1. If you're receiving a message that you are banned from the Current Events or Politics forums, it's not you specifically: those forums have been hidden for all users. For more info, see here.
    Dismiss Notice

The Amiga Thread

Discussion in 'Retro & Arcade' started by breech, Jul 24, 2010.

?

What Amigas do you currently have in your collection?

  1. Amiga 500

    48.1%
  2. Amiga 500+

    8.0%
  3. Amiga 600(HD)

    11.2%
  4. Amiga 1000

    9.1%
  5. Amiga 1200(HD)

    29.4%
  6. Amiga 1500

    0.5%
  7. Amiga 2000(HD)

    13.9%
  8. Amiga 2500(HD)

    1.6%
  9. Amiga 3000(T/UX)

    3.7%
  10. Amiga 4000(T)

    5.9%
  11. Amiga CDTV

    2.1%
  12. Amiga CD32

    17.6%
  13. Amiga Prototype (1000+/3000+/3500/Other)

    0 vote(s)
    0.0%
  14. Amiga Compatible (Draco/Access/Minimig/Pegasos/MIST)

    2.7%
  15. AmigaOne (SE/XE/MicroA1)

    1.6%
  16. AmigaOne (500/X1000)

    0.5%
  17. Emulators - I don't have real Amigas anymore :(

    20.9%
  18. Emulators - I've never seen/owned a real Amiga, but wish I did!

    7.5%
  19. My Amigas are all original - no mods other than original expansion cards, accelerator cards, RAM car

    12.3%
  20. My Amigas feature modern-day modifications such as DVI/HDMI/IDE/SATA/USB/RAM/CPU/GOTEK and CF upgrad

    9.6%
  21. I own a mix of original and modified Amigas.

    12.8%
Multiple votes are allowed.
  1. dalek

    dalek Member

    Joined:
    Jun 27, 2001
    Messages:
    58
    Location:
    Wollongong
    chip mem performance is perfectly fine with a TF536 - as in chip mem speed exactly the same as without
     
  2. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    Using WHDload? Unlikely.

    For the record, chip mem 'speed' doesn't change just because you fitted an accelerator. The problem is the wait states needed to sync a much faster 32 bit CPU with the 16 bit bus.

    The TF536 really shouldn't be that dissimilar to my own ACA1232. Granted, one runs at 50Mhz with 64MB of fast ram vs the ACA1232 running at 40Mhz with 128MB of fast ram (running at an overclocked speed). I can plug my ACA1232 into my A500 and replicate the slow down no problem, as soon as I plug the accelerator into my A1200 the issue completely disappears running the exact same CF HDD and OS install.
     
  3. dalek

    dalek Member

    Joined:
    Jun 27, 2001
    Messages:
    58
    Location:
    Wollongong
    You can't compare the two. An ACA1232 plugged in via an ACA500 in to an A500 is an entirely different beast to a TF536. The ACA500 would have to arbitrate between the ACA1232 and the Amiga because the accelerator doesn't know that it is talking to a 16bit data bus.

    The TF536 (and any good accelerator designed for the A500) should have the same or better performance to chip RAM as it should synchronise to the 7MHz chipset clock domain when dealing with the A500.

    Here are some bustests to prove it. The first image is a vanilla A500 Rev 6 with a 68000. The second is with a TF536 first with caches enabled then without.

    The TF536 doesn't use wait states, when accessing the Amiga in the 7MHz domain it slows to 7MHz and accesses the bus cycle exact. Then you also get the better performance from the 68030 architecture and cache.

    68000.jpg tf536.jpg

    Moral of the story - ditch them apples for some oranges.
     
    Grant likes this.
  4. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    Do you see slowdowns under WHDload running certain games/especially demo's? As I could do those bus tests and I doubt they would pick up the issue.

    WHDload uses wait states. The ACA500 has little to do with it as the Zorro port has full access to all data bus and address lines, the problem is there is half as many on the A500 vs the A1200.
     
    Last edited: Jul 10, 2021
    Grant likes this.
  5. Grant

    Grant Member

    Joined:
    Jan 23, 2002
    Messages:
    1,705
    Location:
    Wollongong
    This is one of the videos I watched demonstrating the slowdown, it's not a nicely edited production and takes a while to get to the point, but shows WHDload games with a plain ACA500 vs. with an ACA1221 and a Blizzard 68030 accelerator

     
    flu!d likes this.
  6. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    The issues he's experiencing under R-Type I never experienced on my own A500 running an ACA500 and ACA1232, the machine never crashed. Having said that, compared to the presenter in the video I'm total ass at R-Type.

    There were some demos that highlighted the problem really well, I just can't remember what they were. The slowdown issue is the very reason I switched to an A1200, since getting my A1200 I've never taken the A500 out of storage, the A1200 just runs everything perfectly via WHDload.
     
  7. dalek

    dalek Member

    Joined:
    Jun 27, 2001
    Messages:
    58
    Location:
    Wollongong
    That slowdown in R-Type II is purely down to OCS/ECS vs AGA performance as it makes heavy use of the blitter. Nothing to do with the accelerator.
     
  8. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    It doesn't happen with the 68000 @ 7Mhz. R-Type is an OCS/ECS title.

    I'm not trying to spread come conspiracy here, the fact is the 16bit design of the OCS/ECS Amiga's runs into problems running software that hits chip ram directly and it's most evident under WHDload where wait states are utilized. It's a very well reported phenomenon that people used to blame specifically on WHDload.

    From memory the Spaceballs- Spasmolytic demo used to highlight it really well whether it be run from disk or WHDload, in a certain part of the demo both sound and graphics would slow down to literally half the speed of a standard A500.
     
    Last edited: Jul 10, 2021
  9. Sledge

    Sledge Member

    Joined:
    Aug 22, 2002
    Messages:
    8,920
    Location:
    Adelaide
    bester and flu!d like this.
  10. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
  11. Sledge

    Sledge Member

    Joined:
    Aug 22, 2002
    Messages:
    8,920
    Location:
    Adelaide
    Being made by 1 of the guys involved with the vampire, yes
     
  12. DonutKing

    DonutKing Member

    Joined:
    Mar 21, 2004
    Messages:
    1,469
    Location:
    Tweed/Gold Coast
    This is going back a while, but when I had an ACA500 I remember running that bustest program and I did in fact notice lower bandwidth with the ACA1232 than without.

    I was mainly playing games and demos and was disappointed to find that in many cases my performance was reduced by the accelerator. I ended up selling the ACA500 and just using a Gotek, and I'm much happier for it.
     
    flu!d likes this.
  13. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    I wonder if I'll get a faster result running the same test on the A1200? Surely I would.

    I'd try the A500, bit then I've gotta pull the accelerator out of the A1200, install it in the ACA500 housing....Bleh, I CBFed really.
     
  14. matto20v

    matto20v Member

    Joined:
    Nov 26, 2003
    Messages:
    299
    Location:
    Perth, Western Australia
    Fluid, I think you're gunna have to back up your theory's and beliefs with a bit of evidence because I'm not sure that you're on the right track here mate I am sorry.
     
  15. Grant

    Grant Member

    Joined:
    Jan 23, 2002
    Messages:
    1,705
    Location:
    Wollongong
    There's plenty of mentions of it on eg. EAB (I don't often read those boards but it showed up in a Google search), but I don't have a way to test it at the moment. From what I gather, the WHDLoad slave code sometimes adds worst-case wait states, and sometimes uses the snooping available in WHDLoad: http://whdload.de/docs/en/opt.html#ChkBltWait, which uses the MMU to detect when it can move on.
     
  16. flu!d

    flu!d Never perfect, always genuine

    Joined:
    Jun 27, 2001
    Messages:
    19,692
    I know I'm right mate. Got both machines here and there's plenty of discussion on the topic. I've experienced and tested it first hand over and over again before dumping my A500 for my A1200.

    The 16 bit bus is a limitation regarding WHDload and faster 32bit processors with a 32bit bus. There's a solid reason why the A1200/4000 are 32 bit designs.

    Anyway, this is pretty cool:

     
  17. bester

    bester Member

    Joined:
    Feb 14, 2002
    Messages:
    977
    Location:
    Adelaide
    This arrived in the mail last week:

    [​IMG]
     
    adz, flu!d and power like this.
  18. power

    power Member

    Joined:
    Apr 20, 2002
    Messages:
    66,933
    Location:
    brisbane
    nice package bro.
     
  19. bester

    bester Member

    Joined:
    Feb 14, 2002
    Messages:
    977
    Location:
    Adelaide
    Upgraded to AmigaOS 3.2 no problems, just chucked in the ROMs and pre-imaged 32GB compact flash card (going for a fresh install this time), copied the Blizzard 68040/68060 libraries over, and then copied all of my data/utils off of another compact flash card (via pcmcia) onto the new one.
     
    power likes this.
  20. bYrd

    bYrd Member

    Joined:
    Sep 14, 2001
    Messages:
    2,287
    Location:
    NE Melbourne
    Can't be arsed going to Amiga OS 3.2 having just put 3.1.4 ROMs into my A1200, especially when it seems like community release they still want to charge for.

    Bring on that PiStorm 1200.
     
    DerekP likes this.

Share This Page

Advertisement: