Digitimes had an article yesterday from
AMD Taiwan (Update: Digitimes has now corrected the article to cite their sources as being Taiwanese mobo makers) saying that Athlon 64 sales are “better than expected” at that they expect to ship 600,000 of them this quarter.
A figure of 600K Athlon 64s is far below any estimates or projections that AMD has previously made for the chip, roughly half the previous lowest estimates, and about a tenth of what AMD was promising a year ago.
True, this doesn’t include Opterons, but toss them in and you only get about 750,000 Hammers sold. This accounts for roughly a tenth of AMD’s production capacity at 130nm (even after figuring in additional die size).
Not too good for a processor that’s been out a year. Prescott, which has been out a shorter period of time, now accounts for a far higher percentage of Intel’s production, probably more than 50% at this point.
Let me put it another way. Six months after launch, Intel is making more than one Prescott for each Northwood made, maybe approaching two Prescotts per Northwood.
In contrast, twelve months after launch, AMD is making one Athlon 64 for every ten Athlon XPs they make.
As we pointed out about a month ago, low yield and high prices have been a problem for AMD.
Next quarter, though, the Digitimes articles says that production ought to roughly triple, even though 64-bit Windows will still be in beta.
We’ve heard this before.
Let’s face it, in its first year of existence, the Athlon 64 has been a sales flop. (Opterons have done OK, but they’re still rather less than 10% of the server marketplace).
Total Athlon 64 sales have probably been less than 2 million units. Granted, most of the fault is probably self-inflicted, either through inability to make the things and/or pricing them too high.
But let’s give AMD the benefit of the doubt here. Let’s assume that now they can make a lot more.
What can they now sell them for?
Price Parity Cracking?
For the first time since introduction, prices for some Athlon 64 processors are now selling at a significant discount to AMD prices (which in return are connected to Intel prices).
While they are nowhere near the 33-50% discounts usually seen in Athlon/AthlonXP days, this is a new development.
AMD Official Price
Difference (in USD)
Percentage Discount From Off. Price
Please note that the discounts listed actually underestimate the actual discount
by a few percentage points since ZipZoomFly’s price includes free 2-day shipping.
In contrast, prices on AMD’s upper-end Athlon 64s are pretty close to official pricing (though FX prices are now a bit weak).
Intel’s pricing at ZipZoomFly are from 2-5% off official prices, which they’ve been for quite some time.
Will They Continue To Erode?
If Intel proceeds according to schedule, we ought not see price cuts on mainstream PIV processors until the 4GHz PIV is announced next year (the price cut one might expect on the announcement of the 3.8 Prescott is already in place).
So AMD’s prices ought to remain stable until then. Will they?
It should be pointed out that the discount we now see could be AMD’s unofficial way of getting rid of socket 754 processors given that affordable socket 939 processors will become available next month.
It’s hard to see, though, how AMD is going to triple sales out of the blue given Athlon 64’s lackluster sales up to now.
Yes, socket 939 CPUs will appeal to enthusiasts, but that’s not going to sell a million extra CPUs in the next three months. Perhaps OEMs will put more models into play.
Then again, we may instead see fire sales on Athlon 64 processors, perhaps socket 754 only, but perhaps more?
Can AMD sell millions of full-fledged Hammers for $150-200 average? Or will the historical discounts reemerge once the supply taps open up?
We shall see.