SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?


To: Scumbria who wrote (91761)2/5/2000 9:48:00 PM
From: Tenchusatsu  Read Replies (2) | Respond to of 1576313
 
Scumbria, <The only time that a deep pipe can hurt performance is on mispredicted branches. As a general rule, it is easier to implement a deep pipe than a shallow one, because the amount of work done per stage is proportionately less.>

On the other hand, at the Microprocessor Forum last October, it seems that pipelines can get too long. I think even Itanium's 10-stage pipeline was considered a little longer than it needs to be by some people. And to think that the P6 core has a 12-to-14-stage pipeline. Even Athlon has a shorter pipeline (10-stage), and many people consider that core to be more scalable in MHz than P6.

Also, many people have a hard time imagining how overhead can be kept under control in a super-deep pipeline. It'd be interesting to see how Willamette does it, if Willamette indeed has a longer pipeline than P6.

Tenchusatsu



To: Scumbria who wrote (91761)2/6/2000 3:12:00 PM
From: tejek  Read Replies (1) | Respond to of 1576313
 
Scumbria,

As a general rule, it is easier to implement a deep pipe than a shallow one, because the amount of work done per stage is proportionately less.

Why is that? Deeper implies more length which would seem to indicate that there is more work involved in developing that length.

ted