4
\$\begingroup\$

Background

I'm working on a free universal parts library tool. In an effort to optimize memory use and performance, I'm trying to work out the practical maximum number of pins you'd ever see on a part.

Goal

Right now, our limit is set at 4096. I'd like to bring that down substantially, but I don't want to compromise our ability to support high density interconnect components.

The Ask

What is the largest package you've ever used/seen?

Thanks for viewing/responding.

high density package

\$\endgroup\$
  • \$\begingroup\$ It looks like Intel is using LGA 2011 packages on some of their newer high-end chips. My guess is some high-end processor will claim the record. \$\endgroup\$ – helloworld922 Feb 16 '13 at 8:01
  • 4
    \$\begingroup\$ Better avoid hard limits. \$\endgroup\$ – starblue Feb 16 '13 at 8:21
6
\$\begingroup\$

The largest I'm aware of at the moment is LGA2011 so I wouldn't recommend dropping your 4096 pin limit to cover the near future as it's already close to the next nearest lower power of two of 2048. I would however recommend looking at your data structures and limits to make them dynamic, if performance and memory use are a problem I assume you're using some static data structures that will waste a lot of space for two pin devices?

Instead maybe allow for a 32-bit number in the metadata (only a few extra bytes) so that if packages of greater density arrive in the future they will be covered. If you're trying to store data in a relational database or similar and currently have 4096 columns it should be normalised and store each pin description in a seperate table.

\$\endgroup\$
  • \$\begingroup\$ Thank you (and other commenters) for your feedback. Actually, my question is really the other way around. We are doing everything dynamically (almost exactly as you described) at the moment and are investigating if we can use pre-allocation to improve performance at a reasonable memory cost. You're right though, at 2048 pins, it doesn't make any sense. \$\endgroup\$ – DrFriedParts Feb 16 '13 at 19:53
  • \$\begingroup\$ I wonder if you should consider pre-allocation for the majority of cases with a more dynamic scheme for exceptional cases. While ICs with huge pincounts do exist they are by far the exception. \$\endgroup\$ – Peter Green Feb 15 '16 at 23:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.