SI
SI
discoversearch

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 : Intel Corporation (INTC)
INTC 48.76+10.9%Jan 28 3:59 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Paul Engel who wrote (23384)5/31/1997 1:21:00 AM
From: Ali Chen   of 186894
 
Paul, this is your third wrong post about PC architecture (at least). If you unable to distinguish between simple linear scalability (aka frequency and bus width) and architectural issues, please do not confuse other investors. Lucius is more than right about x86.

Your memory example is plain wrong. The original 120-ns memory was not capable of page accesses. FastPage DRAMs, then EDO DRAM, then SDRAM ... - all theses steps led to increase in memory bandwidth. For your information, the 60 ns access time is the time from the start of data burst to the first data phase only. But the major data blocks may come at the rate of 25ns for 60ns EDO DRAM, check for reference the Micron web site. This means that the standard cheap EDO DRAM can perfectly run at 80 MHz bus speed, or make up to 40x8 = 320 Mbytes/sec data transfers.

Another ARCHITECTURAL improvements were L1 and L2 caches. Without these MEMORY enhancement even P-ii is equivalent to 286 processor, which just proves the Lucious remark. BTW, contemporary caches today are at least 6 ns, down to 4 ns in PPro, and not 15 ns as you quoted in some of your previous posts. Are you still using a 486 machine?

The most antique element of the X86 architecture is total lack of operating registers - less than 4 on average. This leads to total inability to run more than 2 instructions in parallel, because the resulting CPU code uses frequently two operands that depend on previous results. The modern technique of register renaming was invented to relieve this dependency problem, but if the original code hardly contain any parallel instructions, it is barely possible to recognize this lost intentions and properly execute the code in hardware.

As a consequence of this handicapped x86 architecture, the performance improvement slips to just 20-30% over the previous generation of CPUs, even with all this sophisticated stuff. And there is no way out, except to trash the x86 as a class. Therefore, Alpha will eventually win, and you, guys, better sell off Intel as soon as possible, before too late. Today you had a very good lesson. Accept my condolences.
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext