|Developer Diary · You Heard It Here First · 22 December 2003|
|A Dwindling Rationale for High-End Servers|
The rationale for shelling out $500,000 for a high-end server is rapidly dwindling. With multi-cpu white box Opteron systems proliferating on the horizon the workstation manufacturers, Sun, HP and IBM are on notice. To give a sense for the numbers we are talking about here:
System CINT2000 Cost 4 way opteron 1.8 ghz 46.1 ~$12,000 4 way ultrasparc Sun Fire 4800 1.2 ghz about 30 ~$220,000That's right. The Sun Fire has substantially less performance than a generic Opteron system and costs twenty times as much (as of December 2003--price is dropping by the week). It is getting really scary out there for the workstation manufacturers. I checked the eBay prospectus and it tells a similar story. Within the last month there were 4 4800s listed all for around $90,000. None of them sold. Only one of them drew bids, the highest of which was $515. That is not a typo, five hundred dollars for a system which lists at over $200,000. Even scarier was a 12-way listing. Sun lists the 4800 12-cpu system currently at $860,000. Earlier this month it was offered at a starting price of $53,000 (a 94% discount). It received no bids.
The reason this can happen is that until this year the workstation manufacturers were insulated from the clone boxes because they had proprietary buses that allowed them to make scalable systems. Ordinary PC systems with a PCI bus do not scale. In other words if you put 8 cpus in a PC you get the same performance as 4 CPUs (or less). AMDs new chip architecture allows unlimited scaling so it can directly compete with high-end systems. In many respects its new architecture is superior to the proprietary systems it is challenging at a fraction of the cost.
HP and IBM are reacting like stunned bunnies--total paralysis. I think they are waiting for their accountants to tell them what to do. Sun is gaily planning to offer Opteron systems in its product line apparently oblivious to the fact that they will never be able to compete on a price point with clone manufacturers.
An interesting sidelight is that there is a market opportunity for small computer manufacturers to jump into the game and make a name for themselves--something that has not happened since the clone boom of the late 80's.
|return to John Chamberlain's home · diary index|
|Developer Diary · email@example.com · bio · Revised 22 December 2003 · Pure Content|