home *** CD-ROM | disk | FTP | other *** search
- Path: yama.mcc.ac.uk!usenet
- From: Christos Dimitrakakis <mbge4cd1@fs1.ee.man.ac.uk>
- Newsgroups: comp.sys.amiga.hardware
- Subject: 64-bits
- Date: Tue, 12 Mar 1996 16:11:31 +0000
- Organization: Sirius Cybernetics Corporation
- Message-ID: <3145A233.3FB4@fs1.ee.man.ac.uk>
- NNTP-Posting-Host: meehpa02.ee.man.ac.uk
- Mime-Version: 1.0
- Content-Type: text/plain; charset=us-ascii
- Content-Transfer-Encoding: 7bit
- X-Mailer: Mozilla 2.0 (X11; I; HP-UX A.09.03 9000/720)
-
- Well, usually doubling the bits is considered to be an advantage.
- About the point that the numbers used will be too large,
- there doesnt have to be only ONE instruction per reading.
- In other words, the 64-bit word may contain 2 instructions,
- which is the same as getting the two sererate 32-bit words
- one by one, though it is -ideally- twice as fast.
- But there are problems:
- If a microprocessor doubles its data bus width, its circuit
- geometry will be much more than doubled in complexity. That holds
- true especially for CISC processors. This can lead to diffuclties
- in design, unreliablility (usually because of crosstalk), overheating,
- too much use of chip area by wiring and other less important problems.
- Furthermore, the overall driving ability of a system will drop
- if its data bitwidth is double. This means that some kind of buffering
- will have to be included, to help the microprocessor/memory/whatever
- drive the output. Buffering unfortunately increase propagation delays.
- So, though the speed is ideally doubled, the propagation delay added
- by buffering makes the gain negligible.
- These are main reasons of why we haven't seen yet a 1024-bit
- processor. :)
-