PS5 Pro *spawn

[snc]

Uh, pssr not that good as I thought. Some briliant examples like Stellar blade and FF r but also very poor. Artifacts and stability problem in Dragons Dogma 2, Alan Wake 2, Veilguard to the point its not that much better than fsr 2 in this games.

 

[Charlietus]

Uh, pssr not that good as I thought. Some briliant examples like Stellar blade and FF r but also very poor. Artifacts and stability problem in Dragons Dogma 2, Alan Wake 2, Veilguard to the point its not that much better than fsr 2 in this games.

Those games look broken (DD2 and Veil guard). Maybe some things aren't working well in those engines with the upscale. Alan wake 2 has the problem with post processing that maybe is making it alias more than it should, and still images look strange. I also wouldn't expect a 864p base to produce a good 4k image with PSSR, at least right now.

 

[Charlietus]

Just finished the video, and with all the information that we have now, I can say that it's like if PS5 had been released in 2022, overpriced by at least 150$.

Not great.

 

[Globalisateur]

How does game boost work for 30% or even 40% more performance?

Do PS5 codes really use 36CUs only?

That's puzzling indeed.

 

[pjbliverpool]

Sure, but the complete 16:9 resolution is 2160p, and that is upscaled from 1080p with DLSS performance.

The 16:9 equivalent resolution would be about 2844x1600. I get extra pixels at the side vs the 16:9 render, nothing is cut off at the top and bottom. Essentially I sit as close as I can comfortably manage to the screen which is about 2-3ft. If the screen was 16:9 (same height but narrower) then I couldn't sit any closer. On the other hand if it was simply bigger overall (same width but taller) I would likely need to sit further away, thus giving a comparable image at 2160p as I currently get at 1600p anyway.

 

[Dictator]

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

As i Said elsewhere, this is more 3070 Ti Levels at best If that GPU Had more memory...

 

[Globalisateur]

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

As i Said elsewhere, this is more 3070 Ti Levels at best If that GPU Had more memory...

In elden Ring? Is the 6800 using the same native resolution ? Game on PS5 has DRS and runs at a very high res on PS5 Pro. Besides that game is not patched and who knows how many ressources it use in that mode.

Not sure using unpatched game is the right way to jauge PS5 Pro vs GPUs. What do you think?

 

[pjbliverpool]

I'm a little confused at the comparison between the PS4 image quality enhancement mode and Nvidia DSR at around 15min. I really don't see how they are comparable at all. DSR renders at a higher internal resolution and then scales the image back down to the display output resolution. It's true super sampling and will make the image look FAR better than the base display resolution at higher factors. The image will be sharper with more detail and much less aliasing both in stills and in motion. That's nothing at all like what was shown for the PS4 image quality enhancement which was described as simple post process filter that doesn't improve aliasing and offers no improvement at all on several graphical elements.

 

[Below2D]

In elden Ring? Is the 6800 using the same native resolution ? Game on PS5 has DRS and runs at a very high res on PS5 Pro. Besides that game is not patched and who knows how many ressources it use in that mode.

Not sure using unpatched game is the right way to jauge PS5 Pro vs GPUs. What do you think?

Rich tested it using PS5's Quality Mode settings at 4K on the RX 6800.

 

[Remij]

PSSR's "wow factor" was always going to fall on the games which were terrible to begin with.. ala Final Fantasy 7: Rebirth. PRO will make a lot of games more consistent it quality, and performance.. so for the group that wants the best PS console experience possible, it will serve its purpose.

 

[Cynicalking]

The 16:9 equivalent resolution would be about 2844x1600. I get extra pixels at the side vs the 16:9 render, nothing is cut off at the top and bottom. Essentially I sit as close as I can comfortably manage to the screen which is about 2-3ft. If the screen was 16:9 (same height but narrower) then I couldn't sit any closer. On the other hand if it was simply bigger overall (same width but taller) I would likely need to sit further away, thus giving a comparable image at 2160p as I currently get at 1600p anyway.

That's exactly the point. Using that ultrawide res you're upscaling to a lower res , not 4k but 2844x1600, so of course the input resolution can be lower than 1080p. If you're upscaling to 1440p you can manage lower base res with a competent upscaler like DLSS.

 

[Cynicalking]

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

As i Said elsewhere, this is more 3070 Ti Levels at best If that GPU Had more memory...

Any reason why Elden Ring's RT mode was not tested?

 

[Saidur_Ali]

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

PS5 Pro has faster memory and is more efficient. It is running higher clock speeds for boost mode (Probably 3.85 GHz for CPU and 2.35GHz GPU) so they might cap memory bandwidth and CU utilisation for compatibility hence why games need to be patched to get higher performance. Seemed an odd choice by DF not to compare performance in native PS5 Pro games in terms of performance gain and relative performance of GPU to PC in a review but what is essentially PS5 backwards compatibility mode. It will be a bit like comparing how PS5 is in terms of performance by judging it only on PS4 Pro boosted games in a review. I wonder if GT7 is patched with basic Pro support given it looks to have a much bigger performance gain more in line with hardware. They can't advertise it as PS5 Pro Enhanced due to not increasing target maximum resolution or framerate or feature any new raytracing effects yet.

 

[Boss]

I disagree. I commonly use DLSS performance on my 3840x1600 monitor from 2-3ft away and it generally looks excellent IMO. That's an 800p input resolution.

Eh, that's fair. I use a 3440x1440p 34 inch qd oled monitor and anything other than DLSS quality looks like trash to me. When I decide to connect to my tv which is 77 inches, DLSS performance at 4k looks just barely tolerable.

 

[Gamer09]

Uh, pssr not that good as I thought. Some briliant examples like Stellar blade and FF r but also very poor. Artifacts and stability problem in Dragons Dogma 2, Alan Wake 2, Veilguard to the point its not that much better than fsr 2 in this games.

I think the initial demo of Rift apart was just crafted very well to showcase PSSR, it is always too see the technology in practice for real world results.

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

As i Said elsewhere, this is more 3070 Ti Levels at best If that GPU Had more memory...

If it had more memory, BW. It is closer to a 3070 in raw raster performance then?

 

[oliemack]

I'm a little confused at the comparison between the PS4 image quality enhancement mode and Nvidia DSR at around 15min. I really don't see how they are comparable at all. DSR renders at a higher internal resolution and then scales the image back down to the display output resolution. It's true super sampling and will make the image look FAR better than the base display resolution at higher factors. The image will be sharper with more detail and much less aliasing both in stills and in motion. That's nothing at all like what was shown for the PS4 image quality enhancement which was described as simple post process filter that doesn't improve aliasing and offers no improvement at all on several graphical elements.

I did not mention DSR at all. I said RTX Super Resolution, which is a machine learning based filter applied to video content.

 

[Boss]

If it had more memory, BW. It is closer to a 3070 in raw raster performance then?

I'm still waiting for those claiming to know the raster performance of the ps5 pro to provide their evidence.... When they've tested other gpu's with the cpu in the pro, then they can start making claims about gpu performance. Testing games at a higher resolution and then claiming that it's now gpu bound without the means to verify it is disingenuous. When you look at the apu for the ps5 pro, it's bottleneck ridden. You can easily be cpu bound, bandwidth bound, etc.

As an aside, for the price sony are charging, they cheaped out big time. The cpu should be running at least 4ghz or 4.5ghz. It's on TSMCs N4p if rumors are to be believed. Even if they didn't want to change the cpu, they could have significantly upped the speeds. Also the bandwidth is clearly not enough. Regardless, I'll be picking up my 2 units tomorrow and putting them through their paces.

 

[RDGoodla]

I think PS5Pro is perhaps bandwidth bound heavily visavis its potential compute/shading based on the modest upgrade in unpatched titles (30-35% in the good GPU bounds cases). It is also performing worse than the RX 6800 which is not too great. RDNA2+ on Desktop is leveraging Infinity Cache heavily for 4K and more... It not being here is probably a limiting factor with many more CUs to feed at high res.

How does Sony allocate PS5 codes to 60CUs?

Previously consoles games only use fixed number of CUs. PS5 Pro seems to allocate PS5 games to use more CUs.

As i Said elsewhere, this is more 3070 Ti Levels at best If that GPU Had more memory...

If it had more memory, BW. It is closer to a 3070 in raw raster performance then?

Pro already has 1.2 GB more memory while 3070ti uses 8GB VRAM.

 

[techuse]

I'm still waiting for those claiming to know the raster performance of the ps5 pro to provide their evidence.... When they've tested other gpu's with the cpu in the pro, then they can start making claims about gpu performance. Testing games at a higher resolution and then claiming that it's now gpu bound without the means to verify is disingenuous. When you look at the apu for the ps5 pro, it's bottleneck ridden. You can easily be cpu bound, bandwidth bound, etc.

As an aside, for the price sony are charging, they cheaped out big time. The cpu should be running at least 4ghz or 4.5ghz. It's on TSMCs N4p if rumors are to be believed. Even if they didn't want to change the cpu, they could have significantly upped the speeds. Also the bandwidth is clearly not enough. Regardless, I'll be picking up my 2 units tomorrow and putting them through their paces.

Sony themselves state "up to 45% faster than PS5." That puts it firmly in 3070-6800 territory. Bandwidth contention issues and a lack of infinity cache will result in cases where performance falls below that. More efficient code and an API advantage mean sometimes It will hit above that.