Replies to post #6431 on Advanced Micro Devices Inc (AMD)

[wbmw]

FnF, do you think 2.0GHz is the magic number to success at this point? I doubt it. AMD is probably late in delivering frequency enhancements to Opteron, but this doesn't indicate a fundamental problem which would inhibit them from achieving higher frequencies later. I still think that 2.2GHz or 2.4GHz is the ceiling for Opteron on .13u. The time frame for these releases is the only unknown.

[kpf]

FnF: I am curious as to what is dictating AMD's actions

Nothing really to be curious about, I guess. We just see a usual pattern for a new process.

Remember Thoroughbred Pattern?

------------- Lauch----------- Production Sweetspot (est)
Q1/02-------- 1700+----------- 1700+
Q2/02-------- 2200+----------- 1800+
Q3/02-------- 2400+/2600+----- 1900+
Q4/02-------- 2700+/2800+----- 2000+
Q1/03-------- 3000+----------- 2200+
Q2/03-------- 3200+/(3400+?)-- 2400+?

Not sure if the production data are accurate, especially not for the current quarter but the basic pattern is clearly visible ex post from the pricing-data.

The timeframes for Clawhammer are likely to be similar - but could as well be a lil bit faster - or a lot slower.

So, Opteron production sweetspot is 1400 MHz currently, that is what pricing tells us.
As soon as production sweetspot is at 1600, we will learn quickly by falling prices for Model 244. I am not talking about AMDs pricelist, but streetprices. Easy to follow e.g. on pricewatch.com.

As for your question: "When will AMD release a 2G Opteron"?
Could well be Model 248 will be launched this months already, but that does not mean anything.

In summary: Suggest you to look at prices, not at announcements. The latters are more disinformation than information.

K.

[sgolds]

facsnotfiction, the priority is to release the 2-way and 8-way configurations reliably. Even if AMD can produce 2GHz Opterons in reasonable quantity, it would be a mistake to release it at this time. The downside of a field problem would be devastating, AMD can not even take a small risk of such a thing.

First they need to build confidence, then afterwards start cranking the clock rate (which isn't even a major selling point on server chips, anyway).