11-19-2007 11:48 AM
11-19-2007 03:07 PM
11-20-2007 07:38 PM
10-10-2008 01:05 PM
I am trying to understand it as well. Based on Arthur's reply, it (10ns) sounds as a maximum pass thru delay between an input and an output pin. If my understanding is right, it would set a maximum frequency for an FPGA of interest, only if the connections are pass-thru.
Based on my understanding, here is a follow up question: We usually have all kinds of combinational and sequential logic in between input and output pins. When we include all that, what is the significance of the speed grade?
10-10-2008 01:59 PM
The pin-to-pin delay description is only valid for CPLDs - as Arthur indicated.
FPGAs have significantly more complicated timing models. If you look at the output from speedprint (run from the command line), you'll notice that most of the timing delays of the elements are affected by the speed grade.
In general, the only good summary is "faster speedgrades are faster" The specifics will depend on the family, speed grades, your design specifics, etc.
The datasheet can also be useful to understand some of these limitations (e.g. max frequency of the global clock buffer).
The practical application of the speedgrade is very much dependent on your design, your ability to reach timing closure, the amount of time you can spend tuning your design, and the limits imposed by the silicon and the associated speed bins.
06-27-2010 07:46 PM
someone can help to sum up a total speed grade list for xilinx devices?
above discussion is really helpful to me, but it will be perfect if the information about spartan3/spartan6/virtex6 family can also be found here:)
06-28-2010 07:09 AM
All of the information above should generally apply to V6/S3/S6 too.
A few notes:
-how a speed grade correlates to actual timing is obviously dependent on the particulars of a specific design, synthesis/implementation options, constraints, placement/routing, etc.
-The tools are timing driven so it can be tough to judge margin in some cases as the tools stop when they meet your goal (e.g. how fast could I really go?)
-you really can't compare speed grades from different families based on the numbers.
-All things, being equal - I would expect a V6 -1 to be slightly faster than a V5 -1. Not because of the #s but because of what I know about the architectures. This may or may not apply to other families. For example, V4 speed grades are -10, 11, and 12. I would expect V5 -1 to be slightly faster than V4 -10.
-There are some exception to the ~15% guideline above. For example, the difference between S6 -2 and -3 is likely twice that. In other words, I would expect a S6 -2 to be slower than a S3 -4, but a S6 -3 may be slightly faster than a S3 -5. But that's just a general guideline.
-speedprint can also be a useful utility - see http://www.xilinx.com/support/documentation/sw_man
06-28-2010 07:15 AM
You should also check the datasheet for the respective family... This is likely the quickest way to see some of the performance differences associated with the speed grade of a particular device.
09-03-2010 11:14 AM
I guess it will be a pretty dumb question since it has not come up yet, but, is the speed grade a fixed hardware caracteristic or what? I mean, has it to do with the technological procedure used to manufacture de FPGA or is it a configuration parameter that can be changed? The reason I am asking is because i XPS in the project windows i double clicked on "project options -> device: xxxxxx" and changed the speed grade from -3 to -4. I thought that if this was a fixed hardware parameter I sould have received an error when building and downloading the bitstream to the board (spartan 6 sp605 by the way) but nothing happened and the bitstream behaved as expected.
Therefore, I would assume that the speed grade is just a configuration parameter used for something i don't know, but i am not sure that this is the case...
Also, how can i change the clock speed for my desing without building a new design from scratch (BSB)? when i runed BSB i fixed the frequency to 66 MHz but now i would like to try a higher one.