Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Joe NYC who wrote (198197)	5/21/2006 4:06:25 AM
From: Joe NYC	Respond to of 275872

Collector's item: Top of the line Itanium Madison, 1.6 GHz, 9MB of L3, current bid $9.99 cgi.ebay.com

To: Joe NYC who wrote (198197)	5/21/2006 4:12:10 AM
From: NicoV	Read Replies (4) \| Respond to of 275872

I wonder why INTC release such a large range of frequencies. IIRC, Opteron was launched with clock speeds ranging from 1.4 to 1.8 (maybe 2.0, I don't remember) GHz. 3 years later, Opteron is at 2.8 GHz. Intel is launching Woodcrest with a frequency range similar to what has been seen over the entire lifespan of Opteron so far. What kind of yield distribution corresponds to this kind of frequency range and this kind of pricing structure? The only thing I can come up with is a very flat yield curve, rather than a yield curve with a high peak. The flat yield curve (I think) corresponds to one of the current big problems in scaling processors: variability, i.e. in a particular process, there are wide uncontrollable swings in transistor performance. Assuming that it is true (for whatever reason) that there is a flat yield curve, what are the consequences? I can see a number of problems: - There must be a lot of CPU's that work perfectly, but only at very low frequencies. This probably explains the existence of the 1.6 GHZ models. Maybe there are also a lot of CPU's that only work at 1.4 GHz, but that are unsellable because to slow. - There may be very little improvement in clock speed as the process matures. The frequency range that we see now is what we will get over the lifetime of Intels 65 nm process. - A significant portion of the volume will be slow processors. Even if AMD looses the top speed crown, due to their much lower frequency spread they will be very price/performance competitive a few bins below the top speed. - The slow ramp of clock speed may tempt people to spend a bit more on their CPU, since their investment will last longer (I know that's what I would do if I would buy a Conroe: pay 50% more and keep the PC for 3 years instead of 2). That could create a lot of demand for the highest speed bins, that can not be filled because of the flat yield curve. In other words, there's a good chance that demand for the 1.6 and 1.86 GHZ models will be very low. Looking at AMD's recent history, I have a theory that nobody seems to have mentioned here yet. For the last few quarters, AMD has been very capacity constrained. That means that they must have turned all process variables towards high yields, sacrificing high speed. They may even have stopped using their stress mechanism if it would sacrifice yields. Now that AMD has more breathing room in terms of fab capacity, they will probably turn all process variables towards high speed instead of yield, given the new competition of Conroe. That could explain the talk of the Inquirer about AMD increasing clock speed again. What would be net result? Intel with a lot of low speed unsellable NGA parts at the low end, and a much more competitive environment at the high end than currently perceived (the current mood is that AMD has completely lost the performance crown). EDIT: Looking back to the launch of Opteron, how much would have been the demand for 800MHz, 1 GHz and 1.2 GHz Opterons?

To: Joe NYC who wrote (198197)	5/21/2006 10:32:20 AM
From: heatsinker2	Read Replies (2) \| Respond to of 275872

Joe, re Theo Valich's Woodcrest article There's several interesting tidbits: 1. TDP back up to 80W 2. Intel's power number is typical, while AMD's 89W is worst case. So Theo pronounces parity, although its certainly possible that AMD actually has the lead. But then again, maybe Theo is wrong and Charlie is right. 3. Theo states "Woodcrest CPU is a monster, that's only in single and dual socket terms". But it shouldn't be that simple , right? I assume that dual socket is less of a monster than the single socket, and, of course, Michael Dell tells us the 4 socket Woodcrest system is still a dog. 4. Theo brings up the north bridge issue and an additional 10-30 watts. Can you or anyone else provide some more details on this? Also, there is memory dissipation, where AMD also seems to have an advantage. Can anyone, other than the extremists Doug and Pete, shed some light on this associated dissipation issue? hs2