| ||||||||
| by 3oh6 | January 5, 2009 | ||||||||
| SLI Testing: 16X VS 8X SLI Testing: 16X VS 8XInstead of an SLI section, we decided we would tackle a question that has been coming up on the EVGA forums quite a bit lately. With the inherit limitations due to the lack of PCI-E lanes on the Intel X58 chipset, the question whether running video cards in 8X instead of 16X PCI-E slots would be detrimental to performance. This won't be the end all say all as our testing is going to be limited, but it should provide some insight and another set of numbers to discuss in this ongoing conversation.  Testing MethodologyWe used the setup that can be found in the Test Setup & Methodologies section with a couple minor changes. The detail settings are outlined for each benchmark below. They essentially follow the same benchmark guidelines as outlined for the benchmarks we just saw except all of the games were played at their highest settings at a resolution of 1680x1050 with the exception of Crysis which only utilized 4xAA. Here is a breakdown of the system settings that changed from the Test Setup & Methodologies:
These settings are a bit different from what any other benchmarks were ran at thus far in the review. Let's get started with the results. Crysis - Sphere benchmark Detail levels are all set to Very High with the resolution at 1680x1050. We ran the benchmarks with a demo of the Sphere level in DX10 and 64-bit.  The first game we looked at is Crysis with the Sphere level time demo, which consists of water, snow, and plenty of underbrush. Aside from the frame rate being very strong, there appears to be no noticeable difference between the three different setups. The margin for error with any benchmark would easily explain any gaps in frame rate from top to bottom at both the minimum, maximum, and average frame rates. We weren't actually sure what to expect from this comparison and so far it is coming up equal. Let's move on to Far Cry 2. FarCry 2 We were able to absolutely max out the Far Cry 2 settings and still maintain very acceptable frame rates with this setup. Game play is smooth as silk, or at least as smooth as Far Cry 2 gets, and the game looks absolutely incredible. All settings were set to Ultra including 8xAA and ran in DirectX 10. The benchmark again utilized the in-game Long Ranch benchmark.  Here we see perhaps a slight pattern to the results with the 16X/16X setup looking to be stronger than the other two setups with the 8X/8X setup being the weakest. The margin is on the absolute edge of not being due to the margin for error, but the gap seems to be consistent between the minimum, average, and maximum frame rates. Both setups running at least one card in an 8X slot seem to fall the smallest bit behind in average and maximum frame rates. Perhaps this is a sign of things to come or an anomaly with just Far Cry 2 and this particular time demo. Fallout 3 These last two titles were run in-game using FRAPS to record the frame rates. This method is a little hit and miss as it is difficult to run the exact same route every time but we stuck to a timed checkmark type route that lasted 10 minutes going from the Vault 101 entrance down to the river then back up to Megaton fighting two ants, and two fire ants along the way, just like in the benchmarks we just looked at. Like Far Cry 2, all settings were absolutely maxed out in game with no fiddling done to any .INI files.  Fallout 3 is the first title that we looked at where we weren't able to run a time demo and had to rely on actual game play using Fraps to record frame rates. These results clearly show that fact with the absurdly high minimum frame rate in the 8X/8X setup. The average frame rate again shows a similar pattern to Far Cry 2 but taking into consideration the fact that we are recording actual game play frame rates, the margin for error easily covers the differences and judgment can't be made either way. The only real conclusion we can make is that there doesn't appear to be anything substantial in difference up to this point between the same setups. Call of Duty: World at War The last gaming benchmark we look at is the latest Call of Duty which was again ran through a portion of a level using FRAPs as the recorder for frame rates. The level is the Little Resistance level recorded from the time we leave the boat until the time we are about to enter the bunker near the end. This means the frame rates see a little bit of cut-scene action, some 'indoors' action while we went through the right side bunker, and a whole lot of explosions and gun fire. In-game settings are all maxed out and the game ran at a resolution of 1680x1050. The frame rates are a little high because of the indoor time but like Fallout 3, we had a timed check point list setup that was stuck to for each run ensuring as close a replication as possible.  What we have here is just another set of inconclusive results. Well, we shouldn't say inconclusive because the results thus far have shown very little, if any at all, difference in frame rates between the three setups. Like Fallout 3, COD was actual game play using Fraps to record frame rates and this needs to be kept in mind when looking at the results. The most important set of numbers are the average frame rate results and those are all clustered closely together between the three setups. Futuremark 3DMark 06 & 3DMark Vantage We decided to throw in some 3DMark results as well to see what synthetic benchmarks picked up for differences between these setups, if any. Vantage is run with the Performance preset like in the previous Vantage benchmarks.  Like all of the results so far, 3DMark 06 and Vantage show no perceivable difference between the setups. All of the runs average out to just about the same with the margins in results falling under the variations from one run to the next. Even looking at the three individual runs for each setup, there is no single result that varies more than a couple hundred points from the rest. | ||||||||
| |
| Latest Reviews in Motherboards | |||||||||
|
