ISSCC 2006: Intel Tulsa

Pages: 1 2 3 4

Conclusion

It is relatively common knowledge that Intel’s 90nm Pentium 4 based MPUs were power limited. However, it was unclear whether that would be true for 65nm products. Fortunately, Intel’s presentation included a schmoo plot that answered that question rather decisively, shown below in Figure 4; the point in blue is the target for Tulsa.


Figure 5 – Tulsa Schmoo Plot, courtesy of Intel

This schmoo plot shows that Intel was almost certainly thermally constrained for Tulsa. Based on Figure 4, it seems that a 1.4V, 3.9GHz would probably have the necessary yield. Given that the 1.25V, 3.4GHz part has a 150TDP, such a hypothetical 3.9GHz part would dissipate around 215W. So, while Pentium 4 based MPUs may not scale to the frequencies they are capable of operating at, Intel’s 65nm process is alive and healthy.

All together, Intel has given everyone quite a bit to look forward to for the second half of the year. Tulsa improves over Paxville MP in several important ways: larger and smarter caches, lower power and thermal dissipation. Naturally, the real question is how will it perform, with any luck this spring’s IDF may shed some light on that issue.

Acknowledgements

I would like to thank the following people:

  • Stefan Rusu of Intel – For his excellent presentation, comments, clarifications and an interesting discussion over wine.
  • Scott McLaughlin of Intel – For facilitating Stefan’s clarifications.
  • David Wang – For his short news story which has been cannibalized by this one.
  • Paul DeMone – For donating his transistor illustrations.

Pages: « Prev  1 2 3 4  

Discuss (14 comments)