The last video about Bayonetta on the Switch made by Digital Foundry is a great example to study. The game runs at the same resolution as the XBox 360 version with the same textures, the difference? 30 to 50 fps on 360 with slowdowns and tearing vs 60 fps stable on the Switch...

Having said that, it shows at least 30% increased power over the 360 version... the only way to play the 360 version with the same fps is by playing it on the One X with back compatibility, which improves the game.

https://www.youtube.com/watch?v=R_M0gX0GE0o  (5:30 minutes)

I have made the point myself about better optimisation of bandwidth but still there seems like a shortfall and if anything your values seem to confirm it. As I said previously the xbox one has a pretty recent GPU architecture but still benefits from that 32MB of ultrafast memory to supplement main memory bandwidth which is in itself much higher anyway than Switch. 

My point about graphics cards was a low end chipset of recent times as used in Nintendo hardware vs high performance of the past which share similar gflops output and I'm sure I've seen PC benchmarks where the older cards actually came out faster because they weren't compromised as much. Your comparison seems more like mid range to old high range. I certainly wasn't thinking about such a comparison I was thinking of high or mid range old gpu vs  low end/mobility of more recent times of similar gflops. In the case of the wii u the Radeon GPU after all the debate about 176 or 352 gflops it was really down to the exceptionally low power draw of the wii u that confirmed it couldn't be 352 gflops and likely a mobility radeon GPU  rather than desktop version. That seemed to fit better with the confirmed spec and a better fit for a small console with limited cooling. 

I don't agree with your comparison as I say but even if we accept it, looks like most game frame rates between the 2 are broadly similar and roughly speaking the newer chipset achieves about a 50% frame rate boost for the same gflops. It doesn't seem true to some other comparisons I've seen where they have fitted older GPU's and newer GPU's to the same motherboard and CPU setup. This for all I know could have been historic frame rates with the older gpu matched to an older motherboard and weaker older CPU setup which would make the comparison utterly pointless but even if we accept this that means the 155 gflops is boosted to about 230 gflops or something putting it squarely into the ps3/360 performance level and no more. It doesn't seem right though. With games like Xenoblade 2 dropping to 368p at times and Doom and very low resolutions too both at 30fps as well. Some Switch games suffer from slower frames rates when docked. This must surely be a CPU or memory bandwidth issue. However where a game has to drop resolution significantly to maintain frame rates in portable mode it feels like its a GPU performance issue. The CPU is the same and memory bandwidth better for lower resolutions.

We are clearly both in agreement about the CPU performance anyway, yes the Tegra at full clocks would be comfortably superior to Xbox 360 in CPU terms but Nintendo didn't use full clocks and only 3 CPU's are used for actual games the other is dedicated to the operating system and background tasks. I still secretly hope Nintendo might release some CPU performance with a later firmware but maybe that is unlikely. Sony did increase clocks with the PSP slightly so it is possible.  I'm unsure if at full clocks it would beat a PS3 with all cells firing and fully optimised code I strongly suspect not considering many consider not even the Xbox One or PS4 achieve that and there are some benchmarks that show it stronger in real world performance than ps4/xbone. 

I tend to look at what developers are achieving with hardware as well as the spec numbers. See what's happening on screen and where the compromises are especially where the compromises are repeated by many different developers. I think you always have to factor in real world evidence. 

I'm not yet convinced of the Tegra's graphic superiority in portable mode with such low clocks vs 360/PS3. To me it looks much more like the 4GB of main memory is the real saviour, room for better graphics, textures without the constant need to swop/stream data of 360 and PS3. I feel this is the huge strength of the system compared to 360/PS3. I honestly wonder if the PS3 and 360 had that available memory if it couldn't both speed up and allow more sophisticated game engines. PS3 especially was a nightmare. I remember running Fallout 3 with DLC at 480i on my PS3 just because it helped reduce slowdown which was awful in Operation Anchorage and the other DLC. Capcom's pressure on Nintendo to boost memory from 2GB to 4GB on Switch I feel was a great move I just wish they asked for higher CPU clocks at the same time. You have made a case for the Switch graphic hardware but despite being a big Nvidia fan myself I'm not convinced by it at this point for portable performance. Yes great for docked and easily superior to 360/PS3 no question.

The Xbox One has more bandwidth by default than the Switch.
The Xbox One has eSRAM to lend a hand to lend a hand for things like bandwidth intensive render targets.

But in terms of architectural efficiency... Maxwell beats Graphics Core Next.

Graphics Core Next does come into it's own once you start to leverage Asynchronous Compute however, but when you start to compare the Switch against the XBox 360, the Switch starts to look more favorable.

My comparison was to prove the point that GPU's of different architectures cannot have their performance determined by just using flops.

And yet was most likely a Very Long Instruction Word (VLIW) 5-way derived Radeon GPU architecture, which had all the accompanying advantages over the R580 derived GPU in the Xbox 360.

Doom is doing some good effects though. Unless you disagree? Then you need to point me towards an Xbox 360 game that has hardware accelerated particles which have hardware accelerated lighting and shadowing.

The Switch is doing things with Doom that the Xbox 360 simply can't.

Could be a combination of all of the above. Or none at all.
Until you look at the performance profiling for the engine running on the actual hardware, the only thing you can make is accusations, which is not something I can adhere to.

Well, the CPU is more efficient even at those lower clocks, that isn't really up for debate.
Even having just 3-cores for gaming is fine.

But the clocks is what holds things back. I'll probably do some more research on this at a later date and so how both chips compare on this front at some point, just haven't had the time recently.

That would be nice. But probably unlikely. Battery life is fairly short on the Switch, higher CPU clocks would erode that farther.

In some tasks it would be faster than the Cell... Especially with Integers.

The Cell was very specific where it could really show it's capabilities... And games/game engines have tons of different demands.

The best Analogy would be a car race... Where the 8-core Jaguar/4-core Tegra can cruise down to the finishing line at 100 miles per hour.
The Cell however will do 50 miles per hour for the bulk of the race... But when the road conditions are just right, it can peak to 150 miles per hour.

The Cell has the higher potential speed, but it will loose the race every time as it's not able to maintain that speed under all conditions.

Real world evidence is what I live by.

However, keep in mind that the Switch hasn't been on the market for a year yet... We have yet to see the best looking and most demanding games on the platform... The Xbox 360, Playstation 3 and Wii U are done. Finished. Tapped out.

Maxwell beats Graphics Core Next in terms of efficiency every day of the week. (Outside of Asynchronous Compute that is.)
Graphics Core Next obliterates VLIW4 in terms of efficiency every day of the week.
VLIW4 was a substantial improvement over VLIW5 in terms of efficiency as it was a more balanced layout.

I have already shown an example of a Graphics Core Next GPU out-performing a VLIW5 GPU with almost a teraflop extra of single precision capability, that same kind of performance divide would exist between a VLIW5 GPU and the R580 as well.

And the Xbox 360/Playstation 3 GPU predates Graphics Core Next, VLIW 4 and in some aspects, VLIW5.
The newer the GPU, the more work it can do per flop, that's the crux of it.

That's a massive part of it.
7th gen consoles were extremely memory limited.

It could.

Net Immerse/Gamebryo/Creation Engine was just never a good fit for the Playstation 3.

Even just a 20% higher clocked CPU would have helped significantly.

I'm not an nVidia ardent fan. I primarily sit in the AMD camp as they typically provide better price/performance.

Here we can see the Microsoft Surface Pro beat the Geforce 7800GT and 7900GS.
https://www.anandtech.com/show/6877/the-great-equalizer-part-3/3

The Playstation 3 has a 24:8:24:8 core layout running at 550mhz with 22.4GB/s of bandwidth.
Which means it should fall below the Geforce 7800GTX in terms of capability. (As it has more ROPS and bandwidth.)

Tegra K1 is beating the Microsoft Surface Pro by 50% or more.
https://www.anandtech.com/show/8296/the-nvidia-shield-tablet-review/5

The Tegra X1 is beating the K1 by another 60% - 80% and in a few instances by 400%.
https://www.anandtech.com/show/9289/the-nvidia-shield-android-tv-review/4

So we can account for the fact that the Switch is just 30% of the clockspeed of the X1 and still come out on top.

Ergo. Switch beats the Playstation 3 and Xbox 360.

Something is wrong with quoting on my phone. I own Bayonetta 1 on Xbox 360 (which means I have it for Xbox One X), Wii U, and Switch.

I'm personally disappointed with the image quality of the Switch version of Bayonetta 1 on my 4K TV. Unlike most of my other Switch games, it looked really muddy. I also fired up Bayonetta 2 and it looks really sharp and vibrant.

I really can't use Bayonetta 1 as an example of Switch shortcomings because Bayo 2 looks so much better! And, like Digital Foundry said, it may actually be the best console version available. I'll compare the original Xbox 360 version to the the Switch version and maybe even the Xbox One X version myself, tonight. I feel like there's a lack of aliasing on the Switch but I could just be remembering things wrong.

Personally, optional portability makes it the version to own, anyway.

The Bayo ports to Switch are very barebones conversions; they're not bad per se, they just don't do much to take advantage of the hardware beyond just the almost automatic framerate boost that comes from a faster GPU. I mean, it's cool that both games run smoother than 360/Wii U in portable mode, but the docked mode feels like an afterthought.

I knew there was a reason I was subscribed to this guy!

https://youtu.be/DCfR54Wq4x8

This guy thinks Wiiu and switch have around the same Gflops which is laughable.

Few inaccuracies in the video. But he mimics my own sentiments on the GPU front.