Page 1:Memory Scaling On AMD's Trinity Architecture
Page 2:Test System And Benchmarks
Page 3:Results: 3DMark And Aliens Vs. Predator
Page 4:Results: Battlefield 3 And F1 2012
Page 5:Battlefield 3, Frame By Frame
Page 6:Results: Skyrim And StarCraft II
Page 7:Power, Average Performance, And Efficiency
Page 8:When Does Spending 50% More Become A Great Value?
Power, Average Performance, And Efficiency
Our motherboard automatically ramped up voltage whenever we changed the memory ratio to run at DDR3-2400 data rates. Those changes, which we didn't ask for, but are imposed (and maybe even necessary for stability) anyway are reflected in a big power consumption jump compared to running at DDR3-2133. The only way to get normal readings at 2,400 MT/s would have been to use modules rated for that data rate, and not overclocked to get there.
Interestingly, GPU power increases about 1 W for every step up in memory data rate (excluding the DDR3-2400 result, which is abnormal and due to the motherboard’s automatic voltage increases). For this combination of motherboard and memory, the easiest way to have your performance and efficiency is to leave it at XMP values.
DDR3-2400 only yields 2% more performance than DDR3-2133, while DDR3-2133 generates a huge 20% gain over mainstream DDR3-1600.
If all else were equal, we’d expect the performance lead attributable to DDR3-2400 data rates to show up in an efficiency chart. Because Asus' board juices voltage to enhance stability, though, we don't get the efficiency pop we were expecting.
Even still, the 20.4% speed-up from DDR3-2133 is far larger than the power consumption premium imposed by G.Skill's kit. So, that faster memory setting does deliver higher efficiency.
- Memory Scaling On AMD's Trinity Architecture
- Test System And Benchmarks
- Results: 3DMark And Aliens Vs. Predator
- Results: Battlefield 3 And F1 2012
- Battlefield 3, Frame By Frame
- Results: Skyrim And StarCraft II
- Power, Average Performance, And Efficiency
- When Does Spending 50% More Become A Great Value?