I agree with Markus that the best thing to do is replicate essentially everything in the author’s setup.  In any case, you need to know the exact specifications of the original setup.  If you cannot replicate everything, here is some advice for what is important.

If your algorithms are benchmarked in a strictly computational context (no I/O components), then it is probably sufficient to replicate the exact CPU SKU, chipset and DRAM size & speed.  If there is I/O involved, you need to replicate the exact network and storage environment, including HW, drivers and kernel version. 

Of course, it is best to re-run the original published algorithm on whatever platform you are able to obtain, even if it is “identical,” to gain some measure of normalization to apply to the comparison of the new algorithm.

Thanks for your answer. I have contacted the authors and had no answer from them. So, I might replicate their work. Thery were supposed to mentioned the CPU type, i3XXX or i5xxxx, not only RAM and GHz.