Console builders should be looking for CPU cores that can grant a certain amount of performance continuously, not chips that raise the clocks on demand.
I assume it's because Nintendo already ducktaped 2 cubes together, so they can't do that again. Also, 8 smaller cubes make 1 larger cube...
Mass production start in 1Q 2016, so no zen, very probably no hbm, maybe gddr5x
You could do a square shape with 9 cubes. 9 would make an odd cube, tho. (See what I did there? Heh.)Surely that would be 9 smaller cubes?
If NX is home console, they will probably aim to reach ~1.5Tflops so that they could play with big boys. And we all know that Nintendo thinks that the easiest way to achieve that will be by ducktaping 3 or 4 WiiUs together.
4 WiiUs together would get you about 700 Gflops, less than half the goal of 1.5 Tflops.
Indeed, you'd need like 9 WiiUs to get there.
I thought... the wii u was around 350 for the gpu.. Did I miss something?
Both Zen and HBM are finalized. If Nintendo wants to use them, they can.
Sounds like you did 1 flops/clock instead of 2 - 320 x 2 x 550MHz = 352gflops, 176 is half that.
How much "finalized" equals to ready?
Amd will launch the first zen cpu (not apu) in 4Q 2016, Nintendo is expected to start production in 1Q 2016 with what is really ready and tested at that moment.
With GDDR5x the HBM2 looks less appealing. In the wild hipotesis that they are aiming for a decent home console, the first + mb traces can be more convenient than hbm + substrate.
Let's call it a USP rather than a marketing advantage. A games console is bought chiefly for its ability to play games and not it's size, with form factor being one of a number of lower priority consumer considerations.If Nintendo has gone for an HBM low power APU with no optical drive, then this is going to be one tiny little box which will give it a marketing advantage over the other two boxes.
Wiiu supposedly only has 160 shaders