RDNA 3 doesn’t actually have much in the way of architecture or clock speed improvements. RT is improved but that's about it.I wonder how much improvement it can bring. 5nm process more than doubling the transistors? AMD upping the clocks to get closer to 3GHz on console-level chips? RDNA3 arch changes exploited to the hilt? Infinity cache? FSR3 for consoles?
Doubling of 5700XT performance lands at 3080 level of performance in this 4k chart of latest graphics card review at TPU. RT performance should be close to 3060Ti level.
ASUS Radeon RX 7600 Strix OC Review
The ASUS Radeon RX 7600 Strix OC comes with the best cooler of all RX 7600s tested so far and achieves unbelievable noise levels. Our review confirms that thanks to its large factory overclock, the Strix is also the fastest RX 7600 available at this time.www.techpowerup.com
Do you think there will be a mid gen refresh console from Sony and Microsoft?
- Thread starter invictis
- Start date
VFX_Veteran
Regular
How is there no doubt? Did they announce something that I missed?
VFX_Veteran
Regular
Concerning mid-gen refreshes...
I always thought that last gen was a special case and not the beginning of a new norm. I can understand why console lovers would want it to be the new norm to help keep up with the rapid growth of the PC GPU market, but unfortunately that's not the case.
At this stage of the game, you'd need much more than a mid-gen kind of upgrade to garner 60FPS, frame generation (an absolute must), a good AI upscaler, and proper hardware RT. Quite frankly the upgrade would have to be massive.
If I were Sony/MS, I'd weather out the storm on current gen and just end it early (2026). This would all depend on AMD, which is unfortunate, as they are the main hardware vendor for new tech coming to the consoles. They are always playing catch-up to Nvidia and that's not a good place to be in.
I always thought that last gen was a special case and not the beginning of a new norm. I can understand why console lovers would want it to be the new norm to help keep up with the rapid growth of the PC GPU market, but unfortunately that's not the case.
At this stage of the game, you'd need much more than a mid-gen kind of upgrade to garner 60FPS, frame generation (an absolute must), a good AI upscaler, and proper hardware RT. Quite frankly the upgrade would have to be massive.
If I were Sony/MS, I'd weather out the storm on current gen and just end it early (2026). This would all depend on AMD, which is unfortunate, as they are the main hardware vendor for new tech coming to the consoles. They are always playing catch-up to Nvidia and that's not a good place to be in.
Sony have a vastly larger total userbase to sell a Pro model to, VR, and they don't control the Windows / PC ecosystem. So a Pro could well make sense for them.
I'm not sure it does this time for MS.
A reworked series X may potentially make sense though. An extreme version of this might be a shrink to 5nm, and going with a deterministic boost like Sony could see them hit 2.5+ ghz. Double up the ROPs, double caches, move to 18 ghz GDDR5, maybe use all 2 GB chips for 20GB total (and uniform striping across all memory) and allow native support for 8K displays. Avoid any kind of differences in the programmable parts of the chip to minimise developer work. Basically replace the current model with a slightly enhanced version. Can't see something this extreme happening, but fun to think about.
Or MS could use increased clocks to reduce the GPU by 8 CUs saving die area. This would need sustained clocks of about 2157 mhz, which should be doable on 5nm. They could also drop to a 256-bit bus with 18 ghz GDDR5, saving even more die area (PHYs and memory controllers), reducing the number of pins, reducing board complexity, and reducing the physical number of memory chips (reducing costs). Might want to increase the size of GPU L2 though as you be losing 20% of L2.
Series S just needs a shrink and clocks pushing quite a bit higher. There are currently lots of games where PS5/X hold 10 or 20% higher frame rate advantages during stress points, and addressing this would make the S a much more attractive platform with no extra work required from developers.
I'm not sure it does this time for MS.
A reworked series X may potentially make sense though. An extreme version of this might be a shrink to 5nm, and going with a deterministic boost like Sony could see them hit 2.5+ ghz. Double up the ROPs, double caches, move to 18 ghz GDDR5, maybe use all 2 GB chips for 20GB total (and uniform striping across all memory) and allow native support for 8K displays. Avoid any kind of differences in the programmable parts of the chip to minimise developer work. Basically replace the current model with a slightly enhanced version. Can't see something this extreme happening, but fun to think about.
Or MS could use increased clocks to reduce the GPU by 8 CUs saving die area. This would need sustained clocks of about 2157 mhz, which should be doable on 5nm. They could also drop to a 256-bit bus with 18 ghz GDDR5, saving even more die area (PHYs and memory controllers), reducing the number of pins, reducing board complexity, and reducing the physical number of memory chips (reducing costs). Might want to increase the size of GPU L2 though as you be losing 20% of L2.
Series S just needs a shrink and clocks pushing quite a bit higher. There are currently lots of games where PS5/X hold 10 or 20% higher frame rate advantages during stress points, and addressing this would make the S a much more attractive platform with no extra work required from developers.
VFX_Veteran
Regular
Yea.,... no. Too much work and not enough power for full on RT. Not worth the investment. Might as well design the PS6 now and have it release in 2026.
I can't stress enough that the consoles are at the mercy of AMD for any new tech. Frame generation is an absolute must going forward. There simply isn't enough bandwidth to allow for path-tracing or full RT with Nanite type geometry without some sort of AI adding frames that are lost. Trying to solve that problem now just doesn't seem realistic - especially since it's not even available for their current high-end GPUs.
So that's why ryzen 7840u igpu performance is very similar to 6800u
Forward compatibility is the worst.
It completely eliminates the 'generation' model, which has been the foundation for the more fundamental improvements in games we've seen over the years(on a technical level). Get rid of that, and you basically just have a perpetual cross-gen situation, which is lousy. People generally dont want to buy new, $400-500 consoles to play games they already could before, they buy them for the promise of games that couldn't even exist before. That's the excitement.
No it’s not. Not every ps5 owner is an owner of past generations of PlayStations. Same goes for the Xbox and Nintendo consoles. And pubs can’t afford to remake or remaster every classic title out there. One of my son’s favorite games is Fight Night Champion. It’s part of a genre that gets no love on modern consoles. BC removes that barrier and allows gamers access to past titles without jumping through a bunch of hoops. FC allows titles to perform better with newer hardware.
Plus cross gen is a product of market demographic not modern hardware capabilities. No one is building games with last gen performance because the XS/PS5 consoles can play last gen games. They are building cross gen games because there are a significant number of users still on 1s and 4s.
Last edited:
gamervivek
Regular
RDNA3 has certainly come a cropper when it comes to clockspeeds, which is a disappointment. I hope that this will be rectified like they did with PS5 and subsequently RDNA2 over RDNA1.
As for architecture, even if it doesn't improve over RDNA2, just the RT advantage can allow them to market pro as 'full RT' console. Doubling of current-gen to 6900XT level of performance will allow Cyberpunk Ultra RT to run at 30fps with FSR2. That RT advantage can come in handy if they don't even hit 6900XT level in raster.
Similarly, other games with RT based GI like Dying Light 2 and Witcher 3 update can be played at PC-capable RT settings and not just shadows/reflections.
The consoles with the best backwards/forwards compatibility as a product line have been Nintendo's Gameboy and DS line. They dominated the competition and had plenty of impressive titles. Resident Evil Revelations released on 3DS in the US a few weeks before the PSVITA and looked on par with the Vita launch titles, despite the 3ds hardware being essentially a generation behind.
The 7900 series just barely qualify as full RT GPUs. There is no way to put close to that level of performance in a console any time soon.
gamervivek
Regular
I don't see next-gen consoles going over 3GHz even in the more optimistic scenario of RDNA3 having some issue that is preventing it from hitting the max clocks it can in gaming. Consoles are usually clocked hundreds of Mhz lower than desktop cards and RDNA2 had the gap unusally wide by the time the smaller chips came out and were close to the 3Ghz mark.
And then there's the problem of fitting this clock with console-level power restrictions.
The PC market does not accept 30fps, but consoles can. 6900XT can easily get 30fps with FSRQ at 4k and full RT in Dying Light 2. It'd be a decent upgrade for the pro version.
That used to be the case. Playing your library in better quality is now a value to gaming. One would have to measure what the market prefers. Pretty sure a super scientific poll here would show quite the reversal on the one carried out years ago that showed people didn't care for BC and now they do.
Its almost as if folks didn't care for it only because they were told they didn't need it by one console provider.
Disagree. I didn't care for it up to PS4 because games were radically different on radically different architectures and you had to have a clean break to use that hardware well. I played new games, not old ones. Now architecture is homogenised hardware designs constrained by diminishing returns and expensive development that favours multiplatform development, a clean break offers far less gains for the loss of BC. It no longer makes sense. Plus GaaS is very much apart of modern gaming where it wasn't in previous generations with discrete game iterations.Its almost as if folks didn't care for it only because they were told they didn't need it by one console provider.
That BC/FC no longer makes sense now doesn't mean it didn't make sense 10+ years ago for lots of gamers as well as some platform holders. Nor does feeling differently about it now mean I was being spoon-fed beliefs by a corporate entity and incapable of deciding for myself rationally.
As a PC gamer, I agree. But the Steam Deck is also a successful device. Lots of people are using that as a 30FPS device.
I’m seeing lots of drops to the mid 20s with nothing even happening on screen.
Subtlesnake
Regular
Navi 32 is likely outside the cost and power constraints required for a $499 (or $599) console. Microsoft's reason for making the S was that the price/transistor was not expected to come down significantly on 5nm, and we have seen what has happened to the pricing of 5nm PC GPUs.
Frenetic Pony
Veteran
It'll be fine this gen. What, exactly, is any "next gen" set of consoles going to add in terms of gameplay potential? Basically nothing anyone would care about for quite a while.
Sony and MS need to give up the current perception they stuck with for this console gen. My sister wants to play the new Fable, and would end up buying a Series S if it had Viva Pinata has well. But she bought a Switch just for Animal Crossing without hesitation, in large part because it was portable. People love that, they'll pay for that. It's a much easier and more obvious sell than some nebulous gameplay improvement over whatever current gen game might even take advantage of this gens CPU/RAM for gameplay. And it gets harder and hard to sell on better graphics. UE5 and Star Wars Outlaws already look great, how much better are you really going to get. Is it really $500 better, and how many people would consider it $500 better?
The smartest thing MS could do is put out a Series M (for mobile) that matches Series S specs around say, 2026 for $399 (technically it could be earlier if they went 15w, but that'd make the thing a bit too big/heavy for truly mass market). Maybe a Series X mid gen referesh that doesn't cost any more at the same time. Zen 6 and that new GPU arch they have after RDNA4. Look 8k, and/or 3d, or whatever the cool visual specs are then.
Sony obviously has a harder time of it, they might not be able to hit a mobile PS5 till 2028+. But if they do it could include a 4k OLED (micro-led?) and double as the screen and processing unit for a VR headset. Make the controller break apart from the body like Switch, and you've got most of the components of a VR set in people's hands just from buying the console. VR might get much more adoption if the "headset" was just an add on that cost <= $200.
This would also get a lot of support from devs. They're waiting a while to go "This gen only" because of the absolutely massive install base of last gen. And this gen is going to have an even bigger install base by 2028. Asking devs to give that up is going to be a tough sell at best, better to simply embrace that and give people different reasons to upgrade.
Last edited:
Similar threads
