Re: Next, you assume absolutely no growth in production throughout the year...
No, I assumed more than a doubling, from 350,000 to 750,000 as other manufacturers besides Samsung began volume production.
I did miss that samsung expects to be able to produce 1.5 ramping to 3 million of each, I apologize for that mistake.
But you are assuming a trouble free ramp from zero to 6 million units per month over a 6 month period. You see no chance for less than a perfect, on schedule ramp? There is zero experience with large volume production of rambus.
Microsoft.s minimum requirements will allow you to boot the machine, but not much else. With present memory prices what they are, it's silly not to have sufficient RAM - and Microsoft know it. I'm running W2K Beta 3, (although RC1 arrived last week - I haven't gotten around to installing it yet), and it needs to do too much swapping with 128meg. This may be because it still contains debug code, but as it stands, running Office 2000 on W2K, 256 meg will be what the top 25% buyers want.
Our developers are running 256 meg machines now with NT4 and visual studio or smalltalk (visual age). I expect that by the middle of next year that will be a standard high end memory requirement. Note that I only expected a quarter of the top 15% of machines to call for that amount of RAM - about 6% of the market. I'm expecting the top 15% of the market to be split between Rambus Coppermines and PC133/DDR Athlons.
You don't expect at least 6% of the top end will be using 256 meg next year?
...But you tried to gloss that over.
I didn't try to gloss over anything. I've been talking about share of the X86 PC market from the beginning, and how the constraints of price and supply on rambus limit its capability to gain a significant share of that market. I notice you failed to mention the large server market, which though limited to a few hundred thousand units a year, will be rambus free next year and each of those machines will use several gigabytes of first PC133 and later DDR dram. That should bring ram share back to being pretty close to unit share - but I was talking unit share of the market.
If rambus doesn't achieve sufficient market share near term it will become an orphan - and be avoided by all who have been stuck with token-ring, microchannel, etc. over the years.
Dan |
