Sony and AMD have announced a new partnership aimed, unsurprisingly, at improving the technological basis for high-quality graphics at greater performance and for better gameplay.https://www.notebookcheck.net/AMD-hardware-on-board-Mark-Cerny-provides-an-early-peek-at-the-PlayStation-6-gameplay-to-benefit-from-AI.934889.0.html
And when will RDNA 4 come out? Everyone is waiting for it. Possibly the next console will be with ZEN 6 + RDNA 4 but first the AMD 8000 graphics with RDNA 4 have to arrive.
RDNA 4 won't be a big bump over RDNA 3.5
RDNA 5 is supposed to be "the good bump".
Quote from: heffeque on December 19, 2024, 21:07:29RDNA 4 won't be a big bump over RDNA 3.5
RDNA 5 is supposed to be "the good bump".
RDNA5 (or UDNA now I guess?) is supposedly going to be a major leap for high end yeah, but where are you getting that RDNA4 isn't a significant improvement from?
From what I've read and heard online it is has a lot more changes and newer GPU instructions to rdna3 than rdna3.5 had to it.
It is also significant step up in RT workloads.
Rdna3.5 is actually the one that feels a little underwhelming. Looking at the benchmarks for 890m it's barely any faster than 780m at similar lower wattages. Only at high wattages it's a bit faster. The 880m with same amount of CUs actually shows zero improvement, in some cases regressions. Rdna3.5 feels even worse than rdna3 which was already barely any improvement from rdna2.
I've high hopes for rdna4. Kind of wish it was in strix halo but deadlines had to be met I guess.
Quote from: RumourMill on December 20, 2024, 01:42:58RDNA5 (or UDNA now I guess?) is supposedly going to be a major leap for high end yeah, but where are you getting that RDNA4 isn't a significant improvement from?
From what I've read and heard online it is has a lot more changes and newer GPU instructions to rdna3 than rdna3.5 had to it.
It is also significant step up in RT workloads.
Rdna3.5 is actually the one that feels a little underwhelming. Looking at the benchmarks for 890m it's barely any faster than 780m at similar lower wattages. Only at high wattages it's a bit faster. The 880m with same amount of CUs actually shows zero improvement, in some cases regressions. Rdna3.5 feels even worse than rdna3 which was already barely any improvement from rdna2.
I've high hopes for rdna4. Kind of wish it was in strix halo but deadlines had to be met I guess.
I take it from the fact that AMD won't even try to make a high-end version of it. Most people could care less about RT, so it'll probably not be much of an improvement elsewhere.
Quote from: RumourMill on December 20, 2024, 01:42:58Looking at the benchmarks for 890m it's barely any faster than 780m at similar lower wattages. Only at high wattages it's a bit faster. The 880m with same amount of CUs actually shows zero improvement, in some cases regressions. Rdna3.5 feels even worse than rdna3 which was already barely any improvement from rdna2.
I've high hopes for rdna4. Kind of wish it was in strix halo but deadlines had to be met I guess.
But wait!! Read elsewhere and 'Strix Halo' is going to 'trade blows' with RTX4070's!! LOL!!!!
ROTFLMAO!!!
Yeah, I have this bridge for sale...
Maybe when RDNA 4.0 comes along it MIGHT be >%10-15 improvement but I would not hold your breath.
A process die shrink & optimization can only go so far.
Besides AMD's 'higher end chips' are met to be coupled with external dGPU's doing the 'heavy lifting'. I'm guessing whatever space+thermal+power budget AMD could of tossed the iGPU's way was spent on a 'NPU' of some such for WIN11 'AI' functions. :)
Quote from: heffeque on December 20, 2024, 01:53:45the fact that AMD won't even try to make a high-end version of it. Most people could care less about RT
Rx 480 performed like a R9 290x at half the wattage. Don't think we're gonna see efficiency gains like that as node improvements aren't as substantial now as they were back then but point being is don't think you can draw conclusions like that just because of lack of high end. I mean, the rx 7000 series has high end and is widely considered a dud because it barely competes with Nvidia's midrange.
Personally I agree about RT but we're entering 2025 now and landscape is changing. There are already games setting RT capable hardware as minimum prerequisite. The switch 2 is also RT capable which will make developers target it as the minimum specification and make it even more commonplace in games.
There are other factors too why AMD wouldn't like to go big expensive die and high end in consumer market. Because every time they do, they just dont sell well at all. The typical demographic that goes for $1000+ GPUs are not gonna settle for second place in compute / AI software stack, upscaling and RT. AMD probably waiting for their software stack to be more polished first which takes time before going at high end again.
Besides you compared to rdna3.5 which definitely isn't high end either.
Quote from: GeorgeS on December 20, 2024, 02:24:30ROTFLMAO!!!
;) fair.
Tbh, ignoring all the hype, I'd just be happy if it's roughly equivalent to a desktop RX 6600. Because that's still 300%-400% faster than the next closest igpu in certain games. And it should be having comparable CU count to ps5's. Even if we assume rdna4 brings 0% improvement.
I've been waiting for a 40 CU igpu part for a very long time now, since 2016 PS4 pro. What architecture it is whether it's rdna 2, 3, 3.5 is not as important to me as just having a much bigger GPU and higher raw bandwidth.
Not so interested in dgpu in mobile space as it causes a whole bunch of other issues.