| ||
| by 3oh6 | January 25, 2008 | ||
| Memory Benchmarks Memory Benchmarks: It is benchmark time and there is plenty to look at so let's get started. We begin with a look at a couple bandwidth measuring programs and move on to latency. ![]() The obvious stand-out is the fact that the Project X modules at their rated frequency and timings absolutely destroy the DDR2 setup clocked to DDR2-1200 in all bandwidth tests. The FSB on the rated Project X results is 450MHz versus the 400MHz for the DDR2 setup which helps in bandwidth but the CPU frequency is lower. Moving down the graph and looking at the various overclocked numbers, we can see that Read bandwidth in Everest cracks 12k and comes awfully close to 10k in SiSoft Sandra. Overall the numbers are quite impressive and appear to scale evenly with increased memory frequency, regardless of timings. ![]() The bandwidth trend continues into the Science Mark results with both 7-7-7 timing sets of the Project X outperforming the 6-6-5 numbers, even with CPU frequency the same. Again, in order to run 900MHz, a FSB of 450MHz has to be obtained and this will influence the bandwidth results a little bit but that is the nature of the beast. A new addition to the memory results is the inclusion of just the PC Mark 05 memory bench results. Like the bandwidth results, everything scales appropriately with the exception of the grey stock results not out-performing the 6-6-5 numbers with a higher CPU frequency. This would indicate the fact that CPU frequency influences the PC Mark 05 memory benchmark rather heavily. ![]() Moving away from bandwidth numbers we land on the SiSoft Sandra and Everest latency results. The latency results of the top three timing sets are almost identical and as the memory frequency and CPU frequency go up, latency goes down according to Everest and Sandra. There are no real surprises here so let's move on to some more benchmark results that rely on the memory sub-system and overall system performance. | ||
| |
| Latest Reviews in RAM | |||||||||
|