Could alternative CPU architectures like ternary/quaternary and/or analog etc be better at matrix multiplication?
Analog is certainly better, so long as you're OK with some noise.
Is that the point of intelligence anyway? The fuzziness
85.3 GFlops: Optimizing FP32 Matrix Multiplication on a Single AMD Zen 3 Core
21–24 of 24 posts
Re: 85.3 GFlops: Optimizing FP32 Matrix Multiplication on a Single AMD Zen 3 Core
#22For comparison, the best performing GPUs today can do FP32 at > 100 TFLOP/s
My go-to rule that's held since 2008 has been that a GPU can do about 10x the math that a similarly-priced CPU running properly optimized code can.
My 5070Ti can realize about 20 TF/s while a 12-core Zen5 can do 2.
Re: 85.3 GFlops: Optimizing FP32 Matrix Multiplication on a Single AMD Zen 3 Core
#23Earlier quoted context omitted.
Analog is certainly better, so long as you're OK with some noise.
Is that the point of intelligence anyway? The fuzziness
How is intelligence fuzziness?
Which shows a sign of intelligence? "The problem is fuzzy to me." vs "The problem is understood by me."
Are you aligning fuzziness with lack of information / what you know now is all you got at the moment?
Re: 85.3 GFlops: Optimizing FP32 Matrix Multiplication on a Single AMD Zen 3 Core
#24Earlier quoted context omitted.
Is that the point of intelligence anyway? The fuzziness
How is intelligence fuzziness? Which shows a sign of intelligence? "The problem is fuzzy to me." vs "The problem is understood by me." Are you aligning fuzziness with lack of information / what you know now is all you got at the moment?
As in "fuzzy logic"