because clickbait. and with the fervor of new systems, they gotta throw in pieces with nintendo somehowWhy do we keep getting Samsung.AMD and Switch 2 OPEDs? Here's another one from today
Nintendo Switch 2 could be a true powerhouse with this breakthrough
Nintendo Switch 2 could tap into the joint power of AMD and Samsungwww.tomsguide.com
Why do we keep getting Samsung.AMD and Switch 2 OPEDs? Here's another one from today
Nintendo Switch 2 could be a true powerhouse with this breakthrough
Nintendo Switch 2 could tap into the joint power of AMD and Samsung
Yeah there is a lot of unrealistic expectations about AMD, the Unreal Engine 5 demo ran 1440p at 30fps with dynamic resolution and one of epic's developers confirmed that their 9tflops RTX 2080 powered laptop ran the demo better.
AMD invested a lot in their CPU business over the last decade, their GPU investment was comparatively small. Nvidia has never stopped investing large amounts of money on their architectures and unlike Intel and AMD, Nvidia doesnt sit on their architectures for 3+ years, they are constantly changing out to newer, better technology. They aren't lazy and they have both the best graphics hardware engineers and graphics software engineers around. AMD has used open source projects to try and catch up and they just tried to turn Radeon rays into a closed software environment but the backlash had them open the code up partially.
That said, 2080 is on par with 2 years older 1080ti without RT or tensor...
It's the laptop 2080, a 9.4tflops gpu, which is a lot closer to the 2070 super's 9.1tflops than the 2080's 10.6tflops. Though pascal has aged, the Turing 2070 super does trade blows with it in games that use new features like vrs.
Yup. The fact that TVs are still 16:9 and the main point of the Switch being able to play portably and on your TV and have a comparable experience rules out any chance of a wider Switch screen. Devs also would have to invest additional time to optimize their games for changing ratios on the fly.
There's only so much you can do on the same node after all. Nvidia didn't put their efforts into per core performance but rather future-proofing Turing with DX12U support. If gaming Ampere was just a die shrunk Turing, I honestly wouldn't be surprisef
What would the supposed needed gpu teraflops be considering that ps5's objective is 4k@30fps while portable switch 2 is likely aiming for an upscaled 1080p@30fps?
A76 cores already match Zen architecture iirc, but the clock speed is limited. I'll be frank though, there shouldn't be a CPU issue with a Switch successor, A78 cores on 5nm should be clocked over 2GHz and while they won't match PS5's CPU, you'll get close enough to deal with any sort of port. GPU offloading will also not go away either, and I think there is going to be less offloading with PS5/XSX because there isn't much the CPUs can't do in those boxes, so why kill GPU performance to save the CPU here? This is why the focus in speculation should stay on GPU performance IMO, there is a limit to what you need a CPU for in games, even exclusive PC games don't really use the CPU for that much.
www.nintendolife.com
www.nintendolife.com
If we accept the suggestion that an 8 TF NVIDIA GPU equals a 10.3 TF AMD (PS5) GPU, then the speculated 3 TF GPU would be about 40% of the PS5 in terms of GPU. A drop from 4K to 1080p is a 4x resolution drop, so that gap should be easily bridged by the 3TF NVIDIA GPU in the Switch 2 (of course, GPU load doesn't scale 1-to-1 with resolution, but a 4x drop means a very big difference in rendering cost). Let put that up front: I think you should be able to create a 1080p that is of similar quality as the PS5 has. However, the discussion for the graphics doesn't stop there. First off, we want to do DLSS2.0 in order to construct a 4K image? Well, a portion of the rendering frame budget needs to be allocated to this, since it can't be done in parallel with the creation of the image itself. However, at 30 fps, DLSS takes a smaller portion relatively speaking, because the time to construct DLSS is independent of the frame rate. Check out this video at the 11:58 mark: you can see that using DLSS takes only 2.5 ms on an RTX 2060 Super. Let's be conservative and say that the Switch 2 GPU has enough RTX cores to do the process in 4 ms. That leaves 29 ms of frame time for the rest of the process. However, instead of rendering a 4K image, it only needs to render a 1080p image in that time. The video suggests that cutting native rendering from 4K to 1080p reduces the workload by about 2.7x (16 ms -> 6 ms). So, you have a workload of 2.7x as small, to be done 88% of the time. This means that your GPU capability needs to be about 42% of the PS5. So, a 3TF NVIDIA GPU seems almost good enough to produce a native 1080p image upscaled to 4K via DLSS2.0 using the same settings as the PS5. Drop these settings a notch, and the Switch 2 is able to run native gen 9 games really well. Of course, plenty of games on PS5 are probably going to opt for upscaling methods as well, instead of rendering at native 4K, so things will need to be pared back a little further on Switch 2. But it doesn't seem like a very big problem.
I don't understand why people in this thread keep writing that Nintendo won't upgrade their storage.
Their digital sales have been skyrocketing in recent years, they bring in more money than physical sales and are only limited by internet access and local storage: why wouldn't they try to solve one of the bottlenecks?
Nintendo's Digital Sales Are Nearly Double The Amount Of The Previous Fiscal Year
Up and upNintendo Earned Almost $2 Billion In Digital Sales This Year, A 72% Yearly Increase
And mobile earnings have jumped, too
They don't have that luxury with game dev paradigms changing thanks to faster ssds being the defacto storage. The problem is they don't have an expandable storage solution that can keep up without someone getting fucked on costs
So just like PS5 and XSX? There doesn't need to be a change, Nintendo has supported SD cards since pretty much the moment they were introduced, they will likely keep going and just tell the end user that X AAA game needs to be installed on internal memory (just like Sony is planning to do), there will be plenty of games that will run just fine on an SD card though, I don't expect Super Mario Party 2 to require 2GB/s data streaming.
The XSX/PS5 CPU is based on the Ryzen 4900u I thought? A 35 watt 8 core, 16 thread, with 8CU GPU at 1.75GHz. You are actually likely looking at 15 to 20 watts at the 3.5GHz for just the CPU in PS5.
Do you happen to have some indicators for how to compare different CPUs with one another? I've been trying to find a good one, and happened upon Geekbench. That website suggests that the A76 single core performance at 2.00 GHz is 2/3 that of the Ryzen 3700x single core performance at 3.6 GHz. Those numbers are wildly off from what you suggested, so I'm doubtful this benchmark gives a good indication (or perhaps I'm misreading the provided information).
Your information is correct for that benchmark, it's a general performance indicator that is definitely better than just guessing some round number based on nothing but power draw or expectations, however it doesn't give you how that CPU performs in a game, it should be somewhere along those lines, but since different benchmarks can find different bottlenecks, there is no great way to do it.
I don't understand why people in this thread keep writing that Nintendo won't upgrade their storage.
Their digital sales have been skyrocketing in recent years, they bring in more money than physical sales and are only limited by internet access and local storage: why wouldn't they try to solve one of the bottlenecks?
Nintendo's Digital Sales Are Nearly Double The Amount Of The Previous Fiscal Year
Up and upNintendo Earned Almost $2 Billion In Digital Sales This Year, A 72% Yearly Increase
And mobile earnings have jumped, too
Thanks, that makes sense. For the Ryzen 4900u you mentioned earlier, the benchmark is 1225 for a 2 GHz clock, so pushing that up to 3.5 GHz max would suggest a 75% power improvement per core over the A78 core, assuming a 2 GHz A78 is used. Conversely, if Switch 2 has 8 A78 cores at 2 GHz, then it would have 57% of the CPU performance that the PS5 has. Again, assuming that this specific benchmark can be extrapolated to general performance (and we don't know how well it generalises!).Your information is correct for that benchmark, it's a general performance indicator that is definitely better than just guessing some round number based on nothing but power draw or expectations, however it doesn't give you how that CPU performs in a game, it should be somewhere along those lines, but since different benchmarks can find different bottlenecks, there is no great way to do it.
However, A78 should have a much higher performance per clock over A76, and I'd suggest over 2GHz on 5nm is likely going to be a thing, since A78 is designed for both 7nm and 5nm, and I can't imagine that it will be designed to run at ultra low clocks, where something like a successor to A55 would fill in better.
Just to give an idea of expected performance gain of A78, A77 is 20% faster than A76 in some benchmarks, we can see from the first picture that A78 has a similar jump in performance expected, so if you look at that 2.0GHz A76 single core test, and compare it to the expected results for A78, you'll find that a 2GHz A78 core would get very close to the ryzen 3700 3.6GHz single core benchmark of 1254, somewhere over 1200. I'd expect Ryzen 4000 series in PS5, so whatever IPC increases found in the new Ryzen chip will be there, but PS5 is also a 3.5GHz chip, and I expect over 2GHz for A78 in a Switch.
This is again not an ideal comparison, and 8 cores @ 2GHz on 5nm is going to draw between 2 and 3 watts probably, but until A78 comes to market, we can't say that for sure, an A77 powered Snapdragon 865 with 4 A77 cores over 2.4GHz (1 over 2.8GHz) and 4 A55 cores + Adreno 650 (a 1.25TFLOPs FP32 GPU) draws only 5 watts on 7nm.
I actually think one thing Nintendo might do is clock 2 A78 cores higher as well, something like 2.4GHz on 2 cores and 2GHz on the other 6, if they use A55 cores, I'd suspect that the OS cores might be clocked differently than the ones reserved for devs, but considering 8 threads is probably enough for next generation, they might not need it.
I think the SoC power budget will go up by virtue of the screen being much more energy efficient in a Switch successor, much like the screen in the new models of Switch, the launch model saw a 7.1 watt to 9 watt power draw based on how bright you had the screen, expect something better than that, leaving room for Nintendo to increase the SoC budget from ~5 watts on Switch to ~6.5 watts on the successor.
I'm curious why you say that. Of course, the game won't magically run at 60 fps when a 30 fps lock is in place, but a small patch could very well lift that restriction and make it 1080p/60 fps, right? Unless they have tied the game's logic to frame rate, of course, in which case they would need to redo the game (but is there any indication for that?).
You don't drop the CPU clock to not mess with your game logic. Suppose you're making a racing game and you model the tire model to do one amount of calculations per second. If you drop the CPU clock you'll have a different tire model on handheld than on docked. Of course it can be done but it's just an example of why you should not do it.
Hm, if that's the case, then yeah, you won't get the game at 60 fps. It's a poor design decision to tie game logic to frame rate for this reason, and it handicaps your ability to bring the game forward easily. But well, what can you do?
That makes sense, but if your game isn't too demanding, then maybe Nintendo can just offer a complementary low-CPU mode in docked? For example, we get a low-power profile on handheld with 1 GHz CPU clock and 384 MHz GPU, and then a complementary docked profile with 1 GHz CPU clock and 921 MHz GPU clock. And then a high-power profile on handheld with 2 GHz CPU clock and 691 MHz GPU, and then a complementary docked profile with 2 GHz CPU clock and 1.25 GHz GPU clock. In that case, you can target your profile of choice (low power or high power) and still have a consistent set of two power profiles for docked and undocked.
1/3 (or even 40%-50%) is basically the same as current gen though in GPU (switch vs xbone)
The clock performance won't change that much between 3700 and 4900, not single threaded performance. The IPC gain for the new ryzen cpu is not 75%. It is likely finding another bottleneck at a higher clock that we aren't seeing, however it won't be long before that cpu is on the market and we can get these scores for any clock from it.
Chances of a pro are very unlikely.
The real question is 'are 4xA55@2Ghz good enough (=4xA57@1Ghz) for running OG switch games?'. If so, I don't see Nintendo not wanting them.
