On Thursday 09 March 2006 21:42, Rick Pali wrote: > > But of course the reviews don't talk about Prime95. Despite the near > identical performance with general applications, does the superior > Pentium tuning of Prime95 also carry over to the Pentium D? If so, how > much of an advantage would the 920 have over the 3800+?
I don't have examples of either to benchmark. But my reading of the specs is that the "superior Pentium tuning" is related to SSE2 performance, in which case the Pentium D will outperform the Athlon by a very large margin (at the same clock speed) - think 2 or 3 times faster. > Also, my > understanding is the L2 cache of the 920 is 2MB for each core versus > 512K each core of the 3800+. Does this also make a significant > difference? If you're working in an exponent range where the work vector would fit into 2MB cache but not 512KB, then yes. However, if you're taking PrimeNet assignments (either LL or double check) then the issue is moot, as you won't be getting exponents under 10 million to work on. Query, isn't the 2MB cache of Pentium D chips shared across the cores? > > I ask all this because I can return the 3800+ within another week and > a half. I'm tempted to do so because the 920 deal has a superior case > and a few extra advantages I'd appreciate, but they're not enough to > make me do the exchange. A significant prime95 performance improvement > might tip the scale... On Prime95 performance terms, the Intel system is going to be the winner. There may be a downside in terms of power consumption, hence heat output and cooling fan noise - though it is possible to build quiet Intel based systems, just as AMD based systems are not always silent. Regards Brian Beesley _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
