That works when all things are equal and AMD and Nvidia floating point operations are not equal and flops/cycle are not equal. This is as dumb as the people who tried to compare Motorola and Intel CPUs based on clock speed.
So back in the day you said that a Motorola MHz isn't the same as an Intel MHz? That a 30 MHz Motorola is clocked faster than a 30 MHz Intel??
This isn't about comparing the platforms as being able to do equal work with the same amount of resources. It's about clarifying what a FLOP is, what that number means when comparing GPUs, and how to best communicate that.
Two cameras described as 8 megapixels - we don't say Nikor megapixels != Canon megapixels; we say they have the same megapixel count but resolving power, contrast, and image quality can be different, such that one can produce a better result with the same objective measure of one aspect.
Two cars reported as 400 BHP - we don't say Ford horsepower != Lotus horsepower; we say the Lotus better uses it than the Mustang, such that one can produce a better result with the same objective measure of one aspect.
Two pianos with 88 keys - we don't say a Bechstein keys != Baldwin; we say both have the same keys and notes but the Bechstein executes far better, enabling superior playing and producing a nicer sound, such that one can produce a better result with the same objective measure of one aspect.
Two GPUs with 1024 4-wide SIMD ALUs clocked at 1 GHz - for some reason we say the same calculated 1024 x 4 x 1 Giga, 4 teraflops, is not the same. Unless AMD teraflop numbers are literally counting something different to nVidia ones, a FLOP is a FLOP is a FLOP just as a pixel is a pixel and a Hz is a Hz and a BHP is a BHP.
