The PS5 is about 4x more powerful in CPU/GPU than PS4 so if we get a similar jump in power with the PS6, a portable version of it simply won't be possible. Only possible if the PS6 isn't designed to replace the PS5, which this topic is all about. If PS5 remains the base console for the entirety of next gen, all the PS6P has to do is match the base PS5's CPU, and have a 1/2 strength GPU and use FSR4 (or better upscaling tech) to make up the deficit.  

Just look at the dude...  got the face of a console killer...

Left a mess of green goo...  now hungers for red splatoon...

Has tasted freedom...  will not give up that which he has tasted...

Spreads managed democracy throughout the galaxy... by dispensing peace with the ultimate weaponry...

Yeah, about that 4x for PS4 to PS5...more like 6-8x.

What you could expect is that PS6 will be 2-3x in raster, around 4-5x in hybrid RT workloads (that depend both on raster and RT) and up to 8-10x in pure RT tasks over PS5.

PS6P is more than possible - the thing about the next gen is that raster workloads are becoming less important, so that modest jump in raster performance is what makes PS6P viable option - it will have to render only for 1080p, instead of 4K+ of PS6, and even with its cutdown capabilities it will be more potent in RT than PS5.

Of course, for the umpteenth time, this all depends on UDNA/RDNA5 delivering on rumored uplift of 2xRT over RDNA4, which already slaps the crap out of RDNA2 in PS5.

It's not the GPU side that's the issue, It's the CPU. There is no AI machine learning you can use to make a weak CPU perform better. Look at Switch 2, all it's issues are CPU and bandwidth related. It has a PS4 level CPU and it's battery life is still quite bad relative to Switch 1, so imagine how hard it is going to be to have a PS5 strength CPU in there with reasonable battery life.  PS6P's APU is going to have to be very CPU focused, leaving very little for the GPU. Seriously, the absolute best they can achieve with handheld is just matching PS5 performance and I think running PS5 games natively is essential for the success of the handheld. 

Also, even though the tech in PS5 is very average by todays standards, due to worsening economic conditions Sony is already struggling to maintain the price and low price points are essential for consoles because they are meant to sell in the tens of millions. So, how would it be possible to have a 2-3x increase in raster and 4/5x RT and still keep it at a mass market price? When you consider the extra cost for the screen for the PS6P, I don't think possible, and because MS is no longer trying to compete with Sony in consoles, they doesn't have to worry about having good specs that compete with the competition anymore. I still believe that the high GPU clocks and the use of liquid metal in the PS5 is because Sony saw the specs of the XBSX and thought that they needed to increase the power of the PS5 last minute. That's the kind of thing that competition can make you do. 

All this means that Sony can just re-release the PS5 and call it the PS6. A PS5 with ML for FSR4/5/6 will still be a major upgrade over PS5 with no ML. It also means a PS6P and a PS6 vita TV is possible. You can also have a PS6 Pro for those that want 60fps 4k.  

PS6 will use Zen6 CPU - even current Zen5 is miles above Zen2 that is in PS5. I'm not expecting PS6P to run at Switch 2 low 10W wattage, I expect it more toward PC handhelds of 20W-30W.

As for GPU - there are already higher mid-tier PC GPUs on current architectures that do what PS6 GPU should do. And we're at least 2 years away from PS6 which will use UDNA.

As for competition, PS6 will be not competing with XBOX console. It will be competing with XBOX platform, STEAM and, most likely, GeForce NOW.

And last, but not least - mass market price is not what it used to be, as Switch 2 has so obviously shown.

PS5 is not only old tech, but very, very limited (to be polite) when it comes to what next gen will be focused on, and that is RT.