POST OF THE DAY
Advanced Micro Devices
In Reply To:
Dell and Opteron

Format for Printing

Format for printing

Request Reprints

Reuse/Reprint

By duttermohl
December 2, 2002

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

Chuck is right. I think the deal is already done. AMD hired Newisys for the design, started the Hammers as all Sledgehammers, and dropped the Clawhamer DP all at Dell's request. The only thing left for the formal announcement is to see if AMD can make enough of them.

Dell's entire cost model is limited number of suppliers, low inventory, and very high product turnover. This is how Dell keeps it's total cost low. I can also imagine the "Intel Inside" payola is pretty good at Dell's volume. AMD comes to Dell with "We can save you $20 ASP over your Intel cost per processor for high volume desktops. Dell looks at the extra money they need for more parts, service, advertising and re-educating the consumer and the loss of the volume discount from Intel and say "Uh, Thanks, but no thanks."

But, if AMD goes to Dell and says "We will supply a full range all your DP and MP needs. AND our Newisys designed and validated server will blow the doors off any Xeon you carry. All you have to do is chose a color for the box and slap a label on it." Dell would be interested. The margins are higher in servers and advertising costs are lower so the added inventory costs would not be as big a disadvantage as it is in the consumer desktop space. Intel will still be the sole source for desktop and mobile and the Intel Jingle will still play on commercials, but who buys an 8-way server from a TV ad?

Dell has been selling the h*ll out of low-end servers; great volume, but not much margin. The vast majority of those are 2-ways. The Itanium was supposed to cover 4-way and above, but, the take-up of IA64 has been glacial. Flogging Xeons won't work 'cause the P4 is bandwidth limited in more than a 2 way configuration. P4's need huge quantities of cache to make it scale adding to cost and heat problems, and the 3.06 and above will be hotter than ever with 120 watt cooling! .

So how does Dell get to the promised land of high margin enterprise servers? Hector said the Hammer yields are good (50%) so AMD would be able to provide enough chips from Dresden to supply most of Dell's server needs (unlike the desktop volume). If you noticed, the DP Clawhammer is gone and the Sledgehammer now covers all servers. Think of a 2-4-8 way SLEDGEhammer lineup designed, validated, warranted and serviced by Newisys, all well below the Xeon's current costs! The volume is small, but high margin, and the sales would be into a market where Dell is weak (Dell would still carry Xeons for those who want them in 2-ways where the P4 is competitive and even Itaniums with Intel built re-branded boxes), so the risk is low. The price performance delta between the Opteron and Mid-range Xeons would ensure a rapid adoption. Sortta "Intel isn't getting us there, maybe AMD will."

IMHO, I think this is the cause of AMD's $450M reorganization and Hector's statement that "UMC was qualified" for Barton production, but the JV is on hold. AMD can't make money in the low-end desktop space- the last two quarters showed that. Intel is too big, too rich, and AMD just doesn't have the capital. Now the 3400+ Clawhamer against a 3.2 GHz Northwood or 3.4 GHz Prescott will be a good competitor, but not so dramatic as to gain that much share in the desktop market. Intel has Banias coming and it will be tough to beat in the mobile space. BUT, given the choice of a 2 way Sledgehammer or a 2-way Xeon or a 2-way Itanium, the Sledge is a clear winner, and at 4-way and 8-way AMD would have a true "clean kill". So I think AMD is going to "de-emphasize" the volume desktop space and try to ride the Hammers to the top of the heap. This is not because it is the best, low risk, corporate strategy; I think it is their only hope to stay in the CPU business. I also think Hector has given up Jerry's dream of ever overtaking Intel in volume. 50% market share is a dead dream.

Dresden will have to be sold out before any Bartons will be made at UMC. As the cost-per-chip will be higher at UMC, but the capital cost is 0, AMD will only make enough Bartons to keep Intel honest. If they can sell them at even a small profit, they're worth making. If they cannot make them at UMC at a profit, Bye Bye Barton. The low-end chips will moderate in price as AMD acts more like an overflow valve on demand. When demand is good and prices are high, AMD could make a few bucks a chip. When demand is soft, AMD will cut back and cede market share, until demand returns and stabilizes, or, if the high end-stuff from Dresden is selling out, exit that business.

Before anyone thinks I have had too much turkey, let me re-state the fact that AMD has no choice. AMD has NEGATIVE margins in PC processors. Intel has a huge capital expansion project underway; they will be making 9nm chips before AMD and will have the capacity to flood the market with low cost high volume chips. Intel will price them to sell. What can AMD do? Spend billions that they don't have to compete in a low margin business? No, their only hope is to make a high margin product. As has been pointed out several times before, merchant fabs won't be able to produce high volume, low cost parts well, AMD will need a JV for that. At this time, I don't think AMD knows if they will still be in the CPU biz when the JV is due. So they have to wait to see if their strategy works.

Will it work? Well, the server space is conservative and Intel has a strong presence, but the Athlon MP has done well so AMD has a foothold. A couple of wins like Dell and/or IBM would make the Opteron acceptable to IS managers, so it could work. We will know if the announcements come.

Dave


Become a Complete Fool
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.