gistfile1.txt

I *suspect* that the benchmark summary line might be wrong

  A    B      A / B
 10    8       1.25
100   50       2.00
-------------------
110   58       1.90

A straight average of 1.25 and 2.00 would be 1.62.

But it seems like 1.90 is a better descriptor.

gistfile2.txt

                                                       |  AR 2.1 | DM 0.9.6 |    DIFF |
---------------------------------------------------------------------------------------
Model.new (instantiation)                       x10000 |   0.348 |    0.020 |  17.04x |
Model.new (setting attributes)                  x10000 |   0.984 |    0.595 |   1.65x |
Model.get specific (not cached)                 x10000 |   4.455 |    6.413 |   0.69x |
Model.get specific (cached)                     x10000 |   4.382 |    0.481 |   9.12x |
Model.first                                     x10000 |   4.061 |    4.671 |   0.87x |
Model.all limit(100)                             x1000 |  10.885 |   10.293 |   1.06x |
Model.all limit(100) with relationship           x1000 |  19.869 |   24.472 |   0.81x |
Model.all limit(10,000)                            x10 |  11.407 |   10.960 |   1.04x |
Model.create                                    x10000 |  23.092 |   11.864 |   1.95x |
Resource#attributes                             x10000 |   1.769 |    1.035 |   1.71x |
Resource#update                                 x10000 |  10.840 |    6.578 |   1.65x |
Resource#destroy                                x10000 |  15.681 |   15.869 |   0.99x |
Model.transaction                               x10000 |   3.419 |    5.186 |   0.66x |
=======================================================================================
Total                                                  | 111.193 |   98.436 |   3.02x |

Another way to present the total would be:
111.193 / 98.436 = 1.129

Is it more accurate to use 1.13x than 3.02x?