Decision for timing closure based on seeds

[Altera_Forum]

Hi  

 I ran a seed sweep for 14 seeds for my design and my design is passing in 3 seeds and in some seeds its failing by 10 to 15 Mhz. My Target frequency is 165 Mhz . So can I say that my design has closed timing based on just 3 seeds? And what should my justification be for bad timing failures of 15 Mhz. Can anyone help me regarding this.

[Altera_Forum]

The fitter starts by making a pseudo-random initial placement, which is affected by the seed. Thus, different seeds can lead to different final configurations with different properties (timing, resource usage). 

When your design is on the edge, you may end up with cases where your design works and others where it doesn't work (doesn't meet timing or doesn't fit).  The cases which passed timing analysis will work correctly and you can use them. So, you can say that you have achieved timing closure.  The problem is that since you're on the edge, any small changes you make may push you into the "fail" range. Best case scenario, you'll have to run another sweep to find a seed that works. Worse case scenario, you'll no longer find any cases that meet timing requirements. It's not a comfortable situation to be in.

[Altera_Forum]

Yup my design is on the edge as even in cases where i am meeting timing is with 165.34 ,offcourse with a slight change it might never meet timing again. 

Thanks a lot for your advice. One more question (although deviating from the current topic) I have found that when i increase the fitter and router efforts to 8.0 from 1.0, these constraints seem to worsen the timings which are unexpected but yes i have tried them only with the default seed ie 1. My current results were with efforts level for both at 1.0. They might provide better results if i try a seed sweep with these increased efforts levels but at the cost of a lot of time . So would you suggest me to try out this or probably work more on other constraints as my RTL has passed verification and i dont want touch it since i want to try meeting maximum seeds with some other quartus constraints. Thank you for your time. :)