|
Advice Contact Forum SiteMap Sponsors Home PC Hardware from Made-in-China.com |
. | |
|
Prices: |
Row Cycle Time (tRC)
Historically, the row cycle time has been tRCD + tCL + tRP, meaning a Active to Read latency followed by a Read latency and a Precharge latency. In modern memory components, tRAS is more important than a simple addition of tRCD and tCL. As a result, the minimum tRC value should be no less than tRAS + tRP.
Row Refresh Cycle Time (tRFC; tREF)
Auto Refresh cycle time. DRAM cells lose charge over time and therefore, the contents need to be refreshed by reading them out from the memory cells to the sense amplifiers and then restoring the data in the memory cells of origin. This process requires one entire row cycle time at minimum, however since often more than one row is refreshed by modern controllers, on average, two additional cycles are recommended to complete the refresh in all banks scheduled.
Read to Write Time (tRWT)
Since the memory data bus is bi-directional, reads and writes use the same data lines. This requires that the data are cleared off the data traces before the bus can turn around for example from a read transaction to a write transaction since otherwise data could collide on the bus. Moreover, on each sender and receiver, the input and output buffers need to be turned On and Off, respectively. Since every transaction also needs to prep the bus by sending a preamble or dummy signal across, a turnaround requires a few bus cycles. Lower numbers mean faster turnaround but can also affect the stability.
Refresh Rate
Memory cells are tiny capacitors that are required to hold data for a minimum of 64 msec before they must be refreshed. A refresh involves reading the data out to the sense amplifiers and then writing the data back to the memory cells. In terms of timing, it means that within 64 msec all rows of data have to be refreshed but every row has to be refreshed by itself.
Older memory chips used to have 2048 or 4096 rows per bank, most current chips feature as many as 8192 rows per bank. Therefore, the refresh rate equals 64 ms/8192 or 7.8 µsec (15.6 µsec in a 4096 row design). In other words, every 7.8 µsec, all memory transactions are interrupted in order to refresh one row of memory. Setting the refresh rate to a higher value (for example 15.6 µsec) incurs the risk that the memory cells will have lost their data already and will cause system crashes.
|
AMD Athlon64 X2-3800+ (AM2) |
All advice and educational articles on LostCircuits are free, but if you feel you can, please make a small donation to us!