Nah. That was Kutaragi. They were already on their way to avoiding that ever again long before the PS4 launched.
If I understood his presentation correctly infinity cache wouldn’t help with their implementation of ML. But maybe I got that part wrong.
So what should it have been? At what price? And why?
RDNA3 has hardly any performance improvements over RDNA2 in raster, and RT's performance is still light years behind Nvidia's when looking at the performance of games using it on the Pro.
They are indeed
Probably very unlikely. The only way would be to detect when a game is using FSR and inject PSSR there in the same way some modders did on PC in games like Starfield, but even then I doubt it can be done globally, as the way each game/engine implements upscalers is different.
I guess having someone with first hand experience on game development designing a console has its advantages.
If this was the case, then the first few months post-release where PS5 was outperforming XSX, even if only slightly, should have disabused them of this notion.
Yes, he noted they haven't achieved their full goal yet by having a fully fused network and there are still potential issues with bandwidth bottlenecks they hope to address on future hardware.
Nope have never heard what the BOM was
Honestly the few times Ybarra brought it up as to why they were losing head to head they seemed totally baffled
I personally hope frame gen is not a standard, it adds noticeable input latency and really doesn't look good unless your native FPS is at least 60fps.
The documents taken from the developer portal all seem to have been proven accurate. The big innacuracy was the 33.5 figure, which I suspect MLiD made up based on the other specs. At least Cerny said this was never in the documentation and came from a misunderstanding from someone commenting on the leaked information.In the end, those leaked MLD docs were quite inaccurate in many aspects of the PS5Pro specs. I wonder if they were ever genuine, intentionally edited to be misleading or "supplemented" with assumptions based on certain specifics...
One more reminder that hardware leaks should always be taken with a grain of salt and never as true, no matter how real they may seem. There are always things that are impossible to know until the final hardware is ready.
Building ML models is a multi-year endeavour and Sony are basically where Nvidia were in 2020. I don't think it's case of rushing, since you need a lot of feedback from developers to improve the models, which you're only going to get from a full release. DLSS 3.0 took 2 years for Nvidia to release and XeSS 2 will have taken Intel 2.5 years as well, and they still don't have ray reconstruction working yet.
That hasn't exactly been measured yet.
Sony's approach to developer tools have always been more hands-off than not - ie. give as much freedom as possible. This in turn also means more ambiguity on 'what's right'(and can lead to negative results) - but it's one thing they've been consistent on for ... 25 years or so.
Marc did mention that to come close to those 33TF, developers would need dedicated PS5 PRO libraries and would have to optimize specifically for the new hardware, placing an additional burden on the already overworked developers.
Personally, I’m content as long as they deliver excellent work, providing Fidelity graphics at 60 fps and improved Ray-Tracing, that's all !
It’s not why Sony made the PS4 Pro per Sonys own words
Well, GamePass Mimir can always work as a human broom if Electrical Engineering 101 keeps eluding him.
It sold well its first month. Absolutely nothing wrong with putting this out, and if it's not for you, don't buy it. Easy solution!
It sold well its first month. Absolutely nothing wrong with putting this out, and if it's not for you, don't buy it. Easy solution!
Besides, Sony said nothing close to what you did about why the 4 Pro existed. The whole point was to retain their audience.
Yes, we do have numbers and this is why research is critical. The system doesn't have to outdo or even match the PS4 Pro to be considered a success (we've heard Sony put it in perspective and what their expectations are). As Topher put it, you're saying the higher end shouldn't exist, when the numbers prove that there's a market for these things.
Wait, same power as desktop 4080 the ps5 pro?So it's RDNA2 (for BC) + RDNA3 (vertex & geometry engine) + RDNA4 RT + fully custom ML hardware and GPU customizations with a power similar to a 4080, no 2x sparse matrix shenanigans as it's very custom and powerful. Sony designed and own the ML hardware (and software obviously). Warriors are going to spin this for years, decades even. So many Xbox fans hoping it would not be custom and arguing about the word custom used in the leaks.
It's a 16.7-18 Tflops machine the same way PS5 is a 10-10.3 Tflops machine
The Cerny patent about RT and new acceleration hardware (traversal logic + stack management) was 100% true.
Sony files new patent for ‘Accelerated Ray Tracing’ - speculation that it could be more performant
Source: https://www.psu.com/news/sony-files-new-patent-that-mentions-accelerated-ray-tracing-on-ps5/ Patnet: https://data.epo.org/publication-server/document?iDocId=6759906&iFormat=0 5.3 Publication of the European patent application 5.3.001 The European patent application is published as...www.neogaf.com
As the ML hardware specifically uses the increased L0 and L1 caches, there was no need to increase the L2 cache.
Similarly the new RT hardware units included in the shaders will alleviate main GDDR6 accesses vs PS5 games using hardware RT. It's probably where the claims about memory efficiency come from.
300 TOPs (8-bit) similar to a 4080 for Machine learning tasks like AI upscaling.
I mean, if you actually want a genuine discussion, you need numbers to have it. "Fantasy narratives" come from folks who do no research. Right here on the site you're on has several topics discussing November's Pro numbers, so seek those out so you can at least have some idea of how things are performing.And what are the numbers? So Sony now does charity, spends money on research, development and manufacturing and doesn't have to recoup those costs.
If you don’t know which games look extremely impressive on the Pro, then 1) you have too small of TV, 2) don’t play enough games to have a pre/post comparison 3) have too small of catalogue to compare 4) don’t have a Pro, 5) being a d*ck. In any event, the Pro is quite impressive, for reference Avatar, RE, TLOU, SM2, Outlaws, Callisto, FF, etc all look great compared to base. Absolutely worth the upgrade as everything that I want to play runs ~60 with great IQ now. I’ve also picked up a number of games I would never have before simply to see how they’re doing on the Pro. Sony executed well here, even with the hiccups from lackluster developers.This comparison doesn't make any sense. Nvidia has a wide range of products, from the most expensive to the cheapest, with good profit margins on top of them. In fact, the 4090 is precisely where Nvidia makes the biggest profit, although it sells less.
We are talking about consoles, which usually only have one version that is subsidized and sold for years without modifications.
And what are the numbers? So Sony now does charity, spends money on research, development and manufacturing and doesn't have to recoup those costs.
What games are these? Sorry, but it's your fantasy narrative, your quality criteria are probably too low to be impressed by anything they show you.
But that's my opinion, you don't need to be defensive. You probably don't get money from Sony for this.
The PS5 most likely just breaks even, given that Sony said they didn't have room to cut prices and Microsoft have admitted they were making a loss per unit. The extra BOM for the bigger APU and bigger SSD shouldn't be more than $150 and could be as low as $100, depending on what margins AMD charge. So we would expect Sony to be making at least $100 in profit per unit.
