The quad wars, the fight between Intel and AMD over which company has the “best” quad-core processor, has entered the next phase. Like all wars, there are initial posturings and skirmishes, followed by an attack, then followed by some significant moves by one party against the other. The quad wars have entered the strategic bombing phase.
If youre in the business of reporting on technology, you know this because youre the one in the path of the bombs. In this case, those bombs are press releases from each company detailing how its respective processor is better, faster, uses less power or has a total cost of ownership thats lower.
Now that eWEEK Labs is gearing up for tests of quad-core based servers, I expect the strategic press release bombs to become more intense.
The problem is, its not at all clear that the choice of one quad-core chip versus another will really have a significant performance impact on your network. Past tests have shown that when looked at on an individual basis, server performance times are affected by a number of factors, only one of which is the speed of the processor.
In fact, depending on the application, the speed of the processor in a given server may play a relatively minor role when compared to the servers I/O channels or its pathway to storage.
As a result, the factors that affect how a server performs in your network could very well make one server work much better for you, regardless of claims by one or the other manufacturer or even the tests that eWEEK Labs does. Everything depends on the application that runs on the server and what that application does with your data.
However, that doesnt mean that in larger quantities the choice of a processor doesnt matter, because it does. For example, AMD is claiming that its new Barcelona chip, due out sometime in August, uses less power than previous chips. This new quad core processor is supposed to run at around 2GHz, slower than the processors in your desktop PC. The slower clock speed consumes less power than a processor of the same type running at a higher speed.
Its in power consumption that the equation gets more interesting. Assuming that a server meets your performance needs—that the CPU, storage and I/O are balanced for your application—then a server that uses less power can mean one of two things. The first could be that you will save on those things that are consumption-related, such as electricity consumption for the server itself, as well as consumption for cooling. The second is that you can now run more servers with the same power and cooling than you could without processors that use less power.
Remember, the single most precious resource in many data centers is real estate. IT managers have to deal with having only a certain amount of floor space for their data center—and thats it. If they can stuff a few more servers into that space, it means they dont have to expand the data center. That saves a lot of money, and is, in many cases, a more important consideration than how much the server itself costs.
So when we look at servers as the quad wars commence, its not going to be all about the performance of the processor. eWEEK Labs testing will look at the total server and how it performs in a variety of applications, and well also consider energy consumption.
But on your network, its important to remember that processor performance by itself is irrelevant. Its the total picture that matters, and that might not argue for the solution you initially thought it would.
