home *** CD-ROM | disk | FTP | other *** search
- Path: informatik.tu-muenchen.de!fischerj
- From: fischerj@informatik.tu-muenchen.de (Juergen "Rally" Fischer)
- Newsgroups: comp.sys.amiga.hardware,comp.sys.amiga.programmer
- Subject: Re: Blizzard 1230IV how good ?
- Date: 12 Feb 1996 15:42:18 GMT
- Organization: Technische Universitaet Muenchen, Germany
- Distribution: world
- Message-ID: <4fnn4q$e19@sunsystem5.informatik.tu-muenchen.de>
- References: <42116493@phoenix.owl.de> <1219.6615T886T2318@Redrobe.demon.co.uk>
- NNTP-Posting-Host: hphalle5.informatik.tu-muenchen.de
- Originator: fischerj@hphalle5.informatik.tu-muenchen.de
-
-
- In article <1219.6615T886T2318@Redrobe.demon.co.uk>, Mike@Redrobe.demon.co.uk (Mike) writes:
- |>
- |> Hi, Frank Wille , on 05-Feb-96 20:23:09 you scribbled....
- |>
- |> heh, I just realised when posting this that you wrote the ramtest program ;/
- |>
- |> JRF>> Any known bugs on Blizzard 1230IV ? I guess it'll do 7mb/sec to
- |> JRF>> chipmem. Do the demos run on it ;)
- |>
- |> >7mb/sec??? What a joke! :)
- |>
- |> >These are the measurements from my A4000/040:
- |>
- |> >---8<---
- |> >Memory Speed Test
- |> >⌐ Copyright Frank Wille, February 1996.
- |>
- |> >Type read write
- |> >Chip Ram 4553 kB/s 4566 kB/s
- |> >Fast Ram 11340 kB/s 6020 kB/s
-
- huh ? slower write in fastmem ? fastmem slower than a A1200 chipmem ?
-
- |> >---8<--
- |> From my A4000/030 :)
- |> -------
- |> ╗ Memory Speed Test ½
- |> Copyright 1996 by Frank Wille.
- |>
- |> Type read write
- |> Chip Ram 3742 kB/s 5747 kB/s
- |> Fast Ram 11908 kB/s 15843 kB/s
- |> -------
- |>
- |> This is in a 32col dblpal 720x520 screen (wb)
- |>
- |> I can get chip r/w 4349/6890 kB/s on a 2 colour screen :9
- |> ^^^^
- |> so thats approaching 7Mb/sec
- |>
- could you please also test (a0)+,(a1)+ which might give different
- timings than (a0)+,dn & dn,(a1)+ ?
-
- |> anybody care to explain why writes are faster than reads?
-
- stores to memory are buffered by 020+. So something gets more ideal.
-
- Imho a synchronous 020-28/030-28 should also read 7mb/sec.
- anyone tested it ?
-
- 020-14 also only reads about 4mb/sec, but due to other reasons:
- 020/030 needs more cycles to read in general, also in fastmem,
- so a 020-14 is somehow too slow to squezze out max chipmem read
- bandwidth.
- BUT: when copying, it does 3.5 MB/sec, 16 cycles, i.e a _read in
- 8 cycles_ (that'd be 7mb/sec) and a write in 8 cycles.
-
- reason is the move.l (An),(An) is special optimized in microcode.
-
- NOTE: all above claims are due to timing exeperiences and not due to
- study of manuals. So the numbers are facts, the interpretation could
- be wrong.
-
- |>
- |>
- |>
- |>
- |> Mike
- |>
- |> --
- |> ---------------------------------------------------------------------------
- |> Mike Redrobe - Mike@Redrobe.demon.co.uk MikeRR on #Amiga
- |> ---------------------------------------------------------------------------
- |>
- ------------------------------------------------------------------------
- fischerj@Informatik.TU-Muenchen.DE (Juergen "Rally" Fischer) =:)
-
-