Silicon Investor (SI) -- The First Internet Community

STOCKTALK

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.

Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: Ali Chen who wrote (148034)	11/14/2001 10:51:14 AM
From: Charles Gryba	Respond to of 186894

Ali, your description of the Sandra Benches sounds great but the truth is that the current P4 excels in those benches but lags behind the Athlon in real world benches. So, I cannot deduce anything from seeing improvement in the Sandra benches. I don't see a correlation. ( Maybe a theoretical one but definitely not an empirical one ). Thanks

To: Ali Chen who wrote (148034)	11/14/2001 12:26:35 PM
From: Win Smith	Read Replies (1) \| Respond to of 186894

Ali Chen, 20% performance gain from a 512k cache seems pretty hypothetical. Even 5-7% seems pretty hypothetical to me. Of the three listed NorthWood benchmarks at hardwaremania.com , (http://www.hardwaremania.com/reviews/athlonXP/sandra.gif hardwaremania.com hardwaremania.com ) Only the Sandra memory benchmark showed a (<1%) improvement for the Northwood over the production 2ghz P4. I dug up these graphs from tech-report.com which show the fairly weak scaling of Sandra memory bandwidth numbers with processor frequency; there's less than 1% gain for going from 1.8ghz to 2.0ghz with the production P4: tech-report.com tech-report.com So maybe the memory system can do a little bit better prefetching with the bigger cache. But given the rather weak correlation between the P4/850 system's stellar memory bandwidth and its lackluster performance in other areas, I don't see these improvements leading to much better overall performance. On carefully chosen benchmarks that fit in a 512k cache but not a 256 k cache, 20% or better gain is possible, I'm sure. I haven't seen any indication that there are many existing benchmarks or real world programs where that's the case, though.