You can see that K7 has definitely got the guts to make Intel some serious headache. The one only disadvantage of K7, offering a proprietary platform with Slot A, is more than equalized by that. Today you need to buy a new motherboard every 6 months if you want to stay on the edge of technology anyway. What difference does it make if your next motherboard is a Slot A motherboard with an AMD chipset rather than a Slot 1 motherboard with Intel's upcoming Camino chipset? Slot A and its EV6 protocol is the superior solution. K7 may well be superior to Katmai as well.
Let me summarize the situation again:
- Right now AMD is close to releasing an improved K6-2 core running at 400 MHz. This CPU will fight the lower high end Intel CPUs and will plug into a Super7 board. Super7 boards are now a reliable solution, offering a very good price/performance ratio.
- In a few months AMD will launch 'Sharptooth', the K6-3, in speed versions of 300-450 MHz. This CPU will beat Intel's Pentium II CPUs at the same clock speed and it will still fit into a normal Super7 board. Sharptooth should keep all Super7 platform owners from moving to Slot 1. At the same time Intel will have Katmai ready to launch. However, if Katmai shouldn't have some more features than KNI and what has been disclosed by Intel yet, it will not be any faster than the current PII in every day applications. That's why I expect that Katmai will have some more new features than what we know of yet.
- Later on in the first half of 1999 AMD will launch the new K7 processor. This processor will be the first to use the new Slot A. It should mark a good moment in time considering a move from Super7 to Slot A without touching Slot1, because K7 may well be superior to any available Intel CPU. K7 will probably launch in speed versions of 500, 550, 600+.... At that time the worst case scenario for AMD would be if Intel has already got 'Coppermine' ready, a .18 micron version of Katmai with on-die L2 cache. The architecture of Coppermine's core will still be very close if not identical to Katmai, so that Coppermine can only beat K7 with higher clock speeds. Making this happen will be the task for Intel until July 1999.
If you think that KNI vs. 3DNow! will make a serious difference I would like you to consider the following. AMD will already have millions of K6-2 shipped when the first Katmai processor is sold. Thus giving a huge advantage in 3DNow! installed base vs. a non-existent installed base for KNI. K6-2 and K6-3 will continue to ship at a much more attractive price point than Katmai, since Katmai will be the latest Intel CPU available and thus be very expensive. The result is that the 3DNow! installed base will increase even more, whilst the KNI installed base will grow a lot slower. It is very unlikely that there will be more KNI CPUs in use than 3Dnow! CPUs once K7 hits the shelves. The logical equation of the above said is that ISVs must be crazy to only support KNI in their games, they should rather concentrate on 3Dnow! in case they care about the sales of their games. It seems fairly obvious that K7 will be correct in continuing to support 3DNow! rather than KNI.
Looking at the situations and projections is forging a strong believe that AMD has finally got it damn right. They are not just presenting a new CPU, this time there's a whole straight forward road map, attacking Intel step by step with the final crescendo at K7 launch. We know that this is what it takes for the introduction of a new proprietary platform such as Slot A, and I am confident that the superior design will indeed win a lot of customers.
I have to congratulate AMD for a battle plan that's laid out well. I am honestly impressed and I am not impressed easily, as you know. Intel will have to come up with some real hard work this time. I believe that only a fully new designed CPU will have a good chance against K7, so that they will need to get Willamette done a lot earlier than planned.