By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Soundwave on 12 January 2024
Edit Thread
Pemalite said:
Jules98 said:

New specs leak just dropped:

Not a leak. It's an unsubstantiated rumor. We need to start knowing the difference, otherwise people start to peddle rumor as fact.

Jules98 said:
  • TSMC N4 node process (4 nanometre?)

TSMC N4 node is advertising. It's NOT 4 nanometer.

TSMC N4 is actually based on it's 5nm technology which ironically has a bigger gate pitch and interconnect pitch than IRDS 7nm definition, take that as you will.
But the scaling just isn't happening at the moment.

Jules98 said:
  • 8-core A78C CPU, clock rates unknown, don't know what's meant by GA10F (this could be the GPU line)

GA10F does refer to a Geforce Ampere class GPU.

Jules98 said:
  • 12 stream multiprocessor GPU, performance ranging from 3.5 to 4.5 TFLOPs docked and 1.7 to 2.0 TFLOPs handheld

Teraflops is meaningless. It's theoretical, not real-world.

Jules98 said:
  • 100GB/s memory bandwidth docked and 88GB/s handheld

720P again confirmed? ;)

Jules98 said:

  • Memory cache specifics uncertain, Tegra GPU cores may be able to access CPU cache

This has been a Tegra feature for some time.

Cache coherency is however, only 1-directional, Tegra X1 in the Switch should be able to do this.

https://docs.nvidia.com/cuda/cuda-for-tegra-appnote/index.html

Jules98 said:

  • Display is 8" screen with 1080p and 60hz refresh rate

Good luck getting 1080P happening with 100GB/s of bandwidth.
You will not get the fillrate necessary to drive that.

Jules98 said:

  • Cartridge specifics unknown, but 3D-NAND may provide a cost-effective way to significantly increase storage

And will not last long.

Bitflip is a big issue with NAND, you loose data extremely quickly.
3D NAND would exacerbate that issue.

ROM is the way to go.

Chrkeller said:

Seems reasonable.  I've been expecting the memory bandwidth to be a bottleneck.  100 gb/s for a portable is fair but very slow compared to home consoles.  Ps4 is 176, series s is 225, ps5 is 500 and a good gpu is 700.

As a reference point, steam deck is 88 gb/s.  

I'm sticking with my ps4 like visuals given the bottleneck.  

Steamdeck gets away with it because it's only running 720P levels of resolution.

Keep in mind that Delta Colour Compression is a technology that the Playstation 4 lacks, so that can give Tegra an extra 60% bandwidth boost if the patterns fit with DCC.

Playstation 4 level of capability but with a few extra tricks (Ray Tracing) is the most logical scenario here.

JRPGfan said:

https://www.youtube.com/watch?v=4Arp3H_mgR8



Iphone 15 pro Max, looks like it would need like 4 times its current performance, to match a base PS4 (in this game).
However.... that is still insane.... its a PHONE..... its not ment to be able to match a console, that again, draws 100's of watts of power.

Keep in mind is also running CUSTOM silicon and CUSTOM software to MAXIMISE performance.

You do NOT get that same privilege on other platforms.

Soundwave said:

And again, you have to understand this is not even the max performance, these games are not optimized properly for hardware, with consoles you get a dev team to actually sit down and tailor make a version of a game specifically to one hardware spec, these are just PC games with settings sliders moved around, if a a dedicated dev team actually sat down and worked on a port for like 6-8 months, fine tuning every area of the game just for one piece of hardware, you'd have better performance than this. 

You do realise that is exactly what console developers so? Essentially just move a slider?
Digital Foundry does it all the time and "adjusts" PC sliders to get an identical visual representation of the console release and gets those games running on equivalent PC hardware.


RedKingXIII said:

Personally, I'm expecting it to be a lot more powerful than a PS4, just like the Switch was way ahead of the PS3. It won't be as powerful as the PS5, but the power gap between the Switch 2 and the PS5 will be smaller than the power gap between the Switch 1 and the PS4.

It doesn't need to match PS4 specs to beat it.

"weaker" hardware on paper today is leagues ahead of "faster" hardware from 10+ years ago, efficiency does actually improve, which is why the Switch is able to beat a Playstation 3, WiiU and Xbox 360, nVidia Maxwell is just a significantly more efficient chip despite the "flops" not representing that visual leap.

Why would you ever run any game at native resolution on Switch 2 to begin with? There's no point when you have DLSS, undocked really never needs to go above 540p as far as I'm concerned. Yes native looks slightly better, but not good enough that it's worth forcing the system to render 4x the pixels. On top of that DLSS gives you basically a "free" form of anti-aliasing. Running at native + wasting resources on top of that for AA is just brain dead, in fact I would postulate that DLSS implementation is the automatic default for Switch 2 dev kits, the system will be designed to run with that on. 

Especially on a freaking small 7-8 inch-ish display, the regular joe, even most "game enthusiast joes" are not really going to know or care that their game is actually only rendering from 540p, shit I think you could go even lower than that. 

Yes you can move sliders around on PC games that have a performance overhead to "match" lower console settings, that doesn't really have anything to do with what I'm saying. I'm saying if the Steam Deck version of Ratchet & Clank runs at 30-40 fps, if Insomniac sat down with a team of 20-30 people who worked on the port for 6-7 months JUST for that one hardware, do I think they could get that up to a solid locked 40 fps and/or maybe even bump the settings from Low to Medium 30 fps ... yes, I do. Optimization does matter.

Last edited by Soundwave - on 12 January 2024