The best implementations of PSSR are far superior to DLSS 1. The meh implementations are a bit better. The bad implementations I would agree are the worst upscaling.
Sony PlayStation 5 Pro
- Thread starter chris1515
- Start date
Or a tech enthusiast who's followed gaming technology for 20 years who asks questions and performs unbiased comparisons.
My point can be discussed without calling me an internet troll. I've presented my point. Argue with it on merit.
Yes. So then quantifying and qualifying faults across upscalers would lead to being able to rank them somewhat best to worst, and inform choices on which upscalers to use. My posit is: so far in aggregate, PSSR might be the worst. However, that's with little proper research and a better, rounded investigation might prove otherwise. The argument 'one game handles this upscaler well' would not counter that that upscaler isn't the poorest option available at the time for a game.
Precisely.
Where did anyone say, or even suggest, anyone is wholeheartedly giving up on the tech? The argument is that, at this point in time, PSSR is not proving its worth and PS5 Pro 'enhanced' games are now offering the option to disable it because it's not good enough. True or false?
And so I'm asking for explanations and suggesting a tech interview from a trusted source, DF, to go ask the guy responsible who also showcased PSSR in a very different light pre-launch and find out what's happening and where the future lies. What's wrong with that?
When DLLS first launched, it was similarly critiqued and questions asked about how scalable it was. That's what thoughtful minds do.
Okay, refining my thought process a little, what were Sony's expectations for PSSR at launch, and how do the current results compare? Is this exactly what they were thinking when they advertised PSSR, knowing there'd be plenty of poo implementations?
And, as above, on aggregate how does PSSR fare versus other upscalers?
PSSR is still decent or even good most of the time. There’s a small subset of games in which it struggles mightily and worsens their IQ.
Frontiers of Pandora
Alan Wake 2
Silent Hill 2
Jedi Survivor
Those 4 particularly stand out and it seems that it’s in large part due to it not being able to denoise. According to a COD dev, that’s one of its flaws, but he said he managed to fix it in half a day. Now, whether it means he found a solution to denoise alongside PSSR or switch it off entirely remains to be seen. Still, this causes PSSR to have significant issues with RTGI as the shimmering artifacts look to be a result of its inability to denoise the RT.
It’s far superior to DLSS 1.0 and usually better than FSR as well. I would say it's right behind DLSS and XeSS, but it can easily compete with them and perhaps even surpass them in some respects when implemented properly.
Last edited:
The narrative is perhaps much worse than the situation for PSSR. They seem to be taking it on the nose here much more than what nvidia went through.
Mainly, this technology is old. And people were willing to put up with issues because we didn’t know where this was headed, and what it could be.
Secondly, people are used to “challenges” when you’re at the frontier of gaming. You pay lots of money to get access to features that many developers may not be willing to support, it just comes with the territory on PC.
Thirdly, console people are not used to being beta testers. Consoles have traditionally just worked and people are used to buying and seeing upgrades right across the board. They didn’t have to make choices, the choice was obvious and made for you. And so PSSR being launched half baked is a problem here, people paid significantly more for this console for this feature, only to turn it off is a bit of a disaster.
But the light is, come January I expect most of these problems do be solved? If the timing is to be believed FSR4 should launch with RDNA4 at CES. I am assuming it is the same technology, just fully baked and not rushed out for Christmas. PSSR should be fixed by then and we should see better implementations as developers look to integrate FSR4.
I'm certainly not suggesting the thing will never be solved, nor that PSSR won't turn out to be the bees' knees*. I'm just really confused to its current state and positioning. Repeating my point above:
"What were Sony's expectations for PSSR at launch, and how do the current results compare?"
* Is it bee's knees, the knees of one bee, or bees' knees, the knees of many bees?!
"What were Sony's expectations for PSSR at launch, and how do the current results compare?"
* Is it bee's knees, the knees of one bee, or bees' knees, the knees of many bees?!
I imagine Sony knew exactly how things would turn out. They would still sell to the niche market that deems this refresh worthwhile.
My personal view is that I would still take this over FSR and DLSS 1. At least with PSSR you get some games where the results are far superior. FSR and DLSS 1 always look bad. I don't expect significant PSSR improvements until PS6. The wildly variable results will persist through this console's lifespan.
No tech has the luxury of time travel to be compared to how something was. DLSS1 is now two gpu generations old.
PSSR is no exception. The best thing going for PSSR atm is that DLSS doesn't exist on the consoles.
What would help is Sony having some sort of a timeline and communication around PSSR's roadmap.
There is dedicated silicon, but not in the idea of Matrix Cores or Tensor Cores. I’m about 100% positive on that, 5Pro should mainly be RDNA4 CUs. And PSSR is largely FSR ML based upscaling due out for FSR4.
Though FSR is meant to run on Strix processors and talk of handhelds. It’s likely not as heavy as people believe it to be.
That's talking about PS5 Pro appeal, not PSSR results. Did Sony know results would be this mixed, or did they think their technique and hardware would fair much better and are surprised by some of the failings?
I meant they knew results would be mixed and determined it wouldn't affect sales.
Metal Snake Eater
Newcomer
Digital Foundry Talked a bit of God of war Ragnarok in its unlocked performance on the PRO but "open surprise" talked about the locked performance a bit and according to him the PRO with PSSR drops in performance into the 50s ..
if his Data is accurate i assume the unlocked performance "may" have a wider Dynamic resolution window ?
source:
if his Data is accurate i assume the unlocked performance "may" have a wider Dynamic resolution window ?
source:
GhostofWar
Regular
Is it really worth comparing current AI reconstruction to dlss 1? wasn't AI used for the whole thing with no temporal component? wasn't that why it also had to be trained per game? The other techniques following already new to skip straight to temporal + AI especially if training the models was maybe too much even for nvidia? Seems like having the ability for apples to apples comparison but using apple to orange just because the first mover saved everyone else going down the orange path.
Strix Point has a NPU. PS5 Pro may also have a NPU?
nah, the NPU introduces other issues.
NPU based upscaling is outside of the rendering path, so after the frame is completed, it's sent to the NPU to be upscaled. This form of upscaling introduces a 1 frame latency.
PSSR operates within the rendering path. So there is no latency.
That's a big issue. We've no insight what's failing, which is frustrating. Is it an integration problem or an algorithm problem?
It seems to require individual games to be updated if the talks about certain games using earlier SDKs (from August/September instead of October/November) causing the PSSR issues is to be believed and how patches to those individual games showed improvements.
Similar threads
- Locked
