LandoTheLost
Cappy
- Pronouns
- He/him
Look at the glass 60% fullOoof, less than double the handheld speed. That's pretty conservative.
Look at the glass 60% fullOoof, less than double the handheld speed. That's pretty conservative.
Just seems kind of odd to me when we take into account the average amount of bandwidth used by Ampere per Tflop. Like, at 1007.3, pushing around 3.095 TFlops, which would saturate at around 77.38 GB/s. With LPDDR5X-7500, that leave about 42.62GB/s remaining. Seems like a lot left over. If it were using LPDDR5-6400, it would instead be 25.02GB/s left over.Huh. That's still a pretty nice number but lower than double the handheld speed?
Then they'll be unnecessarily wasting power for the first few years. DLSS is more efficient for creating an image of a given resolution than not, otherwise it wouldn't be very useful.What if Switch 2 uses DLSS only in dock play for now? And in handled mode will be unlocked down the line in a couple of years?
That's a nice perspective for people who give a damn about handheld mode.Look at the glass 60% full
I'll take it. It's not the optimistic 3,4 Tflops but isn't that far away What I find curious is that the docked is below double of the handheld speed. Could the 560MHz be a higher portable clock like the 460MHz on Switch?
The Switch 1 does.Would be a bit strange for nintendo to have such a big gap between handheld and docked performance. Would've been cool though.
Great point.Actually the difference in ratio of pixels in 1440p to 1080p, (1440/1080)^2=1.77 aligns rather well with the ratio 1007.3/561=1.79![]()
Switch 2 has 3 as many shading units as Steam Deck. It's expected for it to hit such a number with a third or so of the clock.the steam deck produces 1.6teraflops of raw power at 1-1.6ghz is that correct? if that’s the case the switch 2 doing that at half/less than half the clock speed is incredible no?
DLSS isn't a TFLOPS multiplierOk if that's pre-DLSS, 1.7TF is definitely good enough for a handheld.
because we've seen that Switch's handheld mode was too deficient, so Nintendo wants to avoid that againWait, this is even more confusing.
Why would you have a pretty high handheld clock, but then go conservative for docked? You'd think they'd have more wiggle room since they don't have to worry about battery/
Given all that I think it's feasible to expect to be mainly "1440p" machine~1.76TF portable, ~3TF docked sounds very. Interesting. However, I do think it makes a lot of sense. A raised floor.
Even at double the speed outright, the system is still doing a 4X pixel output jump from one mode to the other. This device was always going to have to lean on upscaling - if these numbers are true, and importantly, if they are final*, then indeed, TV Mode is going to lean on upscaling a lot. I think this is reasonable, as the device is already designed around upscaling. DLSS Performance Mode targeting 4K, with internal dynamic resolution would have an input of 1080p or less, which sounds like a reasonable render target for a modern game at 3TF. Games that push beyond that can use Ultra Performance Mode (which I think we will see a lot of), or lower their output resolution (perhaps especially in titles that prioritise high, stable performance).
I would say that it would be reasonable for games to be 560p in Handheld Mode and 720p in TV Mode, using Performance Mode in portable modes and Ultra Performance Mode in TV Mode, for 1080p and 4K output respectively.
Games that push the system will be able to do something as drastic as a dynamic internal resolution Ultra Performance Mode targeting 1080p for portable modes, with an internal resolution cap of 360p. I'm sure games will even have the option of 720p target Ultra Performance Mode for portable mode, an internal resolution of 240p, if that's the only way to get the game running. 360p to 1080p in TV Mode.
I do not think we will see any games dip below a 720p output target, a huge improvement over the sub-HD and even sub-SD resolutions some Nintendo Switch titles have at output.
There's also the consideration that developers will have access to more than just DLSS, and can pick and choose to get their desired performance, or when stack a spacial and temporal upscaler if they choose to do so, for instance if 1440p with DLSS then to 4K with FSR(1) provides good image quality at a lower performance cost than 4K DLSS, it may be used.
Others still, as I've said before, I think will likely settle for a sub-UHD output resolution, and trust in the built-in upscaler which I hope is greatly improved.
As before, 1440p seems to be the intended target resolution for the console as 900p was for OG Switch.Wait, so these really are the clock speeds that we were waiting for the last ~4000 pages? Feels a bit anticlimactic - I thought the Cat was messing with us
Actually the difference in ratio of pixels in 1440p to 1080p, (1440/1080)^2=1.77 aligns rather well with the ratio 1007.3/561=1.79![]()
It's not like they can make the gap so high with a typical GPU f/v curve. Too high and your device is too loud and/or too hotThat's a nice perspective for people who give a damn about handheld mode.
These are GPU clock speeds, DLSS does not rely entirely on the GPU for upscaling.Ok if that's pre-DLSS, 1.7TF is definitely good enough for a handheld.
I would absolutely be okay with a thicker but overall smaller device.I hope miniaturisation is at the forefront of the next generation Lite, I'd welcome the return of a clamshell design, too, though with touch I understand why they might not want to do that.
Do you think people would be willing to accept a Lite that's actually THICKER than the standard console in order to achieve adequate battery life in an overall smaller footprint? Say the size of the screen alone on the new device?
what does that mean for portable efficiency if we were to compare the two? because the steam deck chugs the batterySwitch 2 has 3 as many shading units as Steam Deck. It's expected for it to hit such a number with a third or so of the clock.
Biggest question there would be why? I see basically no benefits to keeping DLSS docked only.What if Switch 2 uses DLSS only in dock play for now? And in handled mode will be unlocked down the line in a couple of years?
It's more likeJust some quotients:
Switch 1: 768 / 393 = 1.954
Switch 2: (allegedly) 1007.3 / 561 = 1.796
I think you mistook that 393 for a clock frequency when that number is GFlops for Switch at 768Mhz.Just some quotients:
Switch 1: 768 / 393 = 1.954
Switch 2: (allegedly) 1007.3 / 561 = 1.796
(I was hoping for something closer to 2.25 or 2.5, but if that how it is then that's how it is)
Significantly higher efficiency. More cores at lower clocks like the Switch 2 is more efficient than fewer cores at higher clocks like the Steam Deck.what does that mean for portable efficiency if we were to compare the two? because the steam deck chugs the battery
Nobody really knows. It will likely achieve this with a combination of being more efficient than Orin (how much? don't know), a bigger battery and possibly better node.what does that mean for portable efficiency if we were to compare the two? because the steam deck chugs the battery
I will ABSOLUTELY take 1.7TFLOPs and 3.1 docked.
OK now I want these numbers confirmed
I’m a complete ignoramus, but based on how everyone is interpreting these, this is good enough.
not gonna complain about a huge leap compared to the Switch.I will ABSOLUTELY take 1.7TFLOPs and 3.1 docked.
I can see the initial model having atrocious battery life if the portable mode performance is that high, so perhaps Nintendo wanted to counteract that by making it easier to charge while docked while also not wanting one mode to outshine the otherWait, this is even more confusing.
Why would you have a pretty high handheld clock, but then go conservative for docked? You'd think they'd have more wiggle room since they don't have to worry about battery/
It seems to be pretty damn good, 1.7 / 3.1 TFLOPs of raw GPU compute from a thin portable machine before DLSS is in the picture.I’m a complete ignoramus, but based on how everyone is interpreting these, this is good enough.
Unless Nintendo starts allowing docked only games, I think one should really care about handheld mode performance, since it will determine the system’s baseline and what may or may not come to the system.That's a nice perspective for people who give a damn about handheld mode.
My only theory is heat.Im surprised that the handheld profile is that decent but the docked being 1007.3 when there is no battery to worry about ... thats odd!
I think this is the right perspective.Handheld performance will matter more in the long run for the overall performance of the system as it’s an important floor/baseline compared to the ceiling of docked mode. There is room for Nintendo to overclock the system in docked if needed, not as much room in handheld mode.
Very good point. Could be a coincidence. Aiming for 1080p on the handheld and aiming for 1440p docked upscaled to 4k would be great.Wait, so these really are the clock speeds that we were waiting for the last ~4000 pages? Feels a bit anticlimactic - I thought the Cat was messing with us
Actually the difference in ratio of pixels in 1440p to 1080p, (1440/1080)^2=1.77 aligns rather well with the ratio 1007.3/561=1.79![]()
From the clocks that were just given to us by a reliable dataminer.Where does the 1.7 tflops and 3.1 tflops come from? Why wouldn’t it be 3.4 in dock?
I believe 1440p will be the target for most first party titles (after DLSS), and for that we'll need a pixel increase of 1.77x (if you keep the same DLSS profile between handheld and docked, which is the best decision IMO)
So I'm expecting the docked clock to be at least 77% higher than handheld.