by Rob Williams on March 31, 2009 in Motherboards
At CES earlier this year, Gigabyte showed off two new motherboards that promised both a great value and of course, good overclocking abilities. The EX58-UD4P was one of those, and we’ve now been able to put it to the test. We’re happy to report that as we had hoped, the board delivers on all fronts, and coupled with a reasonable price, it looks to be well-worth a look.
While application performance shouldn’t vary much between motherboards, one area where we can see greater differences is with synthetic benchmarks – at least with those that test both the storage and memory bandwidth/latency. Even still, if differences are seen, you are very unlikely to notice the difference in real-world usage, unless the performance hit is significant, which we’ve not found on any board we’ve tested in the past.
To test the storage I/O, we use a tool that we’ve been using for a number of years, HD Tune. The developer released a “Pro” version not long ago, so that’s what we are using for all of our storage-related benchmarking. The drive being tested is a secondary, installed into the first available Slave port, and is not the drive with the OS installed. To avoid potential latency, the drive is tested once Vista is idle for at least five minutes, and CPU usage remains stable at >1%.
It’s safe to say that for the most part, differences seen in I/O tests is minimal, as this graph shows. All winners and no losers, which is what we’d hope to see.
SiSoftware Sandra 2009
Yet another classic tool from our toolbox, SiSoftware’s Sandra is one of the ultimate benchmarking sidekicks around, allowing us to test almost every-single component in our PC, from CPU to GPU to memory to storage. In the case of our motherboard reviews, we stick with the memory bandwidth and latency tests, since its an area where some differences could very-well be seen.
As mentioned above, the results here don’t represent real-world performance, and if one motherboard sees the memory 4ns slower, the chances of you noticing the hit in real usage is highly unlikely, if not impossible. If any scenario would be effected, it would be processes that last the course of a few hours, not a few minutes.
Like most of our tests here, the results shown above are quite minimal, with the largest differences being around 500MB/s. That’s not too significant when dealing with 20,000MB/s+, and given that no realistic desktop scenario will ever touch that kind of bandwidth (at least today), it’s safe to say all the boards perform near-identically.