number of M9K

[Altera_Forum]

I have a meory with size 8704*32. Iw ant to implement it on Cyclone 3 device. Megawizard shows that I need 31 M9K+ some LUts for this memory. Fitter fails and asks for 34 M9K. Any idea?

[Altera_Forum]

A single M9K has 8192 bits and can accommodate 256 x 32. 

 (8704 / 256) = 34.  Not sure what Megawizard you are looking at, but the Memory Compiler for "RAM: 1-Port" only lets you pick 8192 and add another 512, which individually the wizard reports them as requiring 32 and 2 M9K respectively.

[Altera_Forum]

The 'Megawizard' seems to be assuming you can use all the bits - 8704*32/(9*1024) is 30.22, rounding up to 31. 

This isn't true since you can't use the 9th bit (an M9K is actually 1k by 9 bits). You could probably do something horrid in logic on top of 8192 by 36 bits (using two cycles for writes).

[Altera_Forum]

Hello, 

 I have a question on similar lines.. I have a memory with the configuration : 32 (width)x512 (depth). The sysnthesis shows that 12 instances of block RAM were inferred. The fitter report shows that: M9Ks used are : 8  Total block memory bits = 6,144 / 5,621,760 ( < 1 % ) Total block memory implementation bits = 73,728 / 5,621,760 ( 1 % )  I have 2 questions.  1) Why are 8 M9Ks used ? So my total memory size is 512 * 32 = 16384.. which can be accommodated in just 2 M9K s right? I don't understand how Quartus is mapping the block RAMs. 2) What is the difference between Total block memory bits and Total block memory implementation bits ?  Looking forward to your reply.  Thank you.

[Altera_Forum]

You should examine the Fitter report. Fitter -> Resource Section -> RAM Summary 

 It will tell you what it used and where.  You can also see Fitter->Resource Section -> Resource Utilization by Entity which may be clearer, depending on how complex your design is.

[Altera_Forum]

I think the '12 instances of block RAM were inferred' means that the fitter decided to use memory blocks for some other latches. 

Since M9K are dual ported, one M9K can be used for two RAM blocks.  The fitter can also replicate RAM in order to meet timing requirements.

[Altera_Forum]

I think the '12 instances of block RAM were inferred' means that the fitter decided to use memory blocks for some other latches. 

Since M9K are dual ported, one M9K can be used for two RAM blocks.  The fitter can also replicate RAM in order to meet timing requirements.