欢迎访问ic37.com |
会员登录 免费注册
发布采购

6400 参数 Datasheet PDF下载

6400图片预览
型号: 6400
PDF下载: 下载PDF文件 查看货源
内容描述: 高级内存缓冲器 [Advanced Memory Buffer]
分类和应用:
文件页数/大小: 250 页 / 3863 K
品牌: INTEL [ INTEL ]
 浏览型号6400的Datasheet PDF文件第26页浏览型号6400的Datasheet PDF文件第27页浏览型号6400的Datasheet PDF文件第28页浏览型号6400的Datasheet PDF文件第29页浏览型号6400的Datasheet PDF文件第31页浏览型号6400的Datasheet PDF文件第32页浏览型号6400的Datasheet PDF文件第33页浏览型号6400的Datasheet PDF文件第34页  
DDR Interface  
initialization, DDR I/O calibration, and normal operation where the FBD channel issues  
DRAM commands. The DAREFTC CSR controls the refresh interval and period, and  
includes an enable bit to turn auto-refresh command generation on and off. The auto-  
refresh FSM may be integrated or separate from the MemBIST functions in hardware,  
but the auto-refresh function runs independently of the MemBIST state. Multiple  
MemBIST tests can be started and completed with the auto-refresh FSM running  
during, in between, and after all MemBIST tests complete.  
The Auto-Refresh FSM generates auto-refresh command requests to an arbiter in  
hardware at regularly spaced intervals defined by the DAREFTC.TREFI CSR. For a single  
rank DIMM the command spacing is equal to TREFI cycles. For dual rank DIMMs the  
command spacing is TREFI/2, alternating between the two ranks so that each rank  
receives an auto-refresh command every TREFI cycles on average. Hardware abitrates  
between the Self-Refresh FSM command requests and commands generated by the  
MemBIST function. When MemBIST is not running, the auto-refresh command is  
immediately issued on the DDR bus. When MemBIST is running, an auto-refresh  
request is posted until the MemBIST FSM can interrupt its command sequence,  
precharge all open banks on all ranks, and all DRAM timing requirements are met so  
that an auto-refresh command can be accepted. The MemBIST FSM resumes its  
command sequence TRFC later. The MemBIST FSM must interrupt its operation within  
TREFI/2 in order to avoid causing an auto-refresh command to be dropped by the  
arbiter.  
3.6  
Back to Back Turnaround Time  
The host controller is required to observe a turnaround time on the DRAM data pins  
within a DIMM when switching between read and write cycles, and when switching from  
reads from one rank vs. the other rank on the DIMM. The nominal turnaround time is  
one clock for each parameter, which is the minimum time required to prevent a collision  
of the postamble of the first transaction and preamble of the second transaction.  
Additional clocks may be required by the DIMM, especially at higher speeds.  
The three timing parameters are:  
• Read to write turnaround time. The number of additional DRAM clocks in which the  
DRAM data bus must be idle between a read from either rank and a write cycle to  
either rank.  
• Write to read turnaround time. The number of additional DRAM clocks in which the  
DRAM data bus must be idle between a write to one rank and a read from either  
rank. The write to read turnaround time to the same rank will generally dominated  
by the tWTR specification, which must also be observed.  
• Read to read turnaround time. The number of additional DRAM clocks in which the  
DRAM data bus must be idle between a read from one rank and a read from the  
opposite rank.  
These parameters are dependent on the AMB design and the DIMM PC board layout.  
The parameters are stored in the SPD. Each parameter is a 2 bit field allowing 0 to 3  
additional clocks of turnaround time. See the FB4300/5300/6400 DDR2 Fully Buffered  
DIMM Design Specification for additional details.  
Figure 3-1 shows the nominal turnaround times, with no additional clocks. Note that  
the DQS preamble and postamble may merge slightly when the first transaction is at  
worst case timings and the second transaction is at best case timings.  
30  
Intel® 6400/6402 Advanced Memory Buffer Datasheet  
 复制成功!