Developing a new test method is absorbing. And it implies obtaining a multitude of test results on a multitude of PC hardware to make sure that the used benchmarks are adequate and precise. The more testbeds are involved the smaller are the chances of stumbling upon bugs during regular tests.

In particular, from this year on, the release of each new test method will be preceded by a series of 'pilot tests' with low-performing hardware. Because that is what some benchmarks may behave strangely with. This time we'll use Intel Atom and AMD Zacate for this purpose. Along with a couple of other competitors to make things more interesting for you.

Speaking of these processors, we have already conducted one testing, so competitors are quite obvious: 1) Intel Atom 330 combined with NVIDIA Ion that, in turn, features a GPU decent enough for this class of devices; 2) one more Intel Atom, D510, with the built-in Intel GMA graphics; 3) AMD E-350, the only Zacate available these days, on the example of the ASUS E35M1-M Pro motherboard. The 4th competitor is more exotic, but also more interesting. As you may know, aside from regular Athlon II processors, AMD makes low-voltage versions marked with the letter "e" in the Model Number (e.g. Athlon II X4 615e) as well as ultra low-voltage modifications marked with the letter "u". And what do you know, one of the ULV solutions — Athlon II X2 250u — has exactly the same clock rate as AMD E-350 and both Intel Atoms we've chosen. It will be very interesting to see how it performs, because for a nettop or even a miniature HTPC the extra 10 or 20 watts of consumed power are not that critical. And an ULV modification of a regular desktop CPU may be faster than Zacate which is tailored for mobile and miniature devices from the start. So, the 4th competitor is Athlon II X2 250u with a mATX motherboard based on AMD's most advanced integrated chipset 890GX. Obviously, along with its graphics core.

Testbeds

¹ — The installed 4096 MB minus the memory used by the integrated graphics core as set in BIOS by default (as determined by AIDA64).

As you can see, it's fair and square, more or less: all processors have the same clock rate and same number of physical cores. Although Hyper-Threading 'doubles' the cores of both Atoms, don't forget that Intel has deprived them of out-of-order execution. And the benefit of Hyper-Threading is not that great.

The amount of available RAM is perhaps the most significant difference of the testbeds. Every graphics core seems to take a different share of the installed 4GB. ASUS E35M1-M Pro and MSI 890GXM-G65 use approximately 512MB (a standard, more or less), ASRock AD510PV humbly requires 10MB, and ASRock A330ION occupies as much as a whole 1GB. (As you will see, being either overly modest or too greedy is not beneficial.)

We had a choice: 1) make all testbeds as equal as possible by adjusting BIOS settings (which differ from one motherboard to another, by the way); 2) use default BIOS settings, because that's how 99% motherboards work in user computers. We decided to go with the variant #2 first, and see whether additional tests were in order. But results were nothing out of the ordinary, and additional tests would've only indicated a 3-5% difference where it was expected anyway.

Write a comment below. No registration needed!