Sony PlayStation 5 Pro

They dont have to know how to calculate, 300 top was in sony document

PS5-pro-pssr-and-gpu-leak-mlid.jpg
Ah thank you.

For reference that says 8bit 300 TOPS.
It’s highly likely this is a really large XDNA2 in there.

I can’t math out any other way. If it were 4bit 300TOPS. Maybe. I don’t recall how dual issue works, but 300TOPS at 8 bit. Seems like a stretch too far for CUs to process.

If you replace the XSX with a smaller bus and you can probably get away with enough XDNA2 in there to make 300TOPS. I’m a bit surprised they got away with both.
 
Last edited:
  • Like
Reactions: snc
Not really benchmarks but 'figures' from Cerny himself ...
I just re-listened to the presentation and there were no performance figures given for RT modes. However Cerny talks about the Pro doing ray calculations at double or triple the speeds. So that supports the 2x-3x RT figure in the developer portal being a synthetic measurement.
 
Just did this very quickly, I could shave some money off (Maybe £100-150 more, and that would cover the CPU cooler and OS and price would be same as it is now)

pc.png

  • With a 7700XT = £977.90
  • With a 7800XT = £1027.91
  • With a 4070 = £1067.90
  • With a 4070 Super = £1137.90

Considering the benefits of PC, best backwards compatibility, mods, more choice, cheaper games, free online........I would certainly be looking at a PC instead.
 
I'm genuinely baffled as to why they even bothered if they couldn't hit a better price point. I expect the impact of this on the market will be minimal at best. Given the expenses inherent in developing and supporting a new sku on the market, what ROI can justify it? This won't increase market share. I don't see how it prevents market share from eroding if that is a thing they are afraid of. Is this just a halo product? Does PlayStation need a halo product?

Genuinely baffled.
I'm baffled by the presentation of the reveal. They showed a whole bunch of cross gen titles and bragged about how much better they run. Last of Us 2 is over 4 years old at this point, and while I would assert that I think it still looks great and I can see the enhancements they offer on Pro, I think that if I were marketing a new device I would try to present that new device with software that was an example of a new experience you could get on the new hardware, and not a PS4 title.

They didn’t give ant details. This is fucking crazy.
They showed the only number that mattered to me, and that's the price. It's up to double the frame rate of fidelity mode at exactly double the price.
 
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
No one is denying that, they are using PSSR in all comparisons. I am just denying your logic which only leads to two possibilities: PSSR is software and thus cannot deliver good enough results to narrow the gap, or it's hardware and can deliver good results.

You guys need to stop writing like you have the answer. Without confirmation that tensor silicon exists, you’re just as correct as Lurkmass.
I am using real world data from XeSS DP4a applications and from Cerny's presentation to deconstruct a failing argument from months ago, I am not blindly speculating. We are not the same.

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.
Sony hid all the details of the RT enhancements as well, we already postulated they did this because they are hiding upcoming RDNA4 details.

Cerny mentioned the RT improvements are not available in any AMD GPU right now. They are improvements in upcoming AMD products, he proudly stated they motivated these improvements. Which is why it's logical they are not ready to talk about it just yet. The AI improvements will get the same treatment.
 
I just re-listened to the presentation and there were no performance figures given for RT modes. However Cerny talks about the Pro doing ray calculations at double or triple the speeds. So that supports the 2x-3x RT figure in the developer portal being a synthetic measurement.
Mark Cerny just said that he saw a broad achievement of fidelity visuals at performance mode levels of framerate across titles and in the the video presentation proceeded to demonstrate/mention this for Spiderman 2 (all modes have RT) and R&C Rift Apart (fidelity mode has RT) as well as an example of the new model near doubling the power of the original hardware ...

I presume that they're reaching the above goal with PSSR unless stated otherwise ...
 
No one is denying that, they are using PSSR in all comparisons. I am just denying your logic which only leads to two possibilities: PSSR is software and thus cannot deliver good enough results to narrow the gap, or it's hardware and can deliver good results.
Even if what you're stating is true that still leaves you and everyone else being wrong about the RT perf details ...
 
I just re-listened to the presentation and there were no performance figures given for RT modes. However Cerny talks about the Pro doing ray calculations at double or triple the speeds. So that supports the 2x-3x RT figure in the developer portal being a synthetic measurement.

Alex at DF has already suggested in a video that it's only 2-3x fast in a certain part of the RT pipeline and not the whole thing.

And I get the impression that the ray traversal might be 2-3x faster, but doing one part of the pipeline 2-3x faster doesn't increase frame rate by 2-3x.
 
I think we can safely say Sony's narrow GPU-spiel from the beginning of the Gen was a statement meant to muddle the conversation.
PS5pro is even more "unbalanced" and "wide" than Xbox Series X and yet... here we are.
Nothing sustains this. L0 and L1 caches have actually been doubled on PS5 Pro (RDNA3.5) from leaks. Main problem of XSX is lack of some GPU caches (ratio of cache size vs TFlops, notably the L1 cache) vs PS5. MS never disclosed the amount of L1 cache on XSX and usually they do this because they know it wouldn't be good PR. They did speak a lot about flops, fixed clocks and others stuff though.
 
Mark Cerny just said that he saw a broad achievement of fidelity visuals at performance mode levels of framerate across titles and in the the video presentation proceeded to demonstrate/mention this for Spiderman 2 (all modes have RT) and R&C Rift Apart (fidelity mode has RT) as well as an example of the new model near doubling the power of the original hardware ...

I presume that they're reaching the above goal with PSSR unless stated otherwise ...
He didn't break down what resolution the modes were running, so I don't see how we can evaluate the relative impact of PSSR vs. other improvements.

But this is a side point to the discussion of whether the 300 TOPS figure is accurate. Since the presentation backs up the 2x-3x figure in the developer portal, as a synthetic measurement, that is another reason to believe the information is accurate.
 
lmao you guys are arguing over MLID? That guy is a fuckin idiot. Our oc group has fed him all sorts of made up shit he posts as “insider info.” Even after being laughed at repeatedly he still does it! Because he needs to get “insider info” out there to feed an audience and generate revenue.

Please do yourselves a favor and just say no to techtube tabloids. Def don’t make them a hill to fight on.
 
lmao you guys are arguing over MLID? That guy is a fuckin idiot. Our oc group has fed him all sorts of made up shit he posts as “insider info.” Even after being laughed at repeatedly he still does it! Because he needs to get “insider info” out there to feed an audience and generate revenue.

Please do yourselves a favor and just say no to techtube tabloids. Def don’t make them a hill to fight on.
MLID was just reporting on the information in the Sony developer portal, independently verified by DF.
 
  • Like
Reactions: snc
Back
Top