Indivision AGA Mk2 CR and 4x IDE Buffered Interface Fitment

Caution: Non registered users only see threads and messages in the currently selected language, which is determined by their browser Please create an account and log in to see all content by default. This is a limitation of the forum software.
Don't Panic. Please wash hands.
  • HI there,


    I'm looking to add one of these to boost my IDE speed:


    (link removed - questionable content with dead links to downloadable software - Jens)


    The card appears to occupy the same space as the Indivision and I'm not sure if it will be possible to have both cards installed without one interfering with the other.


    Do you happen to have any guidance on how best to accommodate both cards inside an A1200 case?


    Thanks in advance,


    AndyC

  • Do you happen to have any guidance on how best to accommodate both cards inside an A1200 case?

    Sorry - I don't know the product, and to be honest, I'd look at an accelerator with built-in IDE speeder, as that will give you even more speed. That AmigaKit interface appears to be the same as IDE-fix express (I developed that back in 1998), which has a "natural upper limit" around 3MBytes/sec, with exceptions of synchronous accelerators that may get near the 4MB/s area.


    An accelerator with built-in IDE speeder, such as ACA1233n or ACA1221lc, will give considerably more performance, as they are not tied to the 14MHz bus of the A1200 any more. The ACA1221lc is in stock, and we're about to release a few 40MHz ACA1233n from safety stock later this week.

  • HI Jens,


    Thanks for coming back so quickly.


    I am now using a Blizzard 1260, having previously used your excellent A1233n card, which I have since sold on (I was unaware this had functionality to boost IDE speed). I couldn't pass up the opportunity to move to the 060, however, having been offered one for a very good price by a friend. I'd consider adding the Blizzard SCSI module and SCSI2SD adapter, but I'd be looking at around €300 for that combination.


    There is a Gayle adapter that can be added to the Amigakit board to boost speeds to 7Mb/s async amd 10mb/s sync respectively. It's considerably more cost effective than the SCSI route, so the issue really is whether or not it will fit with the indivision installed.


    I appreciate you can't answer this question so I'll see if any users have posted in other forums about this, or reach out to Amigakit directly.


    Thanks,


    AndyC

  • (I was unaware this had functionality to boost IDE speed)

    This was added in summer 2019. An upgrade is available for those cards that were shipped before that.

    There is a Gayle adapter that can be added to the Amigakit board to boost speeds to 7Mb/s async amd 10mb/s sync respectively.

    I thought you were referring to the combination with Gayle adapter only - the IDE interface itself will not change anything in terms of IDE speed. It simply can't - and if it does, it's only software, and as I found out, the download links of the supplier are dead, so I don't want to link to it at all. For the IDE-fix software, just go to Aminet, it'll also work on a normal A1200 without additional interfaces.


    That said, the 7MB/s figure is an absolute theoretical maximum that is so theoretical that you could already argue this figure to be false advertising. 10MB/s is plain impossible on a 16-bit wide 14MHz bus. Back-of-the-envelope calculation: An access takes 4 cycles at best, and you transfer two bytes per access, that's 7MB/sec. There is no hardware on the combo you're looking at that can speed up to anything beyond that.


    However, you will lose cycles here and there - most notably when accessing the target memory, which happens on every other word-access. Even if that's fast, you'll miss one 14MHz board-cycle, so two consecutive IDE accesses plus the 32-bit wide memory access will take 9 14MHz cycles, resulting in 639ns duration for a 4-byte transfer. If you assume that this can be kept constant (best-case with longword-alignment of target buffer), you end up with a bast-case performance of 6.2MBytes/s.


    The bad news is that you have a Blizzard 060 card, which loses two additional 14MHz cycles per IDE access, which is most likely the reason why you want to speed up your IDE in the first place (I assume you currently get around 1.3-1.5MB/s). Adding these two cycles to the two 16-bit IDE accesses, you're at 13 cycles (=928ns) for a 32-bit word or a best-case performance of 4.3MB/s. Again - this is "on a good day" when the target buffer is longword-aligned and may be lower in 75% of the cases, where the buffer is shifted by 1, 2 or 3 bytes.


    Still, this 4.3MB/s figure won't be reached in real-life, because the transfer itself is only half of the time that's spent in data transfer. The other half is the CF card or HD that needs to gain access to the requested block before it can be transferred. For really fast CF cards, this time may be shorter (Lexar cards appear to be really fast in that regard), but the general rule of thumb is that total transfer time is "half search, half transfer".


    With the high degree of similarity between the interface combo you're looking at, and the IDE-fix express that I made in 1998, I can say that the general rule of thumb of that approach is that the IDE speed you're currently getting is about doubled. The best speed that an accelerator has ever achieved with an IDE-fix express was a 28MHz synchronous 68030 card from Neuroth, which got almost 5MB/s with the fastest HD that was available at the time (an IBM DTTA if I remember right). This one had a great cache management and still serves me as a HD to benchmark best-case scenarios of my IDE performance: It goes beyond 10MB/s with the 55MHz version of the ACA1233n and activated IDE speeder.


    The Blizzard 1260 with IDE-fix express never went considerably beyond 3MB/s with any harddrive, so I'd expect it to be around 2.2-2.5MB/s with a CF card (CF cards don't have cache, so physical harddrives have an unfair advantage here).


    So although I can't offer any product as an alternative, I just want you to not be disappointed by the solution you've been looking at. The supplier is in Europe (for another six weeks...), so you're covered by EU customer protection laws, including the right of revocation within 14 days after arrival of the parcel at your place. If you're willing to risk the shipping cost, you could post some real-world data here and confirm (or falsify) my assumptions.

  • Thanks Jens,


    I think for the very minor bump in performance this might give me, I'll just leave it as-is.


    I really appreciate the detailed response :-)


    Looking forward to seeing the ACA1260 in the wild once it's available, although I don't think I'm ready to replace the Blizzard 1260 just yet.


    Cheers,


    AndyC