|
CPU & Motherboard Benchmarking
Just as with video cards, many benchmarks used for testing CPU and motherboards hasn't changed much, with very few pieces of software being used by many sites. Some places just use synthetic software in their reviews, such as SiSoft. Lately with some questions regarding the validity and bias of some benchmarks, such as found in Sysmark 2002. While there are many more real life benchmarks for CPU's compared to video cards some sites don't seem to acknowledge them. Software such as Sysmark 2002, Business Winstone, Content Creation Winstone, are examples of benchmarks based on programs that are useful.
But many of these programs aren't open to see the times that a process takes, rather just like synthetic programs they spit out a number. A MS Office benchmark I ran across a while ago was called Office Bench (not very original), and instead of spitting out a number, it showed all the times taken by the program (load up, run, shut down) being tested. As an added bonus it allowed testing in a more 'realworld' area, allowing for background tasks to be running (Media Player, etc.) and incorperating it into its testing. Unfortunetely, this program is now being sold for about $50 which puts it out of the reach of some smaller sites who don't wish to spend money on a program that serve no other purpose than as an benchmark.
While office applications are something that is used everyday, there are other free applications that can be used. Let us look at some other benchmarks that can be used in place of some of these (especially Sysmark 2002 which costs a fair amount of money) and are less biased when compared to the 'larger' benchmark programs. For example those with PhotoShop can get a benchmark , which runs 21 tests on a picture allowing you to see the performance differences to be seen when using Photoshop using many different filters.
Another benchmark that is very useful in testing the differences between CPU's and finding the minute differences between motherboards, is DivX encoding as well as MPEG-2 encoding. Both tests are very CPU dependent, and the software to run them is free/month trial (MPEG-2 - ; DivX - , ). Many people today are now converting video to DivX or even converting home made movies to DVD's. As such these programs are very relevant to current trends. The video to test with can be almost anything (would be good to have a standard video). Apart from video tests there is also the very popular MP3 encoding, using the 'free' LAME codec in conjunction with some front ends for this encoder.
Another very CPU dependent group of programs is that of 3D Modeling/Rendering. Programs for beginners such as Truespace up to the more advanced 3D Studio Max, Maya and Lightwave are all very CPU dependent, especially when rendering the final product. This is one area that some sites have looked into and should look into more so, as it is an area that needs CPU power to create many things. The users of these programs are always looking for ways to improve rendering speed, while keeping the quality of the final render at an extremely high quality level.
What about synthetic benchmarks such as SiSoft Sandra and related software are they good pieces of benchmarking software? The answer is both yes and no. While synthetic tests tell you exact numbers, they focus solely on one area, completely removing the other variables from the equation. Thus while in real life having 128MB of RAM compared to 512MB+ can lead to severe penalties in 3D Rendering, these synthetic benchmarks would perform the same in either case. However, as with any synthetic benchmarking tool, they are adequate in finding the general performance of a certain piece, and can help find the reasons behind a strange performance boost or decrease.
Conclusions
What can we say in conclusion. First we want to see what we can conclude about how we as reviewers can write better reviews by using better benchmarking tools. We can first of all relegate synthetic benchmarks such as 3D Mark 2001 to their proper place, that of a secondary test. We need to focus on the games people play, as well as giving as much information as possible, perhaps by using per-second graphs to show how the card performs during every second. We also do not want to forget the other tests we can do on video cards, most importantly 2D quality testing.
With benchmarks for CPU's and motherboards, we need to look not only at office applications, games and synthetic applications. We must look at other programs that many people use, 3D Rendering, video editing/rendering, and more open office benchmarks.
So before we as reviewers use a benchmark, or you as the reader look at the results, we should ask ourselves a couple of questions. Is this benchmark played (games) or used (applications), by people? Am I using this benchmark because I want to bias the reader into accepting an erroneous conclusion? Is there any other benchmarks that I can use to find out the full potential of this hardware? Do I want to spend time/effort on a benchmark to make the review better? (If you answer no to this last question, please do everyone else a favor and don't write anymore reviews). As a final statement please remember that these are my opinions, based on what I've read so take it as personally as you want to.
Agree, disagree? Discuss it in our forums
Previous Page - Video Card Benchmarks Continued
Home
|
Benchmarking - How and Why?: Benchmarks exist to determine how a particular piece of hardware performs in relation to itself, and to others. Question is, are readers really getting the information they really need?
