Qualcomm latest GPU clocks at over 2tf and has raytracing

People said the same thing when the first Ada rumors showed up.
More power efficient and near double the power of a 3090?
No way, there is no technology that efficient.

Fast forward a year and Ada basically delivered everything that was rumored.
Ada is way way more power efficient than Ampere and legit delivers near double the performance.

4TF docked doesnt even sound that extreme for a 2024 Nintendo console.
Not sure why people think it should be closer to 1.2TF.
Tegra X1 (T210) - (15W) (20nm) (Maxwell) - 0.5TF - 2015
Tegra X1+ (T214) - (15W) (16nm) (Maxwell) - 0.65TF - 2019
Tegra Xavier (NX) - (15W) (12nm) (Volta) - 0.85TF - 2020
Tegra Orin Nano 8gb - (15W) (8nm) (Ampere) - 1.28TF - 2023

Because Nvidia's mobile chips at the same TDP have advanced from 20nm to 8nm and from Maxwell to Ampere, 3 graphics generations, and they've gone from 0.5TF to 1.28TF in 8 years. 4TF for a hybrid console is not happening in the next 2 years.

Though I'm not saying the hypothetical Switch 2 wouldn't be more powerful than 1.28TF, but even if they defy expectations and use a state-of-the-art chip with Ada architecture they are not suddenly getting an nvidia mobile chip that runs at a reasonable TDP, has a reasonable price and and jumps from 1.28TF to 4TF.

I expect the Switch 2 to use a customised Tegra Orin chip which may or may not be more powerful than 1.28TF. Most likely the one that's heavily rumoured the T239.

Let's not get too bogged down by the Series S comparison. Let's be clear even a 2.2TF Switch 2 would be an equivalent GPU jump to PS4 -> PS5 or 360 -> XBO. 0.4TF to 4TF would be a crazy leap.
 
Last edited:
Tegra X1 (T210) - (15W) (20nm) (Maxwell) - 0.5TF - 2015
Tegra X1+ (T214) - (15W) (16nm) (Maxwell) - 0.65TF - 2019
Tegra Xavier (NX) - (15W) (12nm) (Volta) - 0.85TF - 2020
Tegra Orin Nano 8gb - (15W) (8nm) (Ampere) - 1.28TF - 2023

Because Nvidia's mobile chips at the same TDP have advanced from 20nm to 8nm and from Maxwell to Ampere, 3 graphics generations, and they've gone from 0.5TF to 1.28TF in 8 years. 4TF for a hybrid console is not happening in the next 2 years.

Though I'm not saying the hypothetical Switch 2 wouldn't be more powerful than 1.28TF, but even if they defy expectations and use a state-of-the-art chip with Ada architecture they are not suddenly getting an nvidia mobile chip that runs at a reasonable TDP, has a reasonable price and and jumps from 1.28TF to 4TF.

I expect the Switch 2 to use a customised Tegra Orin chip which may or may not be more powerful than 1.28TF. Most likely the one that's heavily rumoured the T239.

Let's not get too bogged down by the Series S comparison. Let's be clear even a 2.2TF Switch 2 would be an equivalent GPU jump to PS4 -> PS5. 0.4TF to 4TF would be a crazy leap.

Assuming a christmas 2024 launch, manufacturing would happen in early 2024. By then Orin can benefit of a die shrink and the NX16gb part would fit the thermal envelope of a hypothetical siwtch 2 at 15w, which means 1.88tf.

Admitedly by then there will be additional nodes available, but we're talking about the more refined and high yielding process versions, since nintendo is unlikely to be interested in the high risk-high cost nodes.
 
The Tegra X1 is 0.5TF in FP32, the underclock for the switch docked brings it down to 0.4TF. You are using FP16 half-precision figures for the X1.

The Adreno 530 clocked at 624mhz in the Snapdragon 820 was also 0.5TF of FP32 at the end of 2015.

The Tegra X1 ran at a TDP of 15w to get those figures, the Snapdragon 820 ran at 11w TDP.

Just goes to show you how much of a used car salesman Nvidia really is. I remember they plastered every headline about X1 being the first 1TF mobile chip. Thought they invented their fake TF numbers with Ampere, but guess they really started it with Tegra.
 
Last edited:
A Steam Deck is ~1.6TF and it's a year old at this point and doesn't have a dock that ups its clock speeds. It wouldn't surprise me if in a year or two nvidia can produce something with better handheld performance and certainly docked performance for around the same cost when you have to imagine the volumes would be much larger than a more niche product like Deck
The Switch Dock doesn't increase the speed of the Switch, there's a handheld mode that decreases it. Deck runs full tilt whether plugged in or not. Depending on the game and settings Deck can pull up to 27-28W total system power consumption. The Switch pulls like 11W docked and 9W in handheld.
 
The Switch Dock doesn't increase the speed of the Switch, there's a handheld mode that decreases it. Deck runs full tilt whether plugged in or not. Depending on the game and settings Deck can pull up to 27-28W total system power consumption. The Switch pulls like 11W docked and 9W in handheld.
I get what you're saying I know the dock doesn't increase the speed but there's more headroom as shown by stable over clocks on the Switch, my assumption was the Deck could possibly have headroom if they had design it to have a docked mode from the beginning. But maybe not because the fan might have to spool up so fast it would fly into the ceiling.
 
I get what you're saying I know the dock doesn't increase the speed but there's more headroom as shown by stable over clocks on the Switch, my assumption was the Deck could possibly have headroom if they had design it to have a docked mode from the beginning. But maybe not because the fan might have to spool up so fast it would fly into the ceiling.
There might have been some additional headroom since I think the Shield runs a Tegra X1 at higher speeds, but Nintendo is conservative and like you said it might have been a fan/thermal issue dumping all that heat.

I'd expect something along the same lines with the Switch 2. Maybe 12W docked and 9W handheld, so whatever chipset it's getting it won't be hitting 27W like the Deck. DLSS 2.x is going to work wonders for upscaling to big screen TVs, though.
 
Last edited:
Just goes to show you have far ahead of everyone else Nvidia really is. The Tegra X1 is nearing a decade old on an ancient process node and surpassed 1TF at it's default clocks (not underclocked like Switch). Took 8 years for their competitor's to Moores Law it. ARM GPU designers make AMD look competent.
FYI, Switch's Tegra X1's 1 TFLOPS is FP16. The Maxwell GPU in Tegra X1 includes a double-rate FP16 feature.

RDNA, Vega GCN, and PS4 Pro have double-rate FP16 features.
 
Last edited:
Windows Central- Surface Duo 3 In the Works

Will most likely be ARM based. I really think we need a third contender in mobile space and might as well make it Windows 11 for ARM. ARM based chips are getting beefy enough where you can run a full-fledged OS.
A single foldable screen is great news! I liked my Duo but having two seperate screens was cumbersome so I returned it. Made no sense to have two screens when a single screen can be split via software and still have a single large screen for media and web browsing.

I've since moved on to the Galaxy Fold 4 and love it.
 
Last edited:
Nvidias Orin is already ~5TFs.
Assuming the Super Switch is sticking with Nvidia then whatever SoC they put in that should easily be more efficient than Orin which is Ampere based and match a PS4 Pro.
Now I know some may be disappointed but in my eyes if a Super Switch had PS4 Pro levels of power and access to DLSS........id be stocked as shit.

I'm scratching my head with you thinking the Switch 2 would match a Ps4 Pro in GPU power…or that anyone would be disappointed in a PS4 Pro level GPU outputting to what's likely to be a 1080p display.
 
Top Bottom