By: Michael S (already5chosen.delete@this.yahoo.com),
Room: Moderated Discussions
Aaron Spink (aaronspink.delete@this.notearthlink.net) on March 16, 2017 11:11 am wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on March 16, 2017 9:54 am wrote:
>
> > No, full QPI link consists from 4 quadrants, each with his own clock.
> >
> Pretty sure it is a single fwd clk per link which agrees with
> "Each link also has one and only one forwarded clock."
>
> While it is true that QPI has flexibility in quadrant allocation, the clk isn't replicated
> according to the public documentation. The individual lanes undergo an in depth training
> procedure and support periodic retraining. Can't remember if there is a disparity mechanism
> or not but it seams reasonable that at higher speeds it would be required.
Yes, see above.
BTW, what is "disparity mechanism" ?
> Michael S (already5chosen.delete@this.yahoo.com) on March 16, 2017 9:54 am wrote:
>
> > No, full QPI link consists from 4 quadrants, each with his own clock.
> >
> Pretty sure it is a single fwd clk per link which agrees with
> "Each link also has one and only one forwarded clock."
>
> While it is true that QPI has flexibility in quadrant allocation, the clk isn't replicated
> according to the public documentation. The individual lanes undergo an in depth training
> procedure and support periodic retraining. Can't remember if there is a disparity mechanism
> or not but it seams reasonable that at higher speeds it would be required.
Yes, see above.
BTW, what is "disparity mechanism" ?