PS5 Pro *spawn

[iroboto]

About bandwidth worth reminding that Sony said the bandwidth available would be comparably more efficient on PS5 Pro, somehow.



On top of better GPU cache architecture (doubled L0 and L1) that could help explain that claim I think memory clock (2.25ghz) been very close to GPU clocks (? 2.18-2.35ghz ?) could also help. On PS5 the difference of clocks (1.75ghz vs 2.23ghz) was not helping bandwidth efficiency according to Cerny.

I think it's good, but I haven't seen any miracles here compared to what dedicated hardware can do.

I think as we've learned more about PSSR, the real comparison point is XeSS DP4A vs PSSR. that is probably the closest we can get to comparing 2 to upscaling networks, running on something like a 6700.

 

[Globalisateur]

I think it's good, but I haven't seen any miracles here compared to what dedicated hardware can do.

I think as we've learned more about PSSR, the real comparison point is XeSS DP4A vs PSSR. that is probably the closest we can get to comparing 2 to upscaling networks, running on something like a 6700.

What are you talking about? The real comparisons will be with comparing the likes of Ratchet, Spider-man, TLOU1 on PS5 Pro vs 4070 with DLSS.

 

[iroboto]

What are you talking about? The real comparisons will be with comparing the likes of Ratchet, Spider-man, TLOU1 on PS5 Pro vs 4070 with DLSS.

You're not measuring the same thing when you do that though, unless you believe there is dedicated silicon outside of the CUs. I suspect if that were the case, the base resolution would be higher here for the game before upscaling because dedicated silicon should be able to run in parallel.

 

[snc]

You're not measuring the same thing when you do that though, unless you believe there is dedicated silicon outside of the CUs. I suspect if that were the case, the base resolution would be higher here for the game before upscaling because dedicated silicon should be able to run in parallel.

since when dlss is free

 

[iroboto]

since when dlss is free

its more free =P

From my POV, it would be setting up PSSR for failure if it is indeed run on compute units. You want to compare it against other algorithms that have similar hardware bottlenecks.