Sweet! So DLSS isn't just good for graphics, but also power savings for a handheld device! We need this tech!!!
I'm not sure myself. Looking it up, it seems like the formula to go from clock rate to bandwidth is (clock * bus_width), divide by 8 if it was in bits to convert to bytes, then multiply by a constant that depends on the memory type. The issue I had was that when I tried to put in the PS4's numbers into that, it wasn't working out.Oh yeah, that was something I was wondering.
How do you compare the speeds of normal DDR/LPDDR to GDDR?
The main thing is the Xbox One S used normal DDR3 with a cache of 32mb of eSRAM while the PS4 used 8 GBs GDDR5 yet the main thing that weakened the One S was the raw GPU cores relative to the PS4 iirc.
Asking as it could be a big factor in comparing a Switch 2021 using LPDDR5 VS the PS4 using GDDR5 and the PS5/Xbox Series S | X using GDDR6.
The solution is to end the pandemic so that demand goes back to reasonable levels.Ah, thank you both for the clarification. Is there any kind of long-term solution for fixing this global shortage crisis (hopefully beyond further stripmining the Earth of resources that continues to accelerate climate change)?
If your frame times are already that low, then DLSS probably has nothing to offer. At a certain point, the upscaling cost is going to exceed any potential benefit you could get by lowering resolution.Edge of Eternity. I brought it up with Dictator and he said an Nvidia engineer told him that rare instances would crop up like that. don't think it's really applicable for the Switch Pro. EoE was already running at 200fps before DLSS. the tensor cores probably just couldn't keep up
You could say that.Sweet! So DLSS isn't just good for graphics, but also power savings for a handheld device! We need this tech!!!
April is when we should have heard something by, earlier is possible, but should be in full production by April.Everytime I see you post about what you've heard and theorize this device will be, I can't help but get insanely excited haha.
Sorry if it's been asked before, but assuming this beauty is coming later this year (fall perhaps), when should we expect to get some leaks about it's manufacturing?
So I'm going to be running with my specs for now, I think 8 A78C cores at a lower clock makes more sense for an SoC designed for games, than 6 A78C cores at a higher clock.
CPU: 8*A78C @1.5GHz. (~1/3rd PS5)
GPU: 768 Cuda cores
@1.1GHz. (1690GFLOPs) docked.
@600MHz (921GFLOPs) portable.
RAM: 8GB @ 102GB/s
Storage 256GB @ 850MB/s (ufs 2.1)
MicroSD(express) for expandable storage.
Docked would be pretty similar to PS4 in raw performance, the clock is higher than I've been suggesting, but still very reasonable for power consumption while docked.
Portable would be about double the current Switch's performance, targeting 480p and using DLSS to reach 720p would allow it to achieve really great results on a 7inch screen. Battery life minimum should be above 3 and a half hours, I don't have a good way to calculate this given 8nm chips and Ampere really don't come in configurations anywhere close to this, but going by previous designs, this should consume less than 7 watts.
The transistor difference isn't that great between 6 and 8 A78C cores, but as to your other point, there is no power/thermal reason, that is controlled via the clock of the CPU. Also with DynamIQ, one CPU will effectively be 'off' at a minimum clock until the OS is called to do something.Wouldn't a 6 core CPU be both cheaper (smaller) and run cooler than an 8 core? Even if 8 is better for gaming 6 A78s would be an enormous jump as is and definitely remove the current CPU bottleneck, while being cheaper too.
I don't think they have much of a reason to spring for 8 when 6 would be more than adequate for what this machine is trying to do.
I'm guessing having 8 CPU cores (or 7 usable CPU cores with 1 reserved CPU core for the OS) would make it at least a bit easier to port some PlayStation 5 and Xbox Series X|S games (especially the "miracle" ports) to the next Nintendo Switch model, given that the PlayStation 5 and the Xbox Series X|S have 8 CPU cores (The Xbox Series X|S has 7 usable CPU cores with 1 reserved CPU core for the OS whilst the PlayStation 5 is rumoured to have 7 usable CPU cores and 1 reserved CPU cores for the OS as well).Wouldn't a 6 core CPU be both cheaper (smaller) and run cooler than an 8 core? Even if 8 is better for gaming 6 A78s would be an enormous jump as is and definitely remove the current CPU bottleneck, while being cheaper too.
I don't think they have much of a reason to spring for 8 when 6 would be more than adequate for what this machine is trying to do.
Wouldn't a 6 core CPU be both cheaper (smaller) and run cooler than an 8 core? Even if 8 is better for gaming 6 A78s would be an enormous jump as is and definitely remove the current CPU bottleneck, while being cheaper too.
I don't think they have much of a reason to spring for 8 when 6 would be more than adequate for what this machine is trying to do.
Total amount of heat wouldn't be dictated by # of cores so much as the power drawn. If you're drawing X watts for CPU in total, that's X watts of heat to dissipate regardless.
Heat density should be better if those X watts are spread out over a larger area though, I think?
Yep, 7 cores at 1.5GHz is about the same as 5 cores at 2GHz, so while individual threads would be better with 6 cores, 8 cores has a lot of benefits, but the biggest is industry parity. Consoles, PCs and mobile are all generally 8 cores, so the lower clock would effectively allow industry parity without creating extra heat.The 8 cores vs 6, 8 would definitely give more parity across the board with XboxOne/PS4 games and even potential PS5/XSX games. Since all of these systems have an 8 core CPU set-up with 1 dedicated to the OS and 7 for games, having the same for this revision and even Switch 2 will make porting games much easier than a system that has 5 cores for gaming.
Plus I think the ARM cores become way more efficient energy wise when using clocks under 2Ghz, which would bode well for a device like Switch that needs sustained performance across handheld and docked modes.
On the previous subject of VR, I would argue that the Switch is the most successful or one of the most successful VR devices in the dedicated gaming device space, while not being a dedicated VR device.
Or more specifically the Switch form factor lends itself a lot easier to do VR when it is not a dedicated VR device. VR in that sense is something they can pursue (not that they will), but not as a dedicated VR device. Oculus Quest2, if you alter it a little, is roughly what a switch "pro" would be IMO. Or I should say, a switch pro suited for VR like the Oculus Quest 2 would not be that different. Of course, that would be a separate peripheral that alters the viewing lens for VR just like how the Ring for Ring-Fit is a totally separate peripheral. Also, it would need a higher quality screen to make it work well, but that goes without saying.
I use Oculus Quest 2 because the specs it has is more in line of what can happen for a switch Pro and still be rather good of an upgrade. 1.2TFLOPs (I think?) 6GB RAM, a custom snapdragon SoC meany to handle more intensive tasks (like gaming related), split controllers that can function in either hand like the joycons, a screen that is a lot higher Res, and 64GB of onboard storage.
Like, they can do the standard switch pro(or 2) with beefed up specs in screen, internal hardware, etc. That can be 300 bucks like the Oculus Quest 2 (64GB model), a peripheral like the ring fit is but for docking the switch to the headset portion and the ring attachment for the joycons on the railings suited better for VR. They'd need to have a higher quality screen though, whether they choose regular degular standard or not is irrelevant really, as it has to be a higher quality screen in the end.
What fueled this was that DLSS 2.1 has VR support for it, so while the switch that I am describing is the switch being described here (mostly), I'm describing only a slight tweak to it and a separate peripheral in the same vein as the Ring-Con that makes the device that has the docking concept but taking it more dynamically for a headset. Switch makes it really flexible for this.
I already argued this before that they can do this, not that they will do this. But I hope it is clear how a Device that can do VR reliably can work, not necessarily a dedicated VR device from them.
Bait02
You could say that.
I'm guessing because of the preliminary annotation of the Orin die?Thanks for the answers, for some reason I thought A78 cores were relatively large, so cutting down two of them would save a non trivial amount of money.
The reason I said a peripheral is that it eases the idea more easily to the consumer who wouldn't mind playing with that, rather than a dedicated VR device which people have trouble putting on their head and that's the caveat. People don't really like things on their head like that. And the peripheral I had in mind would be a 100 dollar peripheral lol so with thought put into it, has the lens and since the switch has gyroscope in the tablet and it can power the peripheral itself to incorporate more functions.They could certainly release a VR peripheral designed for use with a future Switch model but, as someone who thinks it would be borderline tragic to see VR gaming either fade away or go all to Facebook without Nintendo taking a thorough stab at it, I'd much prefer a standalone device. How far do they go with the peripheral? Do they just release a headset dock with VR lenses for cheap that you can slot the tablet into for a 3DoF solution ala Labo, GearVR or Daydream (rotational motion tracking via the gyroscope without freedom of movement)? Or, do they go a little more pricy and try to tack on some cameras to facilitate a full 6DoF solution? If they go for number 2, how expensive is that peripheral device?
I'd expect it might be pricy enough to make the value proposition look somewhat questionable next to a Quest. On the other hand, Mario Kart Home Circuit includes a pretty nice RC car with an AR camera for only $99.99. If Nintendo's peripheral answer to PSVR was somewhere around that price and provided a full 6DoF experience, it could be pretty compelling.
My other concern with going the peripheral route though, like you brought up, is the screen. Quest 2's display is 1832 x 1920 pixels per eye @ 90hz. It's just really hard to imagine Nintendo going beyond 1080/60 on the next iteration of their $299 tablet form factor. However, going all out on a premium standalone device might expand that horizon a bit. The difference between my Quest 2 and the 1080p/120hz display in my PSVR is drastic. Nintendo wouldn't necessarily have to go all the way to Quest 2's near 4K display, but we need something better than 1080/60 for Switch VR to be a viable platform as opposed to something more short-lived, imo.
Since dynamiq allows assigning cores to different voltage/frequency domains within the cluster, I think that in actual practice, it's probably the case that the core reserved for system tasks gets its own domain. Therefore, in real time its clock rate is probably going up and down all over the place depending on need (and independent of the game cores). So saying that the system core runs at a set rate wouldn't technically be accurate.I don't really see the need of clocking the 1 core that would be dedicated to the OS so high to 1.5 GHz when a game only at most would be using 7.3 cores or 7.5 cores and that's absolutely pushing it for them, isn't it more likely that 7 cores are clocked at 1.5GHz and 1 core at 1.07GHz which itself saves energy as well?
They raised it to 384MHz iirc, some games use 460MHz like mortal combat 11.Hmm, looking at it again, 600 mhz for the GPU when undocked seems a bit high? If wikipedia's right that the Switch's GPU runs at 307 mhz when undocked, then that's near doubling it. In turn that's near quadrupling the power requirement before adjusting for architecture and increase in SM count (+50% SMs?).
(I'm tentatively assuming that going from 20nm generation to 10/8 nm offers about 4x power to work with. Thus why I think 8 [email protected] slides right in just fine after architecture adjustment... and that the 1.1ghz for the GPU docked actually leaves a bit of room, I think.)
I'm guessing that since all of the CPU cores on the Tegra X1 on the Nintendo Switch seem to run at 1.02 GHz, I think he's making an assumption that the all CPU cores on the next Nintendo Switch model's SoC are going to run at the same frequency.I don't really see the need of clocking the 1 core that would be dedicated to the OS so high to 1.5 GHz when a game only at most would be using 7.3 cores or 7.5 cores and that's absolutely pushing it for them, isn't it more likely that 7 cores are clocked at 1.5GHz and 1 core at 1.07GHz which itself saves energy as well?
Switch is anemic in a good way where it gives the devs the most resources without being a blank piece of paper despite its deficiencies. Same for RAM, due to the switch anemic OS nature, it likely would not need to be 2.5GB to 3GB of RAM away from a hypothetical 8GB RAM pool, but at most 1GB dedicated to the OS giving devs 7GB RAM available to devs vs the Xbox One that had 5.5GB for devs I believe. Or was it 5GB? Either way I don't really expect it to necessarily be 1 to 1.
Since dynamiq allows assigning cores to different voltage/frequency domains within the cluster, I think that in actual practice, it's probably the case that the core reserved for system tasks gets its own domain. Therefore, in real time its clock rate is probably going up and down all over the place depending on need (and independent of the game cores). So saying that the system core runs at a set rate wouldn't technically be accurate.
...but it's easy/simplification for discussion to just say that it's running at the same clock as every other core. And easier to calculate stuff like power draw if you don't have to break it up into multiple pieces.
That's true, it mostly seems for discussion sake, though in practice Nintendo may opt to not go that high for that core strictly for better energy management in the end. At least, I believe that the chance of it being that high are slim and better off used in a different way.I'm guessing that since all of the CPU cores on the Tegra X1 on the Nintendo Switch seem to run at 1.02 GHz, I think he's making an assumption that the all CPU cores on the next Nintendo Switch model's SoC are going to run at the same frequency.
And it was 5 GB of RAM that was accessible to developers for the Xbox One.
The Snapdragon 690 is very similar to the Exynos 880, except the Adreno 619L GPU is superior to the Mali G76 MP5 and UFS 3.0 is also supported.We don't have many devices on 8nm to compare to, but this Exynos 880 is interesting because it's made on the Samsung 8nm process has 8 core CPU(2xA77+6xA55), 8GB of LPDDR4x RAM, 576Gflops of GPU performance, UFS 2.1 storage, 1080p display and the entire phone draws 7w. So take out the things that Switch wouldn't use such as a modem and the potential for what's possible on 8nm is pretty interesting to think about.
Samsung Exynos 880: specs and benchmarks
Samsung Exynos 880: performance tests in benchmarks (AnTuTu 10, GeekBench 6). Battery life and full specifications.nanoreview.net
So I'm going to be running with my specs for now, I think 8 A78C cores at a lower clock makes more sense for an SoC designed for games, than 6 A78C cores at a higher clock.
CPU: 8*A78C @1.5GHz. (~1/3rd PS5)
GPU: 768 Cuda cores
@1.1GHz. (1690GFLOPs) docked.
@600MHz (921GFLOPs) portable.
RAM: 8GB @ 102GB/s
Storage 256GB @ 850MB/s (ufs 2.1)
MicroSD(express) for expandable storage.
Docked would be pretty similar to PS4 in raw performance, the clock is higher than I've been suggesting, but still very reasonable for power consumption while docked.
Portable would be about double the current Switch's performance, targeting 480p and using DLSS to reach 720p would allow it to achieve really great results on a 7inch screen. Battery life minimum should be above 3 and a half hours, I don't have a good way to calculate this given 8nm chips and Ampere really don't come in configurations anywhere close to this, but going by previous designs, this should consume less than 7 watts.
April is when we should have heard something by, earlier is possible, but should be in full production by April.
I have a question. Nintendo doesn't like loss, and this will be probably max 350$ device. So compromise is the name of the game.Yep, 7 cores at 1.5GHz is about the same as 5 cores at 2GHz, so while individual threads would be better with 6 cores, 8 cores has a lot of benefits, but the biggest is industry parity. Consoles, PCs and mobile are all generally 8 cores, so the lower clock would effectively allow industry parity without creating extra heat.
For a brand new device, they will have lower margins in the beginning imo than something that's been on the market for 4 years.Nintendo currently have great profit on current Switch models (they were selling with profit every Switch unit from day one), by now (4 years later) its easily over $50 of clear profit per unit, so you can bet that with any revision Nintendo will release they will want to have similar profit and in same time not having high selling price point.
Onother point people should have on mind when they talk about hardware, specs and selling price point.
2027 would be a little to late. By 2027 the VR trend will either die off or someone else will take the lead (probably Sony or Facebook).Super Switch first please.
If Nintendo ever does any VR/AR headset, it will be an entire platform, not a peripheral, and probably won't happen for another 5-6 years after Super Switch (2021/2022) meaning 2027.
For a brand new device, they will have lower margins in the beginning imo than something that's been on the market for 4 years.
They will sell it at a very low profit margin initially, which I think they should do because lower priced hardware will sell more and eventually will lead to a lot more software sales. So much so that it's pointless to sell hardware at a high profit initially. When momentum goes up, the cost of the hardware will come down naturally and profits will increase per unit. The first task is though to get the momentum going and sell lots of units as to create a software monster.Little lower margins, but its certain they will want to sell it with clear profit from day one in any case.
They will sell it at a very low profit margin initially, which I think they should do because lower priced hardware will sell more and eventually will lead to a lot more software sales. So much so that it's pointless to sell hardware at a high profit initially. When momentum goes up, the cost of the hardware will come down naturally and profits will increase per unit. The first task is though to get the momentum going and sell lots of units as to create a software monster.
The charger is the dock itself, so there will always be one. The grip will always be there too since they aim this at the core crowdI wouldn't be surprised if they stopped including in-box accessories like the charger and Joy-Con grip.
I agree with you. As long as this is a revision, Nintendo will not want to suddenly drop their margin in any significant way. Especially since many from the first wave of buyers of this revision will be upgraders, and these will already have their Switch games library (so Nintendo can't count on them buying a new game with the revision).Yes, but point is that they have huge profit on current hardware and they are releasing revision, so they will want at least solid profit on that revision also.
Just for record, this is not launch of next gen, but revision that will part of current Switch platform.
Then we are back to the age old question: will Nintendo have traditional gens going forward, or will they just release iterations.Yes, but point is that they have huge profit on current hardware and they are releasing revision, so they will want at least solid profit on that revision also.
Just for record, this is not launch of next gen, but revision that will part of current Switch platform.
I'm in the next gen Switch camp. I think it will be the next big step for Switch and this new model will probably last them another 4 years. That's why I think they will not care all too much for the profit margin in the first year. They need to make it the best it can be initially.Yes, but point is that they have huge profit on current hardware and they are releasing revision, so they will want at least solid profit on that revision also.
Just for record, this is not launch of next gen, but revision that will part of current Switch platform.
Yeah thats expected - pretty sure these Pro-Enhanced games will get featured prominently on the eShop as well, so most will make sure that their games come up when people look for games that have been optimized for the new device.now that Crash is announced, I think Activision will be throwing out a slew of pro patches for launch. Crash games, Spyro, maybe even a couple of digital exclusives in the form of a Call of Duty or two (Modern Warfare Trilogy and/or Warzone)
Even if "2" (iteration), what I wrote still applies - as long as many buyers are upgraders, these will already have their Switch games library, so Nintendo can't count on them buying a new game with the iteration.Then we are back to the age old question: will Nintendo have traditional gens going forward, or will they just release iterations.
If 2: effectively this is next gen even though it will be a while for next gen software to arrive
I agree with you. As long as this is a revision, Nintendo will not want to suddenly drop their margin in any significant way. Especially since many from the first wave of buyers of this revision will be upgraders, and these will already have their games library (so Nintendo can't count on them buying a new game with the revision).
Then we are back to the age old question: will Nintendo have traditional gens going forward, or will they just release iterations.
If 2: effectively this is next gen even though it will be a while for next gen software to arrive
I'm in the next gen Switch camp. I think it will be the next big step for Switch and this new model will probably last them another 4 years. That's why I think they will not care all too much for the profit margin in the first year. They need to make it the best it can be initially.
They still want to start quickly building that install base though, so that when they do make the transition software wise it's already in the tens of millions.Even if "2" (iteration), what I wrote still applies - as long as many buyers are upgraders, these will already have their Switch games library, so Nintendo can't count on them buying a new game with the iteration.
So revision or iteration, unless there is a must have first party game right away that is exclusive, Nintendo will still not be able to count on buyers getting a game with their new Switch.
They still want to start quickly building that install base though, so that when they do make the transition software wise it's already in the tens of millions.
My main issue is really just the battery life estimate. I get the feeling they'll try to keep it around the Mariko model's official estimates (4.5-9 hours), especially if the revision is an outright replacement for it. Don't really know how it'd effect the specs or performance, but I do believe that's what they're going for.So I'm going to be running with my specs for now, I think 8 A78C cores at a lower clock makes more sense for an SoC designed for games, than 6 A78C cores at a higher clock.
CPU: 8*A78C @1.5GHz. (~1/3rd PS5)
GPU: 768 Cuda cores
@1.1GHz. (1690GFLOPs) docked.
@600MHz (921GFLOPs) portable.
RAM: 8GB @ 102GB/s
Storage 256GB @ 850MB/s (ufs 2.1)
MicroSD(express) for expandable storage.
Docked would be pretty similar to PS4 in raw performance, the clock is higher than I've been suggesting, but still very reasonable for power consumption while docked.
Portable would be about double the current Switch's performance, targeting 480p and using DLSS to reach 720p would allow it to achieve really great results on a 7inch screen. Battery life minimum should be above 3 and a half hours, I don't have a good way to calculate this given 8nm chips and Ampere really don't come in configurations anywhere close to this, but going by previous designs, this should consume less than 7 watts.
I agree that there won't be exclusives for years to come. Not sure if the new model will replace the old one straight away though, I think that's a strong possibility. Well see.They already have 90m+ of current Switch users, by time this revision come number could easily be 100m+,
so they will need (and want) keep releasing games for that huge install base in any case for years to come..
Best transition would be when this revision completely replace current Switch models on market, same like New 3DS replaced OG 3DS,
but I dont think that will be right away, or releasing huge Nintendo exclusive game (for instance MK9) for this revision only (not gonna happen).
It seems to me that the question is whether the 2021 revision will include lite or not.My main issue is really just the battery life estimate. I get the feeling they'll try to keep it around the Mariko model's official estimates (4.5-9 hours), especially if the revision is an outright replacement for it. Don't really know how it'd effect the specs or performance, but I do believe that's what they're going for.
If Tegra X1 production is over they'll need to release a Lite revision sooner or later, so yeah, I think they'd want to design something that could reasonably maintain the Lite's battery life too.It seems to me that the question is whether the 2021 revision will include lite or not.
The average battery life of the DS and 3DS family is around 3-7h.
Which is in line with the Lite's nominal battery life, and close to the OGswitch's battery life.
If the 2021 revision Lite exists, it should maintain 4.5-9h in the hybrid model as you say
Otherwise it will be difficult to maintain 3-7h in Lite
But if there's no Lite in the 2021 revision, then they might not be so adamant about maintaining battery life.
In this case, the power available will be a little more, which is an advantage in terms of performance, but whether Nintendo will prioritize it depends on their goal.
I am believing that less and less. If it's true we should get some corroboration from other sources.If Tegra X1 production is over they'll need to release a Lite revision sooner or later, so yeah, I think they'd want to design something that could reasonably maintain the Lite's battery life too.
I am believing that less and less. If it's true we should get some corroboration from other sources.
Doesn't necessarily mean thugstas was wrong though. Plans change and insiders often have old info.