To: paul who wrote (8863 ) 1/20/1998 9:52:00 AM From: Mark Finger Read Replies (2) | Respond to of 14631
>>" Sun is the big loser" - Oh come on Mark, how can you put DEC, >>Sequent and Sun in the same sentance and say this. Of the three, Sun is the loser. The actual losses are more in terms of PR and mindshare, rather than revenue. However, PR and mindshare are very important. In particular, DEC needs the deal worse to be able to support the continued research on its OS. I characterized Sequent as a smaller vendor. But remember that smaller vendors add up. >>Due to the complex code sharing and joint developments these >>agreements entail i expect Sun to walk away from many low volume >>deals - Sequent ships around a 1,000 servers a year, 1/70 - 1/80 >>what Sun ships. Sun has a lot more at stake to improve Solaris in >>general rather than customize it so that every one of the less than >>100 servers Sequent ships a month goes out with Solaris. I suspect >>DEC is a little more strapped for business and has less to lose by >>fiddling around with some vendors. First, consider that Sun and Sequent are in different segments of the server market. The bulk of Sun's servers will be 4-CPU models which compete with NT servers rather than Sequent. Second, consider that Sequent has the NUMA architecture, which I thought that they were also trying to get accepted by certain other vendors. If this is true, the vendor who gets Sequent has the inside track on some additional sales. >>Alpha is an intel killer on paper but not in the real world. One of the reasons for this is that NT is still a 32-bit OS, which cripples the Alpha by limiting it only to the clock rate difference, rather than allowing it the full capability of its 64-bit capability. To demonstrate, consider that the fastest NT TPC/C benchmark is only now in the range that Unix Alpha was setting over two years ago. Look at what has happened to CPU performance in the years in between. Here is what I would consider the relative ratings of CPU's at this time (I may be somewhat off because I do not follow this too closely). I am assuming a range of applications, on fully optimized operating systems (read Unix, instead of crippled 32-bit NT) and good optimizing compilers--based on Intel Pentium II being 1.0: Alpha 1.6 MIPS 1.4 PowerPC 1.2 SPARC 1.0 Intel 1.0 Note that these are subject to change based on releases, but these seem to have generally been good for the last couple years IMO.