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: BUGGI-WO who wrote (238391)	8/8/2007 2:43:39 PM
From: pirasa2	Read Replies (2) \| Respond to of 275872

There isn't anything I "doesn't" get. The Brisbane/Windsor for the most part is already a Sempron. It had to happen and it has happened. Wish it had happened back in Q4 2006. They are getting around 50 dollars for desktop CPUs, how much worse can that get? AMD can thrive in a market structure whereby it can move its low end in volume AND has some meaningful presence in high-end gaming and server. Intel has been unable to effectively mess with AMD's low end for several years now. As long as AMD doesn't make pricing blunders, as it did in Q1 2007, it should easily peddle a low-end range of 2.1 - 3 GHz Brisbanes in large volume in 2008. They will definitely be able to produce them in volume, and the marginal cost is around 25 bucks per CPU. At $75 ASP, for instance, Intel would go bankrupt even if it had %90 of the market, yet AMD would be profitable with just 25% of the market. And all it takes for AMD to move up to $75 ASP is having 1-1.5 million high-end parts in the mix. @Pirasa - K10 intro It seems like you doesn't get the big overall picture, which we discussed here a few times. Barcelona will look good in Servers, mostly in MP systems, no question, but ask yourself what AMD could command for the "old" lines at this time and this will be still the wast majority of sales in the first 1-2 quarters at least. Keep further in mind, that Intel puts new CPU and Chipset modells in place -> you could estimate, what prices you could get for a 2G part, you could be sure not that much. In the desktop area you are hopefully well aware of the ongoing ASP crunch - looks brutal for AMD. Look further to Intels Q6600 which is NOW!!! at 266$. Go ahead in time, when AMD launches Desktop parts (Q1-08) and look what Intel will be able to do at this point. A Q6600 will be a 180$ part - very likely and I'm not speaking about the new Penryn parts, which have even higher IPCs. A 2,6G K10 QC Desktop part will be a 160$ CPU in Q1-2008, not more and try to guess, what AMD could get for old DC parts, not alone K10 DC. And it should be clear, that Brisbane will be by far the major volume driver in 2008. Brisbane = Sempron at this point. BUGGI

To: BUGGI-WO who wrote (238391)	8/8/2007 3:05:30 PM
From: eracer	Read Replies (1) \| Respond to of 275872

Re: Look further to Intels Q6600 which is NOW!!! at 266$. Go ahead in time, when AMD launches Desktop parts (Q1-08) and look what Intel will be able to do at this point. A Q6600 will be a 180$ part - very likely and I'm not speaking about the new Penryn parts, which have even higher IPCs. A 2,6G K10 QC Desktop part will be a 160$ CPU in Q1-2008 The Q6600 is a 2.4GHz part, while the Q6700 is 2.67GHz and currently priced over $500. If a 2.6GHz K10 can keep up with a 2.67GHz Q6700 (not yet known) then a 2.6GHz K10 shouldn't be priced much lower than $260 in Q1 08.

To: BUGGI-WO who wrote (238391)	8/8/2007 3:46:43 PM
From: Petz	Respond to of 275872

When figuring the total of server CPUs, don't forget that there are 121,000 dual core Opteron CPUs in the TOP500 list, mostly in the TOP50. More than Woodcrest. top500.org Petz

To: BUGGI-WO who wrote (238391)	8/8/2007 5:09:55 PM
From: pgerassi	Respond to of 275872

Dear Buggi: Opteron K8 DC is still better in HPC than Xeon DC in 2P and 4P systems. For DRAM bound processes, Opteron 8xxx K8s take the cake even from 4P Penryn based QCs because they are still tied to a 4 FBDIMM channel memory hub. 4 channels of FBDDR2-667MHz based DIMMs have half the bandwidth of 8 RDDR2-667MHz channels with higher latency to memory to boot. And since you need 4 FBDIMMs per channel to equal 16 RDDR2 DIMMs that a 2P Opteron 2xxx or 8xxx can handle or the standard 4P MB has on it, the latency is even worse at a very high power penalty. A 4P HP DL585G2 can have 32 RDDR2 DIMMs and although the new Xeon MP chipset is supposed to be able to handle that many FBDIMMs, I haven't seen one Intel MB with that many FBDIMM slots on it (16 FBDIMMs are the most I have seen and 8 is more the standard amount). With 16 FBDIMMs, they would use about 100W more at load and idle than 16 RDDR2 DIMMs. At 32 for each comes to 180W more on the Intel side. Penryn does not save 45W per quad core over Barcelona or even 45W per DC over Windsor. So even K8 will be better for those memory bound applications like TPC-H or TPC-E than Penryn. And Penryn based DC and QC Xeons won't be able to keep up with 8P K10s on socket 1207+ based servers. The latter will be able to have 64 RDDR2 DIMMs with up to 512GB per server. Thus K8s will still have some demand and given the lower prices they will have after the waterfall, they will still have a premium ASP over their desktop cousins with plenty of demand. And that considers Griffin based K8s to be a new core. If they are included, the picture for K8 is much brighter as laptops will garner the largest revenue portion next year. Pete