Arm Gravition Versus X86 on Fibonaci and FFT Running Time

Summary

According to AWS Graviton2 the Arm Graviton provides 34% better price performance than X86. In some sense, it runs faster 34% compared with X86. So I do a small experiment to check this. AWS CDK and CodeBuild with Arm based instance makes it easy to setup. So the Graviton does run about 30% faster than X86. GitHub

• Runtime Python 3.8
• Numpy 1.22.1
• Lambda memory 2048MB
• Lambda timeout 10 seconds

CDK Pipeline

CDK make it very friendly to build this pipeline, feeling like pure programming infrastructure.

Running time for Fibonaci

def recur_fibo(n):
    if n <= 1:
        return n
    else:
        return(recur_fibo(n-1) + recur_fibo(n-2))

n = 30 used for testing.

Running time for FFT single thread

data = np.random.rand(8192, 8192)
np.fft.fft(data, axis=0)

Running time for FFT multi-thread (Lambda memory 10240MB)

data = [np.random.rand(8192, 2048) for k in range(4)]
with ThreadPoolExecutor(max_worker=4) as executor:
  for x in data:
    executor.submit(np.fft.fft, x, axis=0)