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: pirasa2 who wrote (214802)	10/24/2006 6:57:03 PM
From: niceguy767	Read Replies (1) \| Respond to of 275872

Again, I'm parsing out ATI when talking about any possibility of gross margin in the 55% to 60% range. Let's see how competitive AMD's X2's can become at 65nm before pushing any panic button.

To: pirasa2 who wrote (214802)	10/24/2006 10:01:34 PM
From: pgerassi	Respond to of 275872

Dear Pirasa2: The real problem is that Intel is expected to flood the high end with 10+ million Conroes in Q4. 2.4-2.8 GHz X2's will become progressively harder sells at their current prices of $250 (retail) and up. AMD urgently needs to act proactively this time and promptly move its marketable range to 2.2 - 3.0 GHz, starting the X2 4200+ at around $140 (retail). Yet it does not seem that its binsplits will allow such a move. First, not all of the Conroes will be high end. Anything lower than 2.2GHz will be mid range to low end. Certainly all of the 2MB versions won't be in the high end. Second, I think that Intel won't make more than 9 million or so MCWs in Q4. 10+ million MCW will have to wait for Q1-07. Third, if you include all of MCW production as high end, then all of DC A64 production is high end, X2s, Turions and Opterons with some 8xx(x) SC Opterons mixed in with 4 to 8 way having huge I/O and memory bandwidth and memory size. Fourth, you should include MBs in price comparisons for both as all K8s have an advantage there in cost, quantity, quality and breadth. For Woodcrest, you should also include memory costs as FB-DIMMs are more costly both in price and power at the same specifications. At higher speeds, FB-DIMMs are practically non existant. Fifth, given DCA, expect most of the HPC market to shy away from Woodcrest especially those applications where FPGAs can be used to boost performance. Sixth, expect big losses for Intel in mobile marketshare. In desktops, with a full quarter of Dell ramping, expect some Intel marketshare loss. Lastly, Intel likely gained unit marketshare only because it expanded the definitions of servers to include high end desktops (towers). Using the traditional definitions (multisocket and/or high I/O), I believe Intel continued to lose unit share and the numbers appear to back me up. Intel lost on the basis of revenue share in either case. Given the tendency to split the CPU markets in different ways, the best that can be publically known is for overall x86 share, both unit and revenue. By this definition, AMD gained in both, even if you don't subtract anything for Itanium as far as revenue. I looked over Intel's earnings reports and CC and did not see any breakout of CPU revenue due to Itanium (or even Itanium chipsets and MBs for that matter). So while unit share is not affected very much, revenue share is higher than most have calculated here. Even $100 million in Itanium CPU sales would boost, AMD revenue share by 0.3%. There also would be some boosts in those numbers, if embedded was stripped out from Intel's numbers as they are in AMD's. As to prices, I think that after including MB price and availability (Intel continues to harp of chipset shortages affecting their numbers (a plain red herring IMHO)), the X2 prices are appropriate. Although they may go down as 65nm is released and ramped. Looking longer term, 2007, K8L and 65nm will move the sweet spot up and thus, lower prices on each speed bin. The fact that 4x4 can use quads will put the crown beyond what Intel can do as dual quad cores would be severely bottlenecked by the FSB. We see this now because Intel knows that the dual die approach will fail using quad core dies. They know that their FSB can't handle 8 cores even with tons of cache. It would be a disaster to have 8 cores with even 8MB of shared L2 for each core pair on one FSB. That is why they attempting to downplay native quad core dies. IMHO, they won't succeed. AMD OTOH, shows that they scale well to 16 cores. Pete