I found an interesting post at Gonintendo in reference to graphics, not sure if this guy is full of you know what but it sounds like he could be right.
By Squeak
If Nintendo (and the engineers at IBM and ATI) are clever and cut just the right corners and cheat in just the right places, it *should* be possible for Wii to output comparable graphics to at least 360.
Here’s why (among other things): One of the aspects of graphics, where the diminishing returns of increasing power, is most apparent is with regards to resolution.
320Ă—240 looks a lot better than the legobrick resolution of the Atari 2600. 480i/p looks quite a bit better than 320Ă—240, but not quite the aforementioned leap.
720p without AA looks a *little* better than 480p.
But, 480p with good AA and all effects on, actually looks better than a game where a lot of the power has gone towards achieving the high HD resolution.
720p takes up 3x the bandwidth, 3x the memory and 3x the fillrate of 480p, but it doesn’t look 3x better at all.
Even if Hollywood (the Wii VPU) is only half the overall speed of xenos, it would still be faster at filling it’s smaller resolution with similar quality pixels.
The CPU is a different story.
The CPUs major tasks are
- Keeping track of the gameworld
- Transforming geometry
- Physics
- AI
First of, there’s seemingly an awful long way from a singlecore 700-1000Mhz processor to a triple core 3,2Ghz processor.
For general purpose stuff though, access to memory as fast (low latency) as 1T SRAM, coupled with a healthy cache (at least 256kb) means that the CPU can be feed data continuously without having to stall all the time, waiting for the relevant “random” piece of data.
An important factor is also how much helper logic (OoOe and branch prediction etc.) the Broadway has, something which ms and sony has chosen not to have so much of in their CPUs.
Geometry transformation is of course done to a large degree on the geometry engine of the VPU, which can be fixed hardware and with very high polycount, or flexible but with less polygons per second.
ms and Sony chose the latter.
If Nintendo has a fixed geometry engine (like in the GC) it should be able to throw a lot of geometry around coupled with a CPU that’s good at floating point calc, for the more demanding geometry tasks.
In other words you won’t be getting pervasively destructible environments and models mapped with particles on the Wii, but comparable geometry complexity overall.
The CPU is also used for physics, something which can be very important for gameplay.
It would really be a clever move, to have a small part of either the Hollywood or Broadway dedicated to a physics processor, like it has been hinted by some developers in interviews.
Like geometry, physics calculations has characteristics that are very suited for implementing, at least partly, in fixed hardware, making it run cool and fast but of course losing some of the flexibility.
The last point, AI, isn’t suited for hardware implementation, but in the usual way of implementing it, it depends entirely on general-purpose power. What’s more, AI is one of the smallest posts on the CPU time allocation table.
Branching, scripted behaviour and not “genuine” AI, is still by far the most common way to do AI.
And then lastly to memory:
How much memory is really needed for impressive visuals? Well, that depends entirely on what you mean by impressive. But let me just point out that half of Wiis supposed memory size of ~100Mb is enough for 400 512Ă—512 textures. More than I have ever seen any console game use in a single level!
If the drive is fast enough (as fast or faster than the GC one) it should be a relatively easy to use the DVD as a kind of very slow virtual memory.
Topic: Next Generation steps not leaps. My take on next generation graphical horsepower (Read 32406 times)
