Acronis Resource Center

时事通讯

Benchmarking the Benchmarks

How important are benchmarks when buying a new computer?

While it depends on what you're going to be doing, the answer for most people is that benchmarks aren't very important at all. Leaving aside the very real problem of deliberately misleading benchmarks from manufacturers, there are several reasons why benchmarks really don't matter much.

In the first place, very few users ever push the limits of a modern desktop or laptop computer. Generally speaking today, the computers are so fast even performance differences that look significant on paper simply don't matter much.

In the second place, modern computers of the same type (desktop, laptop, etc.) in the same price range generally produce about the same overall performance. In most cases, the systems are within a few percentage points of each other, often well below the threshold of normal workday tolerance. For example, will you really notice if your system can recalculate a spreadsheet in 25 seconds while your neighbor's computer can do it in 23 seconds? Probably not.

But this isn't always obvious from looking at the performance graphs in magazines and websites because they only show the very end of the bar charts. If you saw the whole chart, it would look, in the words of one reviewer "like a brick."

While it is possible that this or that processor, motherboard, etc. is faster or this or that specific operation, over the general range of applications the differences tend to even out. For example, the Pentium 4 processor is usually said to be faster than the AMD Athlon XP. And it is — for some kinds of applications. The Pentium 4 has a 20-stage pipeline that works to its advantage for long sequences of instructions. The Athlon has a shorter pipeline but it does better on floating point calculations. When the Pentium 4 first came out, slower Pentium III and Athlon processors routinely beat it in benchmarks because most software wasn't optimized for the longer pipelines. Today the P4 still suffers if it has to flush the pipeline to do a context switch or in comparison to the AMD on floating point arithmetic.

Does this mean there's no effective difference in computers no matter what the price? No, for two reasons. First, more expensive computers tend to use faster components such as DVDs, video cards, and memory controllers. They also tend to have more high-performance peripherals like SATA disk controllers built into the motherboard.

The second reason is that there are still some kinds of applications that stretch the capacities of modern computers. The most notable examples are games. Building a really good game machine requires more expensive components, such as fast 3D video cards, a whole lot of RAM and other features that drive up the price and make a very noticeable difference on the latest games.

There are examples in the business world as well, such as graphics work stations or computers that will be used to run full-scale DBMS systems for data mining and such.

The best advice is to pay less attention to benchmarks and more attention to exactly what you want to do with your computer. You will save a lot of money buy buying what you need as opposed to what's the "coolest" or "fastest" component on the market.