Testing methodology

Before we look at the results, a brief word what I tested here. The Gigabyte is going head to head with one of its popular rivals, namely the MSI GD70 (bios 1.9B3). There have been loads of reviews on platforms based on the 790FX chipset. To avoid a boring rerun, we opted to go for more of a duel. This to see which platform works best, when equipped with a 4 and 8Gb ram combination.

So all charts are divided in 2 parts. Top part of the chart are for the 2 boards equipped with 4Gb of PC12800C8 rams and the bottom part of the chart is for the 8GB tests.

With only 2 DIMMs (4Gb) installed we manually tweaked the boards for max performance. Meaning forcing 1T command rate, lowering TRFC etc... Timings and sub timings were compared and set alike. Before each OS install each motherboard had to survive a 5 loop MemTest86 test (V4.00).

For the 8Gb tests we opted to go first for a more relaxed approach. Meaning we only forced the C8-8-8-24 timings and 1600Mhz ram speeds via upping the ram multiplier. All the other timings and sub timings were left at “auto-detect”. The 2nd part of the 8GB test, was more like the 4GB test, forcing timings as tight and performing as possible. The latter are referred as "twkd" for the Gigabyte board and "8GB 1T" test for the GD70. Why the change in name ? During the memtest86 test the Gigabyte UD5 errorred in Test 5 and 7 with command rate 1T set. This was strange as the MSI had absolutely zero issues running 1T. I tried every option possible on the Gigabyte to stabilize 1T, it just wasn't going to happen.



When testing the 3Dmarks, the Giga maintains a small advantage over the MSI. The Gigabyte even with 2T set and equipped with 8GB managed to stay ahead of the MSI board with command rate 1T. Apparently this setting is not as determining for performance as with Intel CPU's, though still gives a small boost.



Continuing the duel with Superpi 1mb and Wprime32. Clock for clock the Gigabyte manages to confirm it's slightly better efficiency over its rival. The difference might be negligible to some, yet it is constant. Clearly indicating the Giga is setup a bit tighter out of the box by its bios engineers.



When pi 1mb was neck to neck, Pi32Mb shows the true colors of this new high end board. With over 16 seconds difference when being tweaked is a quite huge performance difference. Wprime1024 results remains pretty close as this is mainly pure CPU related.



Lavalys Everest is a good indication of the achieved bandwidth, this is nice for sub timing testing. Yet I've got to warn you, sometimes it hardly reflects in real world apps. And as I found out results are sometimes erratic.



Cinebench results are in line. The MSI being surpassed clock for clock by its new competitor. Differences aren't that big, but it still shows the MSI seems a bit less efficient as the new Gigabyte board.



X264 HS Video encoding tests are really close. Yet the observed pattern is still visible. The Gigabyte remains that little bit better. Will you notice is this is daily apps ? I hardly doubt it, but overclockers and co might appreciate this little advantage.



The PCMark2005 HD suite was added as Chew from Xtremesystems reported some slow write speeds on the Gigabyte. It seems the Gigabyte is a tad faster on Read, but a lot slower in Write. Windows 7 install also took a bit longer (about 1 min 40 secs) but take into consideration that it has got a lot more features onboard (USB3.0 and SATA3.0) than the MSI board. So we skipped those results here.

All these results were a bit of a surprise to me as I first tested Max HT and later on Max ram speed. The Gigabyte lost both tests to the MSI. So the initial verdict didn't look to good for the newborn Gigabyte.

Max HT was tested by starting from 200 HT clock and slowly upping the HT clock (keeping CPU speeds around 3.4Ghz with a low NB speed). The clock had to be Superpi 32mb, X264, Everest and Wprime stable. If one of these tests failed (usually X264) the HT would be lowered one clock and retested. This till I regained total stability. Reminding you, all these tests were done on boxed air cooling and with only 2 Corsair GTX2 DIMMs installed. Better cooling yields better results.



The MSI comes out as a clear winner here. Being the bit more stable board and finishing our test without any hiccups. After the HT pretesting, CPU clocks were raised near the 4Ghz barrier, NB speed approached 2.8GHz and the Pi, X264, Everest and Wprime test were rerun. Both boards passed with flying colors at their maximum detected HT frequency.



Ram was tested in a similar way. Starting with 200HT clock and the rams at 1600Mhz C6-6-6-1 1T Unganged mode. The Corsair GTX2 rams are rated 1800Mhz C6-6-6-24. So you know it's not the rams limiting. CPU and NB speeds were kept low to rule out the CPU here. The Gigabyte topped at 1680mhz. Not bad, but as you see the MSI board goes 40mhz higher with similar settings. Overviewing all the results and keeping in mind that the Giga seems to trounce the MSI clock for clock in efficiency I can forgive the UD5 and its early bios. We tested with F2 bios in this review. New beta biosses (F3A and F3F) have just been released and tests will be rerun.