home *** CD-ROM | disk | FTP | other *** search
- Path: ifi.uio.no!usenet
- From: ludvigp@ifi.uio.no (Ludvig Pedersen)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: CHIP RAM speed test results
- Date: 17 Mar 1996 16:49:32 GMT
- Organization: Dept. of Informatics, University of Oslo, Norway
- Message-ID: <938.6650T765T2039@ifi.uio.no>
- References: <4i96hh$ilg@serpens.rhein.de>
- NNTP-Posting-Host: gymir.ifi.uio.no
- X-Newsreader: THOR 2.22 (Amiga;TCP/IP)
-
- >kp@direktor.voima.jkl.fi (Kari-Pekka Koljonen) writes:
- >>register move: 72.6ns
- >Looks like some 28MHz CPU clock.
- >>fast readw 182.0ns 11.0MByte/s
- >>fast readl 216.5ns 18.5MByte/s
- >>fast readm 243.9ns 16.4MByte/s
- >This seems to happen with data burst enabled. The overhead of
- >the movem instruction is no longer hidden by the slower memory
- >cycles.
- >BTW, I see the same on an A3000.
-
- I think you got I point.
-
- Here is my results:
-
- NoDataBurst:
-
- fast readw 122.4ns 16.3MByte/s
- fast readl 163.8ns 24.4MByte/s
- fast readm 151.2ns 26.5MByte/s
-
- DataBurst:
-
- fast readw 109.1ns 18.3MByte/s
- fast readl 133.6ns 29.9MByte/s
- fast readm 143.4ns 27.9MByte/s
-
- readl is faster with NoDataBurst, but anyway its very clear that having the
- databurst ON is an avantage here.
-
-
- <sb>Ludde - Amiga Demo Coder
- <sb>Virtual Reality & Official Be developer
- <sb>ludvigp@ifi.uio.no
-
-