We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

AR# 36413

Spartan-6 - MCB Latency between commands for DDR2/DDR3


There are some situations where the Spartan-6 FPGA MCB introduces additional latency between commands than the minimum defined by the specification. This Answer Record documents the latency that is introduced by the MCB between specific commands.


1. Precharge All to Refresh command

The JEDEC specification requires TRP or TRP + 1 (8 Bank Device) cycles as the minimum time between the commands; however, the MCB introduces TRP + 2 or TRP + 3 (8 Bank Device) cycles. This is due to the MCB adding two cycles of latency internally. This is not true when performing a Precharge Bank to Activate, since the MCB is designed so that it accounts for the internal delay and compensates for it in that case. This is internal to the MCB and cannot be changed.

2. Read command followed by a Write Command

The MCB takes more than the minimum amount of time to turn the bus around. It first must complete the read data stream before starting to issue a write command. Additionally, there are (~2) cycles of delay until the Write command is ready and another ~1 cycle of delay at the I/O.

3. Latency between alternating Activate and Read commands.

When alternating between Activate and Read commands, the first read following an activate is sent faster than the subsequent reads that follow activates. This is because the MCB arbitration only processes one command at a time and then it waits for the read to be completed before capture, which introduces additional latency when alternating between activates and reads. This still occurs with a single port, as the arbiter will not be bypassed.

To increase throughput, you may get better results using BL8 instead of BL4. Also, by re-ordering the data to reduce the amount of activate commands, throughput will be increased significantly.

Note that simulation will show all of these latencies accurately, so it is important to simulate your design withyour design's expected data pattern in order to estimate bandwidth and throughput.
AR# 36413
Date 12/15/2012
Status Active
Type General Article
  • Spartan-6 LX
  • Spartan-6 LXT
  • MIG