AMD CPUs, the choice for bioinformatics research

I have heard from many people that though Intel CPUs out-perform AMD CPUs on benchmarks, it's the opposite way in practice.

And I have noticed this greatly in my bioinformatics research. For researches needing a lot of string operations, AMD CPUs are much much faster.

I just benchmarked on a Python script which mainly computes the Hamming distance between two strings, of lengths between 18 to 28 characters.

The Intel Xeon Box:
CPU: Xeon X5660 @ 2.8 GHz (Six-core, TurboBoost @ 3.2 Ghz)
OS: Red Hat Enterprise Linux 6 64-bit, Linux kernel 2.6.32-220.4.2

RAM: 96 G (I assume only 48GB is accessible by one of the two CPUs in the box)

The AMD Phenom Box:
CPU: Phenom II 1090T @ 3.2 Ghz (Six-core, TurboCore @ 3.6 Ghz)
OS: Ubuntu Linux 11.10 Server 64-bit, Linux kernel 3.0.0-15

RAM: 16 GB

According to CPUBenchmark.net, the Xeon X5660 (score 8557) out-performs the Phenom II 1090T (score 5978) by around 30%. Not to mention that the Xeon box has a much faster SAS hard drive while the Phenom box has a regular SATA hard drive - a bottleneck of the benchmark script is hard drive speed. I used Python2.6 for both. Only one user thread, my script, is running on each box - so I assume TurboBoost and TurboCore are both tuned on automatically.

Time elapsed and prices (the lower the better):
Xeon Box: 17 minutes @ $1,200
Phenom Box: 15 minutes @ $180

I think you have got my idea, considering the price difference.

1 comment:

bhojpurigaane said...

thanks 4 sharing