using a 128x32 bit simple dual port memory with independent read and write clock results in following fmax for both clocks (dout is registered):
Cyclone (I) 256 MHz, Cyclone II 210 MHz (restricted)
That's a little bit strange. Especially since the fmax for the memories in the data sheet is the other way round: Cyclone (I) 200 MHz and Cyclone II 250 MHz.
BTW: according to the errata sheet this configuration does not need any restructuring from Quartus.
The numbers for simple dual port with a single clock are:
Cyclone (I) 256 MHz, Cyclone II 235 MHz (restricted)
I'm using Quartus 6.0. Will this change with a new Quartus version?
For the test they come from pins. Therefore, the setup and hold times for the pins are very long, but this not the issue. The maximum frequency is determined by the memory blocks between the input registers (address, write data) and the memory or output register (read data).
The 210MHz is correct for Cyclone II M4K in SDP (simple dual port) mode with dual, non-PLL clocks. The speed for Cyclone II M4K in SDP mode with a single, non-PLL clock is 235Mhz.
Ok, thanks for the hint/correction. One should not try to figure out performance numbers in the wrong context - no one will drive a system with 200+ MHz without using the PLL.
With the PLL (and now Quartus 6.0sp1) the numbers are (independent of single or dual clock):
AFAIK the example I'm using (same data width on both ports) does not need the additional hardware. With different port widths, which I used first, I got the additional MUX inserted by Quartus. However, in my application I can tolerate some latency, so I will MUX the data (and demux) it in an extra pipeline stage.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.