Based upon recent stories about supply problems it appears that something is going on inside Intel, but it is rather difficult to determine exactly what. The official story from the Intel PR dept is that there was “lean inventory going into Q1, and February will continue to be tight.” When asked about possible yield issues, the response from Intel was that there were ‘no reports’ of any such problems. I also asked about unit shipments in Q3 and Q4, but was told only that revenues were up 12% over Q3, and up 8% from a year earlier. When I asked specifically for unit shipments of processors, I was told that information “isn’t available”.
Still not satisfied, I looked at the financial statements for Q3 and Q4. These documents show that that operating income was just over $80 million less in Q4, yet “Interest and other income” was up by over $260 million. Hmmm… it is obvious that the demand was there for greater shipments (re: Gateway and Dell announcements), but they just weren’t available. IPDs have been told there will be no product in February at all (meaning, no processors of any kind). Pentium III prices have been slowly inching up, reflecting a supply and demand problem and the higher speed Coppermines have yet to be available in any real quantity in the DIY market.
Does this mean that Intel has ‘stumbled’ and that they will soon fall from grace? Well, don’t count on it. There could be several reasons why this situation exists, one of which is a possible internal shift in focus. Given the gains that VIA and AMD have made over the past 6 months, it is not unthinkable that Intel may be bringing out a future processor earlier than expected, and had to give up some fab space to do it. One rumored scenario has a loyal Intel partner receiving such a part in the not too distant future on an exclusive basis to accomplish three goals: regain the focus in the market, reward the loyal customer and punish the wayward ones…
Rumors are interesting, however the most recent Intel roadmap indicates that we should see the Coppermine 850MHz and 866MHz before the end of March (possibly even late Feb), with the 933MHz available in June. The 1GHz Coppermine is scheduled for Sept., which appears to be the end of the line for the P6. October will see the release of the Willamette at speeds starting at 1.4GHz.
The roadmap appears to indicate Intel will be trying to position the Celeron against the K7 in the ‘value segment’ – which likely means the K-75. This will be accomplished by releasing the 566MHz and 600MHz later in February or in early March, followed closely by 633MHz and 667MHz. By June, we should see a 700MHz Celeron, but the speed ramp seems to slow down there. According to the roadmap, there is no 800MHz Celeron in 2000.
At one point last year, Intel had claimed that mobile processor speeds would match desktop speeds by Q4. This obviously did not happen. This year, the roadmap is nowhere near so ambitious, showing only an 850MHz mobile processor by the end of Q4. In the server segment, Intel has ‘added’ a 733MHz Itanium at introduction in Q3, with the 800MHz not being released until Q4. Q1 ’01 shows the Foster introduced at 1.4GHz.
Returning to the current issue of Intel processor availability, there are some reports that Intel has had some problems with the FC-PGA packaging, further exacerbating their supply issues. This doesn’t explain why Slot 1 processors have also been scarce, though one manufacturer suggested that Intel has accelerated their transition away from Slot 1 to kill off the i440BX chipset. An Intel spokesperson denied this, indicating that the transition, which began in October, was scheduled to occur throughout 2000 and is still on that schedule.AMD Processors
In January, we were informed that AMD had increased share in the microprocessor market to around 17% (from 12% a year earlier while making a record profit. This profit was accomplished in part by an unexpected increase in flash memory sales, but appears to have been helped by decent Athlon sales. AMD made my job a bit easier by actually providing the unit sales for the quarter… ;-)
The financial statement released in January indicates that 6 million processors were sold in Q4, with approximately 800,000 of those being Athlons. The overall ASP for all AMD processors was about $80, but the Athlons brought in just under $250 each. Working out the numbers (just out of curiosity) shows that slightly over half of the revenues for the quarter ($500 million of $969 Million) was due to processor sales, and that the Athlon brought in approximately 40% of those processor-generated revenues.
Even though demand in the DIY market was very strong, it does appear that AMD did not ship everything they manufactured (1 million manufactured and 900,000 shipped). This may be due to the limited availability of motherboards during the quarter, as many potential customers have indicated they are waiting for a more ‘mature’ chipset, though it is also possible that OEM demand was not quite as strong as hoped for. According to AMD channel reps, demand is still strong this quarter and expected to increase next quarter as VIA KX-133 based motherboards appear. With more aggressive branding and additional motherboards options, OEM demand may improve also.
The Athlon 850MHz processor will reportedly be released in mid-February, with the 900MHz possibly being released early in Q2. Both of these will be manufactured in Fab 25 (Austin), and the official word is that the K-75 will continue to be manufactured solely in this fab. It would appear that K-75 is indeed destined to compete in the ‘value segment’ against the Celeron, at least at the higher clock speeds.
Sources close to AMD have indicated that bin splits in both Austin and Dresden are “very good”. In fact, AMD just demonstrated a 1.1GHz processor out of Dresden on Monday, Feb 7. According to official AMD sources, Fab 30 is not only manufacturing at .18 micron with copper interconnects but is using an ‘improved process’, which reportedly will allow even faster processor operation. With a maximum capacity of 5000 wafers/week out of Fab 30 it seems that there should be no supply issues once Fab 30 begins shipping.
When asked about what speeds we might see for the Spitfire and Thunderbird at introduction, the answer was that these “must be faster than the fastest K-75, or else nobody buys a K-75”. Note that the K-75 will still have a 300MHz L2 cache (1/3 speed for the 650MHz Athlons and higher), so performance improvements will be smaller with each successive speed grade.
One area that has caused some concern amongst AMD advocates (and allowed Intel advocates some ‘needling’ material) is the fact that Dresden is still not shipping product, despite being ‘operational’ since October. Actually, AMD has been claiming a Q2 ’00 shipment schedule from the beginning. Personally, I think they are sandbagging. After all, if the slowest part is a speculated 1GHz+, what reason would they have to ship parts only to have them run at a ‘measly’ 800MHz? There is also no immediate danger of them becoming obsolete, since they are already ahead of the curve and last quarters profits have given AMD a little more margin of safety. But then, I could be completely wrong…
There have also been many on-again/off-again rumors regarding the K6-III+. This part will be produced on an ‘improved process’ (Aluminum only, AFAIK), and will include the ‘Gemini’ technology for switching to lower speeds when battery power is low (similar to Intel’s Gueyserville). Though it is ideal for mobile applications, there are no plans to restrict it solely for this purpose.
As mentioned in the December update (and a recent article on Ace’s Hardware), the K6-2+ apparently will come from the same wafers as the K6-III+, however only half of the L2 cache will be enabled. The rationale for this appears to be that overall yield will be better, helping to keep the costs down. Contrary to what the Ace’s article indicated, however, there will not be any BIOS option to enable the additional cache, as it will be ‘hard wired’ to 128K.VIA (Cyrix/IDT) Update
As indicated last month, the Joshua processor will be officially launched at the end of February (22nd, it appears). These will be running at 300MHz+, but will use a PR rating of 400 and up. Though based upon the M-II core, it will have a larger cache, 3Dnow! Instruction set, improved MMX functionality and a dual pipelined FPU. VIA is hoping that these improvements will provide sufficient performance increases that it can compete with the 400MHz and up Celerons.
Be the first to discuss this article!