When it comes to business applications, I expected Intel’s current second-generation Core family to perform faster than AMD’s Fusion A-series, known as Llano. However, when I tested spreadsheet performance recently, the numbers were worse than anticipated.
Now, a couple of important caveats: What makes the A-series different from previous mainstream AMD parts is that it is what the company calls an “accelerated processing unit” (APU). In other words, it combines a traditional CPU and graphics onto the same chip. Intel has done much the same thing with its second-generation Core family, known as Sandy Bridge (sold under the Core i3, i5, and i7 monikers with four digits). AMD emphasizes graphics, while Intel focuses more on traditional CPU performance—which is why I expected Llano to do worse on spreadsheets. Llano does perform better on graphics in all the benchmark tests I’ve seen, and has more graphics features, such as Direct X 11 support. There have been a number of reviews focused on graphics and gaming, but I haven’t seen much on spreadsheet performance. So as usual, I wanted to try that out. (I’ve been running most of these tests on every generation of processor for the past five years or so; for instance, last year I looked at six-core desktops.)
This time out, I looked at four mainstream notebooks: an HP ProBook 6465b based on a 1.6 GHz AMD A6-3410MX; a Gateway NV55 based on a 1.5 GHz AMD A8-3500M; a Samsung NP300E5A-A01UB notebook based on a 2.2 GHz Intel Core i3-2330M; and a Lenovo ThinkPad X220 based on a 2.5 GHz Intel Core i5-2520M. All the systems run Windows 7, with 4GB of RAM, and 32-bit Excel 2010.
The Intel chips are both dual-core chips with symmetric multithreading, so they run four threads at a time. The AMD chips are quad-core chips that run at a lower nominal clock speed. Those chips and the Core i5 both offer "turbo modes," that can run one thread faster than the rated speed for a while by reducing the speed of other parts of the chip.
Here are the basic results:
Obviously, the Sandy Bridge-based systems outperformed the Llano ones, sometimes by huge margins. Even the lower-end Core i3 was much faster, most notable on the lengthy tests—specifically the huge data table test, that took the A8 over two and a half hours to run but only one and a half hours on the Core i3 and a mere 70 minutes on the Core i5.
On a number of the tests, the higher-end AMD A8 performed worse than the A6. The A6-3410MX has a base frequency of 1.6 GHz, max frequency of 2.3 GHz, 320 Radeon graphics cores running at 400 GHz, support for DDR-1600 memory, and a total power of 45 watts. The A8-3500M has a base frequency of 1.5 GHz, max frequency of 2.4 GHz, 400 Radeon graphics cores running at 444 GHz, support for DDR-1333 memory, and total power of 35 watts. In this case, the better graphics in the A8 don’t come into play, but the extra power in the A6 may yield more headroom.
I should also note that the A-series-based systems did do better than the last AMD notebooks I tested based on the older Turion X2. Of course, it's not directly comparable since that was a while back and used an older version of Excel.
AMD is planning a new desktop chip called the FX (based on a new "Bulldozer" core), which it expects to be more competitive on CPU-centric applications. Next year, it plans a new APU known as Trinity that will replace the aging Stars CPU core in the current A-series with one based on the Bulldozer design. Of course, Intel is right behind AMD with plans for new processors next year, as well.