Directory  Reviews Motherboard Reviews |
| Gigabyte GA-965P-S3 Review |
|
|
|
Page 6 of 7 Benchmark Tests
We'll step through the benchmarks used one by one, and give you our impressions as we go. We used the following benchmarks to evaluate the system...
Sisoft Sandra is a consistent and reliable synthetic testing suite, and is good for providing a baseline to evaluate relative performance. It's memory benchmarking is also fairly consistent as well. We won't go through the whole suite of tests however, as many of the tests do not seem to have relevance when looking at realworld performance. As you can see when comparing stock clockrate to our 3.15GHz overclock, performance is definitely scaling upwards in a positive manner, and more or less in line with the clockrate increase. Even so, and it would honestly take more in-depth comparisons to other P965 boards to verify this, the gains in memory bandwidth are far less than we would see if we had actual control over memory timings. Though we've witnessed far better numbers when looking at other review sites, we haven't spent enough time ourselves with in house P965 chipset boards to verify this. We just mention it because it's obvious from looking at other review sites and their results. As we have several P965 boards in the pipeline here at FastSilicon.com we will revisit these numbers in upcoming reviews, and we feel this conclusion will bear itself out. Everest Ultimate Edition Everest Ultimate Edition is an excellent software suite that can tell you quite literally everything you'd ever want to know about a given system, and it's become an invaluable tool to us here. From a benchmarking standpoint, there's only one real useful benchmark for our purposes, that being the Everest Cache And Memory Benchmark. We ran tests at stock clock and our 3.15GHz overclocked setting of course.
Though we have no comparison data here against other platforms yet, again we've taken looks elsewhere at other reviews of P965 platforms. In particular the memory bandwidth sections of both screens are rather low compared to what we've seen elsewhere, and this further drives home the Gigabyte boards utter lack of memory timing adjustments being a hindrance. 3DMark 2005 and 2006 are primarily graphics card benchmarks, but any gaming engine is also going to rely on CPU performance to at least some degree, so as long as we understand this context, there is some usefulness to using these test suites to evaluate system performance. While our test systems Sapphire X850XT isn't the latest and greatest of video cards, it's still quite adequate for many people's gaming needs.
 
Cinebench 9.5 Cinebench 9.5 is a rendering benchmark based on Cinema4D's 3D graphics application suite, and thus serves as a useful content creativity test. One of the tests is also multithreaded and shows the potential gain of using a dual core platform. The chart more or less speaks for itself here, with the gains achieved from overclocking quite apparent. As an aside, the gains from utilizing a dual core platform in general are also readily apparent. Even though there are still far too few multithreaded applications available, this is another example that drives home the point to us that the days of single cores are over. Divx Encoder 6.21 DivX is still a quite popular encoding format, and encoding/ripping video's to DivX format has become a fairly mainstream activity, so we include it here to give you an idea of the gains we experienced. We take a stock 1 minute 54 second WMV video file and convert it to DivX format using the encoder's default settings.
Here we see a significant gain in rendering time, of roughly 50 seconds over stock which is an increase in rendering speed of almost 1/3rd. In a much longer transcode, you should expect similar gains, and save yourself a lot of time. Winlame Prerelease4 Winlame is our preferred application for converting audio files to mp3 or ogg format. It's also at least partially multithreaded, though not with the achievable scaling inherent in CineBench 9.5. We took a 2 minute 59 second uncompressed wav file and converted it to 160kbps mp3 using Winlame.
As you can see we gained almost 4 seconds from the overclock which is slightly more than 1/3rd faster. When batch converting lots of files or much larger files, these gains can result in a lot of time saved. Half Life 2 To round things out here, we throw in one realworld game, Valve's Half Life 2, chocked to the gills with insane headcrab head-humping action. Here we recorded a 60 second timedemo in the “Follow Freeman” section of the game, experiencing much in the way of zombie love and head humpers galore. We used Fraps 2.7.4 to record framerates through the 60 second run at two different game settings. Setting one consisted of a run at 640x480 with all rendering features dumbed down to bare minimum to attempt to take the graphics card out of the equation. Setting two consisted of running at the native resolution of my LCD, 1280x1024, with all settings cranked to the maximum, with antialiasing set to 6x, and ansiotropic filtering set to 16x. The game was utterly playable at this setting despite our year old and now dirt cheap video card.
Two things are readily apparent here. In the low resolution run, the framerate gains are huge, with the framerate increase averaging 60fps in most areas of the run. In the high resolution run however, it appears our video card is obviously the limiting factor here. Still you can interpret from the low resolution run that there is a good bit of performance to be gained from the overclock, especially if you utilize a higher end graphics card. Overall, the gains from overclocking are utterly impressive, and further drive home the point of just how powerful Core 2 Duo CPU's can be. For less than the cost of an X6800 Extreme Edition, we've assembled an entire system that can for the most part match that CPU in performance, and that my friends is utterly awesome. Still, we're not here to just talk about Core 2 overclocking. What about this Gigabyte board?
Let's wrap things up. |
|||||||||
