First, a little history. When I went SLI with two GTX-970s back in May, I documented a real and significant performance increase. About two months ago I noticed things slowing down. I played with graphics settings for about a month before I discovered that my SLI had been disabled. So again, just a month ago, I verified a significant performance increase in Elite when SLI was enabled. Using the 3D Mark graphics test, I went from about 8700 to about 15000.
I just reran some tests today. Following are the results. Unless otherwise noted, all tests were done at 2560x1440p, refresh rate 59.95Hz, vertical sync on, frame rate limit off. I did tests at the default ultra, high, and mid settings, and at 1.0, 1.5, and 2.0 supersampling, in a few different situations.
Test 1: On a planet surface, in the SRV driving around a hilly area.
SLI Enabled
Ultra 1.0 = ~45-50 fps
Ultra 1.5 = ~25-30 fps
High 1.0 = ~55-60 fps
High 1.5 = ~30-35 fps
Mid 1.0 = 60 fps
Mid 1.5 = ~40-45 fps
SLI Disabled
Ultra 1.0 = ~50-55 fps
Ultra 1.5 = ~25-30 fps
High 1.0 = ~55-60 fps
High 1.5 = ~30-35 fps
Mid 1.0 = 60 fps
Mid 1.5 = ~45-55 fps
These were very close, almost too close to call, but there did appear to be a slight performance with SLI disabled.
Test 2: Hanging around a station exterior near the slot.
SLI Enabled
Ultra 1.0 = 60 fps
Ultra 1.5 = ~50 fps
High 1.0 = 60 fps
High 1.5 = 60 fps
Mid 1.0 = 60 fps (blurry shadows)
Mid 1.5 = 60 fps (blurry shadows)
Mid 2.0 = 40 fps (bad shadows)
SLI Disabled
Ultra 1.0 = 60 fps
Ultra 1.5 = 60 fps
High 1.0 = 60 fps
High 1.5 = 60 fps
High 2.0 = ~40-45 fps
Mid 1.0 = 60 fps (blurry shadows)
Mid 1.5 = 60 fps (blurry shadows)
Mid 2.0 = ~50-60 fps (bad shadows)
Again, fairly close, but still a slight improvement with SLI disabled.
Test 3: Sitting in a hanger looking straight ahead. Wait about 2 minutes for the FPS to settle.
This would be the easiest test to duplicate.
SLI Enabled
Ultra 1.0 = 60 fps
Ultra 1.5 = ~40 fps
High 1.0 = 60 fps
High 1.5 = ~50-55 fps
High 2.0 = ~30-35 fps
Mid 1.0 = 60 fps
Mid 1.5 = 60 fps
Mid 2.0 = ~40-45 fps
SLI Disabled
Ultra 1.0 = 60 fps
Ultra 1.5 = ~35 fps
High 1.0 = 60 fps
High 1.5 = ~50 fps
High 2.0 = ~30 fps
Mid 1.0 = 60 fps
Mid 1.5 = 60 fps
Mid 2.0 = ~40 fps
Again, these were very close, but this time a slight edge went to having SLI enabled.
Anyway, all the tests were so close that I'd be willing to say that they were within an acceptable range of error and in reality there was no significant difference between the two, unlike in the past where there was a significant difference between the two. So not I'm wondering if it is something FD did, or if there is possibly something wrong with my setup - maybe a config file or something? Based on rerunning the 3D Mark test and getting a similar score, SLI is clearly working on my machine.
If anyone else is using SLI, maybe you'd be willing to run some tests? Just doing the hanger tests would be easiest and the most repeatable. If you're at a lower resolution, maybe turn off V-sync and just compare maximum FPS.
Any advice or suggestions appreciated. Thanks.
