Taking a look at Nivida Orin T234: https://gadgetversus.com/graphics-card/nvidia-tegra-orin-t234-gpu-specs/
2048 Cuda cores
GPU base clock 768 MHz
GPU boost clock 1.000 MHz
TDP 50W
4000 Tflops
Samsung 8nm

Nintendo's T239 is rumored to have:
1,536 Cuda cores (75%)
unknown clocks
4-12W range
Samsung 8nm

Now the only way to get that working is either a reduction of cuda cores or very low clocks. I think we could get something around 570MHz docked which would be 4000*0,75*0,57 = 1.7 Tflops at 11W guesstimated. In mobile mode we might be looking at 330 MHz so 4000*0,75*0,33 = 1 Tflop at maybe 5W.

Yea we haven't seen Nintendo talk about specs at reveal since the GameCube. I expect that to continue if those are the specs it would be steam deck with dlss advantage.

To keep backwards compatibility with the OG Switch, the frequency will not go below 307mhz portable and 768mhz docked.

For some games 、gpu runs 460mhz when portable.

I won't be posting here much because I find the discourse is just not very good, but I will take some time to respond to this well thought out post. 

The problem that I think arises is there's really no need for 1536 CUDA cores to get that performance. You could get the same performance from 1024 CUDA cores and just clock them higher (which has no effect on the cost) and the chip would be cheaper and have better yields. Having a massive chip like that for no reason just doesn't make sense, your yields will be worse making production more expensive and you're paying for a more complex chip for no reason. 

If it is 1536 CUDA cores, then I think it is on the 4N/5nm process at TSMC or a smaller node from Samsung, because at 8nm Nintendo is just wasting money by making a massive chip that large (this would be getting into the ball park a similar size to the PS5 chip ... in something that is supposed to be the size of a Switch is crazy, yes you can downclock heavily but it makes no sense for the above mentioned reasons). And the Nvidia leak is pretty clear -- Tegra T239 ... 1536 CUDA cores, there is no other chipset in the documentation that would fit. So if they're paying for 1536 CUDA cores, I think they're probably going to want to get proper use of them. 

But here is Ratchet & Clank Rift Apart, apparently this is supposed to be the high water of what a PS5-only game can do, best looking game on the system, etc. etc. etc. ... yet here it is running just fine on a 1.6 tflop Steam Deck: 

So what is a Steam Deck? Well it's clearly not "just a PS4" or even PS4 Pro, because a PS4 can't run this game at all (or Sony won't bother to try). And the thing with Steam Deck is it could get better performance than even this, because of the nature of the device it's just the PC version of the game, but if you had a dedicated dev team sit down and spend a few months specifically porting this game just for this one spec Steam Deck like console games get, the performance and settings of the game would likely rise a notch beyond even this because of specific optimization. Even if this is "Steam Deck verified" an actual port made just for this hardware would outperform this version. And if you could add DLSS on top of that ... DLSS will make even from 800p resolution a waaaaaay better looking image quality than that. It may not be exactly 4K native, but it will look way, way better than shitty native 800p does, it will look far closer to a 4K game than it does to an 800p game if this chip could do utilize DLSS. 

Also Kopite has gotten node sizes wrong on other occasions, here's one example where he stated a the 50 series Nvidia GPUs wouldn't be 3nm and then later admitting he was wrong.

I was wrong. 🥹 https://t.co/uLyZZN4uiV

— kopite7kimi (@kopite7kimi) May 6, 2023

Really with the Tegra T239 he got only the name correct and everything else was wrong, wrong codename, wrong amount of CUDA cores, wrong CPU from his initial post, wrong architecture (Ampere not Ada) based on the actual Nvidia leaked documentation. 

Lol. Rift Apart runs on the Deck at 720p, no RT and with low settings... and when crossing rifts it takes 10 seconds and isn't instant.

Nobody questions if mobile hardware can run games.... the question always was how comparable.

I don't find the above comparable. It is a significant reduction in fidelity.  Now on a small screen, I'm sure it is fine.  On a large TV....  no thanks.

it looks like a cut down version of t239 so if that happens i win the bet, we don't need to wait for ports lol.

I think what will occur this time that didn't occur with the Switch is a semi-custom APU. One that won't release in other Nvidia products. I've been saying this from the start. So far, the leaks have confirmed this is likely true. T239 features 8 cores in a cluster vs 4, and will feature just one cluster instead of three. In addition, sources suggest that there will be Lovelace features, so it may be a bit of a hybrid between Ampere and Lovelace.

This means that cross comparisons to other Nvidia products is going to be difficult. For example, many people are looking at the T234 and it's 50W for what we should expect, yet the less clusters you have generally the less power you draw. So they may not have to cut back as much as predicted to get the desired power draw.

You can see Lovelace features in one of the Nvidia leaks .... it references "NVENC version 8" on the Tegra T239 ... what is NVENC? It's Nvidia's video playback/recording feature, it's used in Switch games for things like Smash Bros. replays and video capture on the right Joycon. So what's significant about that? Well Version 8 is a Lovelace feature (40 series Nvidia GPUs), regular Nvidia Ampere cards use version 7. So that means the Tegra T239 is using the Lovelace version of this (and who knows what else) whereas other Ampere GPUs don't. 

Also it should be noted like there is no cut down version of the Tegra T239 in any of the Nvidia leaks. It's Tegra T239 and T234 is the full fat Orin for self driving driving cars ... the T239 is already cut down/customized from that. If there was another version of T239 or T234 it should show alongside those two but there isn't.

This guy had a detailed look at one of the Nvidia leaks, note he won't talk about the other one (which confirms 1536 CUDA cores + 12SM for Tegra T239) because the other leak was basically illegal stolen from Nvidia, so just keep that in mind, you can find the exact CUDA core number, he just doesn't want to say it because it was stolen information. 

There is, if you're taking power consumption into account. It scales linearly with frequency but quadratically with voltage, which needs to be higher at higher clocks. A smaller chip at higher clocks would consume more power even if it performs the same.

Also, mind that 5 nm is significantly more expensive than 8 nm. You'll get more dies per wafer in the former, yes, but said wafers are significantly more expensive (I couldn't find the exact figures for Samsung, but TSMC's 10FF process, which is comparable in complexity and size feature to Samsung's 8nm, was ~2.6 times cheaper than N5).

That being said, obviously the bigger node comes with significantly higher power consumption, so lower battery life, more heat, etc. So while I agree it should go for N5, I'm just pointing out that the smaller node would be chosen because of these other considerations, not necessarily SoC cost.

Really? We had our own vgchartz scientist state that it's  cheaper to make 5nm Then 8nm.