Yes it is "considered" however that statement is still correct. I'll explain it to you this way. AMD's 64bit archetecture runs off of, like intel, a 64 bit register, HOWEVER, AMD's 64bit processors process 64bits at a time.
Actually both Intel and AMD have instructions that process 32, 64, or 128 bits at a time. Core 2 and AMD's upcoming core can handle the 128 bit ones in a single cycle (see section at the bottom about SIMD). Core 1 and AMD's current chips take two cycles to do the 128 bit instructions.
This advantage is suitably noted when one runs a benchmark test Intel vs AMD. Yes intel performs at a higher clockspeed, gets better 3dmark scores, but amd gets a MUCH higher score on Mflops.
Intel's clock speeds are typically lower than the AMDs they're competing against in benchmarks these days, actually. Yet they still win every real world test. This is, however, not due to 64 bitness, which provides barely any performance boost for 99% of things on either AMD or Intel.
Also, FLOPS is a completely useless "benchmark". It measures nothing about how fast things actually get done. Clock frequency is also a useless benchmark.
Hell my athlon 4600+ x2 outperformed intel's core2 duo on the number crunching benches. The reason why Intel can beat AMD on the other benches is because there isn't a TRUE 64bit OS (unless you look over to the unix side of things. When vista launches you will most likely see AMD performing consistantly better...
I run a true 64 bit OS some of the time, Windows none of the time. AMD doesn't do much better in 64 bit benchmarks (a bit; instruction fusing in Core2 doesn't work on 64 bit instructions iirc, so AMD gains back a little there). If AMD wants to compete against Core 2 they're going to have to release a new core (which they have in the works), and they're going to have to catch up on manufacturing (which they're trying to do).
That's also not to mention when ATI and AMD finally release an answer to the GeForce 8800, you will see a GPU that meshes seemlessly with the AMD processors, giving AMD yet another boost.
More likely we'll see a really nice integrated graphics solution. On-die GPUs is not a good idea at all (yet, anyway), since most cores are thermally limited.
Now, AMD will have Intel by the short and curlies in a few months. The AMD quad cores have already been produced. (HP and AMD did a Real-Time HD videoconfrencing display, and to prove the system was real and totally working they ran a 3dmark benchmark test as part of the display. It was a quadcore AMD processor.) Now, why hasn't AMD announced or released their's yet? Because they want to perfect the interactions and workload distributions for the 4 cores before they release. Expect to see some amazing shit in a few months.
AMD can't control scheduling or process distribution; that's a software concern. The reason why they haven't released it is because their 65nm manufacturing transition is still in progress, and manufacturing quads at 90nm is prohibitively expensive.
As a bit of further explanation of a rather confusing issue:
A 64 bit chip processes single 64 bit numbers. This means it can handle much larger numbers, but it doesn't mean it can handle more things at once. The wider instructions that
can handle more things at once are called SIMD instructions, and they work on multiple 32 or 64 bit numbers simultaneously. Typically 4 32 bit ones, or 2 64 bit ones at once. Both AMD and Intel have had 4x32 SIMD for quite a few years now.
So the situations in which you will see a performance boost from 64 bit chips are only a) when you use numbers larger than 2^32 a lot, or b) when you need to use more than 2^32 bits of address space (4GB of ram). You also get a slight boost because AMD got rid of some crap in the x86 instruction set when they added 64 bit instructions on (low GPR count, some ancient addressing modes, etc...). As a tradeoff, though, your caches get effectively smaller because your pointers are larger.
In conclusion: 64 bit is meaningless until we need more than 4GB of ram. Clock frequency, execution width (that's number of pipelines, not width of pipelines), instruction scheduling flexibility, branch prediction, and memory hierarchy will
all continue to be more important factors than 64 bit for most applications.
<edit> Also, the reason the 64 bit AMDs are so much faster than their predecessors is because they completely redid the memory hierarchy. They just like advertising 64 bit better, so they try to pretend that's why it's faster. </edit>
