Quote:
|
Originally Posted by Cuog
AMD they have been doing 64 bit longer...
|
No, Intel's Itanium (IA-64 architecture—different from x86) dates to 2001; and in any case, x86-64 is based very heavily on the existing x86 standards. Being first by a few months wasn't a big deal, in terms of the learning curve.
Quote:
|
Originally Posted by Cuog
...not to mention the amount of work that is done each CPU cycle is far greater than intel.
|
Not really, it's got more to do with the number of steps that have to be executed to do a task; for a Pentium 4—and this only applies to the NetBurst architecture, not the P6*—it might be more accurate to say that it does more work, less efficiently, to achieve the same result.
Quote:
|
Originally Posted by Cuog
With an AMD to front side bus or the speed at which everything on the computer talks to each other is 2000 Mhz wheras it is i think 800 Mhz on the intel, all the components on an amd system are faster by nature and everything is faster not just the processor also with the AMD a lower clock clock multiplier is better because that creates heat and a bottle neck inside the CPU because in intel chips the info is sent to the chip at a slower rate, sped up and processed quickly then sent out more slowly again vs. AMD its sent in quickly processed at the same speed and sent right back out
|
The Athlon64 doesn't actually have a front-side bus; the memory controller is on the processor die, so it doesn't have to run through the motherboard to a separate chip. This, however, necessitates some changes in architecture, like moving the peripherals to a separate chip (which is, however, still served by a relatively fast data bus). In either case, it's not usually the bus, but rather the processor pipeline itself that is the bottleneck; this is especially true on a NetBurst processor, where the pipeline can be (I believe) up to 28 stages, versus around 16 (again, a ballpark figure) for an Athlon64. Though you can't predict performance solely based upon the ratios of pipeline lengths, it is a significant factor when dealing with the out-of-order execution and branch prediction routines (which basically try to optimize the execution of instructions, based on the available CPU resources); especially with the Prescott-core Pentium 4s, a bad choice in these routines will cause the processor to wait at inopportune times (while it backtracks, to correct the bad prediction), leading to poor performance under some circumstances. Actually, the branch prediction routines are very good on the Pentium 4, so this situation isn't necessarily typical.
By the way, the Pentium 4 will generally use a FSB of 400, 533, 800 or 1066 MHz, depending on the processor. Each of these is known as a "quad-pumped" FSB, where the actual frequency is one quarter of the listed speed, but data is transferred over the bus four times per clock cycle. It's technically more correct to refer to them as 100, 133, 200 and 266 MHz quad-pumped FSBs (but that just confuses people). The Athlon64 uses a double-pumped HyperTransport bus between multiple processors, and in some implementations (like nForce 4), between chips.
*NetBurst (or, unofficially, P7 or 786) is the name of the Pentium 4's architecture. P6 (unofficially, 686) is the name of the previous generation of CPU core architecture, found on the Pentium Pro, the Pentium II (including Pentium II Xeons), Pentium III (including Pentium III Xeons) and the Pentium Mobile.