JustBeingReal said:
No company measures them in the same way, so you using them like they are is completely disingenuous. No AMD are not aiming for 70GFlops per watt, going by the way AMD measures FLOPs, at 28nm's they've achieved 819Flops over 35watts, that's taking everything together into account across a Carrizo APU/SOC, which means they've achieved 23.4GFlops per watt, if that architecture can be shrunk to 14nm it would achieve 46.8GFlops per watt. Perhaps they can optimize their design to gain more performance, but that would take time, 2016 probably isn't going to be when they can achieve 70GFlops per watt at 14nm, maybe they'll get close to 60GFlops, but this is based on how AMD measures floating point performance. GFlops are not a standard for measuring total processor performance, you using PowerVR, Nvidia's Tegra or anyone else's tech as a comparison, with GFlops as a metric makes absolutely no sense, ratings are not comparable.
Flops are a marketing buzzterm, a fake figure to make it seem like you have X amount performance, but it's not accurate or a figure that is commonly always used in exactly the same way, nor does it cover total processor performance. Flops don't tell you Instructions Per Second (IPS), cache size, cache amount, Bandwidth, latency between components inside a chip or between chips. |
Let me ask you then, what's the maximum performance you think they can get in 4-5 watt power evelope for a handheld launching fall 2016?
If it's not enough to at least be able to run Wii U ports ... I think Nintendo's whole conceptualization of this unified concept will probably have been a bad idea, but that's kind of becoming par for the course with their hardware designs of late.







