? Go ask Japan how that went.
A ridiculous strawman. I'm pro-capitalism, and understand the importance of temperate inflation over time.

If you honestly cant understand why the situation we are in now is different, then you're not remotely equipped to have this conversation. smh
 
According to Shifty Geezer, these are just how these companies have to price these products now. They have no choice, cuz they're struggling so badly!
There may be truth here in his statements. Prior to PS4 times, consoles were sold at a loss because they could make it up in game purchases and subscriptions.

That is no longer true today. The most played games are F2P. And they don’t require a subscription. You just need a console and a monitor and you’re in. If you sell your console at a loss, you just eat losses and that’s not a great place to be. Sony already has some of the lowest margins out there despite being the leader by a country mile.

They simply may not be able to afford any sort of loss on their console.
 
  • Like
Reactions: snc
No one believed a 2x overall improvements, it's widely understandable that this refers to isolated ray tracing tests. The overall performance is expected to be in the range of 50%.
So basically the design wasn't as much of a factor of improvement compared to featuring bigger hardware itself then ?
DP4a XeSS is slower than FSR at the same internal resolution. If you lower the resolution of XeSS it would run faster and potentially become as fast as FSR but will look worse or the same as FSR.
@Bold On smaller GPUs, I don't disagree that this is the case ...

Recall that the goal of the PS5 Pro is to "eliminate the gap" between fidelity and performance mode so they probably are indeed using PSSR in these comparisons as well to extract the benefit (higher refresh rate) of a lower rendering resolution and then offset it with a better upscaler ...
Do we have benchmarks from the Pro already?
Not really benchmarks but 'figures' from Cerny himself ...
 
No one believed a 2x overall improvements, it's widely understandable that this refers to isolated ray tracing tests. The overall performance is expected to be in the range of 50%.


DP4a XeSS is slower than FSR at the same internal resolution. If you lower the resolution of XeSS it would run faster and potentially become as fast as FSR but will look worse or the same as FSR.
Neither of you can know what is in there. Either or can be true.

You guys need to stop writing like you have the answer. Without confirmation that tensor silicon exists, you’re just as correct as Lurkmass.

300TOPS is a nice leaked figure, but without knowing its precision I can’t say anything. They were also wrong on the price point.
I don’t trust leakers to know the various ways that tensor flops can be calculated.

This back and forth needs to settle down until more information is required.

Personally I agree that there is a high chance it is dp4a. They sold PSSR as a library and snuck in that custom hardware for ML. Well they also said they made the primitive engine.

So take it how that is. IMO, if they had tensor cores, it would have been announced here. It’s such a massive difference over regular silicon, of course you would market it.

People flip out on console wars over stupid things, and they won’t take 30 seconds to go over bespoke tensor cores designed to chew through NNs?

secondly, we know how price sensitive consoles are Do they have the silicon space to run tensors and the cache with them? What if they are never used ? Just seems very expensive for a feature set that isn’t with the base system.
 
300TOPS is a nice leaked figure, but without knowing its precision I can’t say anything. They were also wrong on the price point.
I don’t trust leakers to know the various ways that tensor flops can be calculated.
They dont have to know how to calculate, 300 top was in sony document

PS5-pro-pssr-and-gpu-leak-mlid.jpg
 
1725994379990.png

If this was true surely Sony would have said so but CNET seems oddly certain. Would developers even need to know about this if its only in final software for unpatched games?
 
Last edited:
Back
Top