"A 64 bit chip should be better than a 32 bit chip"
Don't pay much attention to belgiangenius, he isn't the latter and I suspect he isn't the former...
Like all things in computing, whether or not 64 bits is better than 32 bits really depends on what you are doing. Being able to directly process 64 bit data can make a difference if you do a lot of 64 bit multiplies and divides, otherwise it isn't so useful. If you need more than 2 gigabytes (or so) of RAM, then the 64 bit addressing can be useful, but you also need OS support. Other features of the Hammers can be useful, again depending on what you are doing. The extra registers, both the general purpose and the SSE/SSE2, can make a difference if they are supported by the OS and/or the applications. The lower memory latency can be a big win if you do typical database or database like stuff, pointer chasing and random data accesses. The 1 meg. L2 cache of the initial A64s can make a difference in some applications too.
So what does it all mean? If you have fairly typical needs and you need a system relatively soon, it isn't worth waiting for. If you have a bottleneck in one or more of the above areas and can wait until Q4 and are willing to spend $1k+ for a system, then it could be worth it. Best would be if you can wait for about a year. Like the Itanium, which Intel fans have been telling us for a couple of years "the current Itanium(x) has some problems, but the Itanium(x+1) will be worth the wait", it would be better to wait for "Hammer, the Next Generation". The A64Lite (256k L2 cache, 64 bit path to memory, possibly DDR2 if that is out by then) should carry a price that is more Barton-like. The A64PremiumLager (1 meg. L2 cache, 128 bit path to memory, possibly DDR2) will be priced higher and have better performance. Heck, there might even be dual core Opterons and that gets my heart rate up...
News 
Market Data 
Discover 
