A colleague did a very simple benchmark, and I've just replicated it.
It's a trivially simple:
dd if=/dev/zero of=/dev/null bs=65536 count=1000000
On amd64: 18.52s 3.5 GB/s
On i386: 9.69s 6.8 GB/s
Platform: Ubuntu 12.04 live CDs (amd64 desktop and i386 desktop),
running on a Dell Zino HD with 3G RAM.
My colleague got similar results on a different, more powerful machine:
Throughput CPU #CPUs #Nodes Memory Kernel Bitmode
21.1 GB/s Intel i7-920 1 1 3GiB 2.6.32 32
10.6 GB/s Intel i7-920 1 1 3GiB 2.6.? 64
I realise this is unlikely to be representative of any real-world workload,
although I'd have thought it's essentially copying data within RAM. I'd
still be interested to understand why i386 appears to perform twice as fast
as x86_64 on this measure though.
Thanks,
Brian.
--
To unsubscribe from this list: send the line "unsubscribe linux-x86_64" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
