The hardest thing about speculation is deciding what factors come into play in your assessment. To maintain credibility in any written work, the author needs to have evidence to support his statements, anything less is sensationalism. Luckily, there are three obvious factors when dealing with the hardware industry. First, what the K7 is up against; like any market, the CPU market has competition. Second, what technology AMD has in current CPU's. Lastly, what technology (that we know of) is being shared with AMD. Once we have looked at these factors, we can being to understand where the K7 may be placed in the market.

The industry finally has strong competition at the PC level of the CPU market. Intel had a comfortable reign over the PC market only until recently. The first major success for AMD has been the K6-2. The K6-2 is on par with the P2 in most respects: fast integer performance; decent FPU performance (especially with the addition of 3Dnow); MMX; x86 compatibility; and, with the new Celron 300a by Intel, there is a direct competition over price. We cannot base the performance of the K7 on these facts because the K7 is not aimed at the PC market, it is being marketed as a server and as a workstation; that puts it in competition with the Intel Xeon and the Digital Alpha. Looking at both we see that one major factor for a server chip is a big, fast cache. The Xeon supports 2 MB of full speed cache and the DEC Alpha supports 1 MB to 8 MB synchronous/asynchronous cache. While we can only assume that the K7 will have a large cache, I'm willing to bet on it. Another important factor is SMP or Symmetrical-Multi-Processing. Both the Xeon and the Alpha support SMP to a great degree, 8 and 14 respectively (the Alpha number may be wrong, DEC advertises a 14 processor server, that is all the information I could find.) In the x86 market, Intel has been the only company to produce SMP capable CPU's thus far. As Microsoft progresses, the old Windows/Dos combination is being phased out for the more stable, more robust Windows NT. Windows NT supports SMP, 2 processors for Workstation and 32 for Server. If Windows NT is to become mainstream, SMP will be available for the masses. This puts Intel and DEC in direct competition. If DEC and Intel support SMP, and Windows NT uses it, SMP is something the K7 will need. One of the most recent advances is to the 100 MHz FSB speed. On top of that, Intel is boasting 200 MHz speeds for later P2/Xeon chipsets. K7 will most likely support 200mhz FSB speeds. Lastly, DEC's Alpha already uses 64bit PCI slots and it has been stated that Intel will use 64 bit PCI with the Xeon. On top of that, Intel has AGP. In the server arena, the 64 bit PCI slots will be important for talking to devices like SCSI and NIC devices. For the workstation arena, AGP and 64 bit PCI will be important for advanced 3d video cards. AMD will have to support one or the other, if not both.

Currently, AMD has a powerful CPU. When people argue over which is better, the P2 or the K6-2, they loose sight of just how powerful the chips are. AMD's K6-2, other than having great integer performance, the recent incorporation of 3dnow SIMD, or Single Instruction Multiple Data, has given 3d gaming performance a great boost. As I stated before, though, the K7 is to be a server/workstation chip. While there may be no use for SIMD while doing server operations, people who use workstations for 3d graphics could definitely use the added 3d boost. This leads me to believe that the K7 will have either 3dnow or a similar set of instructions. Another aspect we can take into account is on-chip level 2 cache. While the AMD K6-2 does not have on-chip level 2 cache, the AMD K6-3 will, we have known about this (and waited for it) for over a year. The on-chip level 2 cache's greatness is already evident in the in Intel's Pentium-Pro and Cel 300a, DEC's Alpha, and most other chips aimed at the server market. It is very likely that an on-chip level 2 cache will be incorporated into the K7.

We have several reports of AMD working closely with other processor companies. In the past, AMD worked with Intel to produce older x86 and VIA to produce the AMD 640. We know that AMD has been working with DEC to design the K7's motherboard bus. This leads me to believe that the K7 may have 64 bit PCI and 64 bit memory architecture. We also know that AMD has reported cooperation with Motorola. This not only means greater chip production, but possible use of copper in AMD chips, which would push the chip to the GHz level.

As far as speculation goes, we can assume that the K7 will incorporate all of these things. While that would be nice, that probably won't be true. Now we do know that the K7 will use SMP, on-chip level2 cache, higher FSB speeds, a bus similar to that of DEC Alpha's, a slot 1 like interface, and SIMD. What I assume is that the K7 will have a small on-chip level2 cache with a large level 3 cache running asynchronous to the processor, a cache much like the level 3 cache of the DEC Alpha and the level 2 cache of the Pentium 2. I also assume that the bus will be 64 bit like the bus of the Alpha. The most exciting possibility is the possibility of GHz speed with the incorporation of copper into the die process. With AMD's explosive growth in the PC market, the cooperation of other processor giants, and the migration from Windows 95/98 to Windows NT, we can all look forward to the AMD K7 upping the ante in a new market.