Thread,
I took some time to expand on why I thought Intel has turned from a gorilla to a dog. The following is my thinking on the subject. Feel free to offer counter view points and shoot holes into the thinking. That' should help us all in fine tuning our thinking.
Also, I want to warn you that I am going to be using a few program names and program details that I have not seen in print anywhere. I tried to restrict the whole peice to non-public information but in the cases where I though I have to mention a few things, I did. These have come from my private conversations - Caveat emptor.
Chuck
**************
The following are my thought on how, in the last 12 months, Intel transformed from a Gorilla to a Dog. To begin this piece, let's revisit 1998 to see what was happening at that time:
Intel
- Intel had a superior PPro core that was killing K6 on the marketing front ensuring that K6 stayed at the low end. Intel has good chipset business – an area where AMD doesn't have any expertise.
- For the mainstream roadmap, Intel had a 7th generation project – Wilamette with neat features such as multi-threading and large on-die caches that would have ensured Intel being ahead of AMD even if K7 was a little early to market. And, to cover the gap, Intel had CuMine.
- At the 64-bit front Intel had this massive effort called Merced, that was supposed to change the world – behind schedule but, hey, what revolutionary approach doesn't? On the IA-32 side, there was also Foster to make sure all the bases are covered.
- At the low-end, Intel had this cool project called Timna – that would takeover the low-end integrated market from the likes of Cyrix and also kill K6 business.
- On the chipset side, Intel had neat projects called Whitney and Camino that would redefine the chipset business. To ensure that happens, Intel had a massive graphics tie-in – internally and with C&T. How can anyone compete with this?
- On the process side, 0.25 was doing great and Intel was well on its way to 0.18
AMD
- AMD was knocking on Intel's low end with K6, trying to gain market share.
- AMD had a 7th generation core in the works – the K7. To create a support infrastructure at release AMD was also working on a chipset. Man, that should stretch AMD's resources. But, in case, to hedge before K7 comes, the stopgap would be K6-3. But, K6-3 had the same short-comings as K6-2.
- On process side, AMD was scrambling ramping their 0.25. Now way they are going to get 0.18 on par with Intel and also they have this speculative Cu process in Dresden, that may hurt.
In 1998 - Intel was clearly the Gorilla! What are the chances for this dog called AMD to succeed?
Now, let's get to 1999 and see where we are:
Intel
- Intel has an inferior Ppro core that will get killed by K7 on pretty much every performance test. Yes, CuMine will help but not quite. Intel has great chipset angle for the low-end – Whitney is kicking butt. But, hey what happened to the high-end? Why isn't stuffing DRDRAM down OEM's throats not working? What is that this pesky Taiwanese company called VIA will not listen to the mighty Intel?
- For the main stream roadmap – nothing has changed from 1998 – everything slipped a year. Shoot, the gap is too big for CuMine to cover. What to do. What to do. Ah, 0.13 process to the rescue.
- For the server side Intel has couple of disasters on hand again. Merced is so late, it is behind the power curve and Foster is delayed because Wilamette is delayed. Now, we are talking about McKinley which is couple of years away to make a good impression. How do you spell disaster?
- On the low-end, Timna is shaping up nicely except for this pesky cost problem with DRDRAM. Can we really do low-end and meet BOM costs without getting in native SDRAM/DDR DRAM support? And, Pinecrest, the next generation effort, is hamstrung because we still have this aging Ppro core. Shoot – the low-end is not looking that good anymore.
- On the chipset side, this DRDRAM thing is turning out to be a disaster – should Intel be supporting 2 flavors of everything? DRDRAM and non-DRDRAM as they might decide to do with, say, Almador (next generation chipset) and delay the already delayed part? Man, that leaves the door open for AMD and the Taiwanese junta to take a lead on mighty Intel even in the chipset business!
- On the process side, what happened to the process advantage at 0.18? Shoot, AMD and Intel are going to be getting out 0.18s at roughly the same time for the core mainstream product line? Well, at least AMD is so streched may be they can't meet Intel with 0.13 if they can get it out by mid-00. And, Intel has the cash to foot the expenses and move quickly to 0.13.
AMD
- Looks like AMD has brought the first 7th generation core to the market with minimal glitches. The k6 is dying and K7 made it in just nick of time. Whew! And, yes, they did a decent job for their first chipset and are finally selling some chipsets! Not bad at all.
- AMD has 8th generation core in the works – with stuff like 64-bit processing and multi-threading and all those neat features. Based on K7 data, it probably will compete pretty well with Foster or McKinley. (Merced seems DOA so I will not bother with that one)
- At the low-end, they have a low cost k6-2 core for PCOAC project and have access to a very powerful K7 core. That should compete well with Timna and more than a match for Pinecrest.
- On the chipset side, AMD has built a decent competency. And working on the server side for a good encore. That should reduce some of the gap with Intel. Still don't have graphics for the low-end but there are decent graphics cores in the market that can be had for dirt-cheap for the low-end PCOAC efforts.
- Cu looks like is slightly behind the schedule but AMD may surprise with a decent Cu execution on the yield side. But can they do 0.13 by mid-2000? Probably don't have the resources to pull it off – especially with 2 different processes to manage. But, what if? Can AMD switch to 0.13 as fast as Intel can? Will they have enough success with Athlon family for the next 2 or 3 quarter to finance a massive and speedy process upgrade to 0.13?
Wow! What a difference a year makes!! Every key area where Intel had what looked like insurmountable lead a year back, Intel is at best – on par – with AMD and realistically behind AMD. The only exception is probably manufacturing capacity and process.
That is my take on how the gorilla has turned into a dog. I think, the nay-sayers who were bitten by AMD in prior years and who may not appreciate what Intel lost in the last year think it will not be any different this time. But, mark my words. IT WILL BE.
************** |
