An open bench-table with the following configuration was used for memory testing:
- Motherboard: DFI Lanparty UT X58-T3eH8, Intel X58, BIOS 4/28 beta
- Processor: Intel Core i7-920 D0
- Memory: Exсeleram EP3001A 9-9-9-24 1.50V 3x2048Mb
- Graphics card: RADEON HD 6950 2GB
- Storage drives: Western Digital WD1500HLFS x2 (RAID0)
- Power Supply: CoolMaster RealPower Pro 850W
- Microsoft Windows 7 64-bit SP1
- CPU-z tool (free)
- AIDA64 system information and diagnosis software suite (shareware)
- CPU-Tweaker 1.5 memory tool (free)
- HWiNFO32 information and diagnosis tool (free)
- Prime95 64-bit v.26.6b3 (free)
Exceleram EP3001a was tested with processor Core i7 memory multipliers of 5. Preliminary tests showed system stability with this multiplier and BCLK of 200MHz which corresponds to memory speed of 2000MHz. Timings were set manually. Command Rate remained at 1T during all tests, RC was set to max available for motherboard value of 30. PCIe bus frequency was settled at 100 MHz.
Voltage values were altered in the following way:
- CPU voltage (Vcore) was modified manually in reference with BCLK
- QPI voltage (FSB VTT) was adjusted manually in reference with BCLK
- IOH voltage (NB Core): 1.1V
- Memory voltage (Vmem): 1.50V / 1.55V / 1.60V / 1.65V
The CPU and motherboard (North Bridge and VRM ) were water cooled. Ambient temperature was around +24°C. The memory has not been additionally cooled, touching it did not reveal any temperature issues.
Max stable RAM speed was defined with 10MHz increment. 30-min run of Prime95 in Blend mode was used as memory stability test.
Test results were tabulated and corresponding graph was built.
The clock rate of 1800MHz with 8-9-8-24 timings at 1.60 Volts was considered to be the best overclocking result. The memory shows great potential with even values for CL, RCD, RP and voltage of 1.55 V. Further voltage increase has failed to any success, except memory settings with 8-8-8-24 timings. With RCD higher than “neighboring” timings the memory reacts positively to voltage increase giving sometimes a better overclocking results compared to even values of mentioned timings.
Below you will find screenshots of memory passing 30-min Prime95 test runs.
Results with 7-7-7-24 timings:
Results with 7-8-7-24 timings:
Results with 8-8-8-24 timings:
Results with 8-9-8-24 timings:
Results with 9-9-9-24 timings:
Out of a sporting interest, we tried to find max clock rate. After a series of trials with settings we managed to run Exceleram EP3001a at 1043 MHz (DDR3-2087 MHz) with 7-10-7-27 timings in triple-channel mode:
The secret for overclocking potential lies in Power Chip chips — increase of RC and RAS values allows further memory clock growth, at the same time, modules respond quite well to voltage gains. We could not get the memory speed to go higher but only due to insufficient processor possibilities.