Benchmark Results: Latency And Overclocking
| Best Timings at 1.60 V | |||
|---|---|---|---|
| Row 0 - Cell 0 | DDR3-1866 | DDR3-1600 | DDR3-1333 |
| Mushkin Enhnaced Redline 997000 | 8-9-5-15 | 7-8-5-15 | 6-6-5-15 |
| Adata XPG DDR3-1600+ AX3U1600PC4G8-2P | 8-9-6-16 | 7-8-5-15 | 6-6-5-15 |
| G.Skill Ripjaws X F3-12800CL8D-8GBXM | 8-9-6-16 | 7-8-5-15 | 6-6-5-15 |
| Corsair Vengeance CMZ8GX3M2A1600C8R | 8-9-8-15 | 7-8-5-15 | 6-6-5-15 |
| Kingston HyperX KHX1600C9D3P1K2/8G | 8-10-5-15 | 7-9-5-15 | 6-6-5-15 |
| Geil Evo Corsa GOC38GB1866C9DC | 8-10-6-15 | 7-8-5-15 | 6-6-5-15 |
| PNY XLR8 MD8192KD3-1600-X9 | 11-11-6-15 | 9-10-5-15 | 8-8-5-15 |
Most of today’s modules only required 1.50 V to operate at their rated settings, but a few were designated 1.65 V designs. Without a proven track record for over-volting the memory controller of AMD’s 32 nm Llano APU, we decided to play it safe at 1.60 V. One of the 1.65 V-rated kits still took first place, as Mushkin’s Redline 997000 reached incredibly low 8-9-5-15 timings at a high DDR3-1866 data rate.
We never thought ASRock’s minimum tRAS of 5 would be a restrictive factor in our memory tests, but AMD’s new memory controller and the high-quality ICs used on today’s top modules combine to facilitate surprising low-latency capabilities.
Mushkin also wins the overclocking competition, though the top six of seven kits finished so close together than we have to wonder how much of today’s results are due to luck, and how much is due to custom binning by memory manufacturers.
