Friday, February 1, 2008

Penryn Performance. The Envelope Please!

Having kept pretty much to its promised schedule for the Penryn microprocessor ...

Intel is now significantly ahead of the industry with the production of 45 nm parts using the new high-K dielectric materials. The claims for the new Hafnium oxide-metal gate technology included:

  • Approximatley 2x improvement in transistor density, for either
    smaller chip size or increased transistor count

  • Approximatley 30% reduction in transistor switching power

  • Better than 20% improvement in transistor switching speed or 5x
    reduction in source-drain leakage power

  • Better than 10x reduction in gate oxide leakage power

The burning question (if I can use that term) was whether or not the new 45 nm technology would reestablish the Moore's law trajectory (so called Moore II) and lead to the production of faster, cooler single CPU parts or just somewhat faster and somewhat cooler multicores. Naturally, part of the Penryn production, viz., 5 models, is aimed at the low-power, longer battery time, laptop/mobile market, with 15 Penryn models aimed at the green server market.

AnandTech is reporting some electrical power and processor performance benchmarks on two Dell Latitude D630 notebooks equipped respectively with a 2.6GHz T7800 dual-core Merom processor operating at 1.23 volts and a 2.6GHz T9500 Penryn operating at 1.15 volts; each being connected to an 800MHz front-side bus. The tests employed the latest versions of the the BAPCo (Business Applications Performance Corporation) SYSMark and MobileMark benchmarks.

Summarizing the results in rounded-down numbers, the comparison between these two laptops shapes up like this:

  • Battery time extension based on MobileMark suite of 3 benchmark codes:

    1. Productivity: +15%
    2. Reader: +5%
    3. DVD Playback: +5%

  • Processor speed based on:

    1. SYSMark 2007: +5% with individual benchmark codes exhibiting a 1--10% boost
    2. MobileMark Productivity: +5%

On the basis of these numbers, which appear to be values from a single run of each benchmark code rather than averages based on multiple runs, it looks more like a 10% power improvement, than the advertized 20--30%, and about 5% processor speed increase. I recommend looking at the individual results for yourself while keeping the following in mind:

All benchmarking is institutionalized cheating.

In particular, severe criticisms have been leveled at the SYSMark 2007 benchmark.

The original media announcement (a year ago) was made jointly with IBM Corp., and they should soon have qualified a 45 nm process using a more traditional silicon dioxide insulator. That matters to AMD, who is relying on the IBM technology for its 45 nm production starting about mid 2008.

No comments: