1233n to 1234 upgrade

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.


Also users that are not logged in can not create new threads. This is a, unfortunately needed, counter measure against spam. Please create an account and log in to start new threads.

Don't Panic. Please wash hands.
  • Quote from Jens

    If it's one chip only, it is not buffered.

    Possible. However, it is sold as 4XEIDE'99 BUFFERED INTERFACE (WITH ALLEGRO CDFS). Supports 4 devices, has 2.5" and 3.5” ports. One large chip. PIO4+ mode ona ACA 1233n 55 works fine.


    Quote from Jens

    Here's a first glance at the prototype we're currently working on:

    I have a one or two questions/suggestions:

    1. I would still suggest to make CF card accesible from outside (as provided by new A1200 cases)

    2. For people who replace RF modulator with Indivision HDMI/VGA port it might be tight fit as is (because of connecting cable).

    3. On the side note: would it be possible to get separate VGA and HDMI ports for Indivisio? These days it is very easy to get replacement back plate with VGA in place of RF.

  • However, it is sold as 4XEIDE'99 BUFFERED INTERFACE (WITH ALLEGRO CDFS).

    That's false advertising, but Elbox obviously doesn't care about EU regulations and honesty towards customers. Any product description is right as long as it lures customers into paying. I'd send it back for a refund.


    On the technical side, the signals that need to be buffered are:


    - A2-A4 (three lines address)

    - D16-D31 (16 lines data)

    - IOR/IOW (two lines control)


    ..and in order to properly control driver on/off, you also need the two select-lines as input.


    So if you wanted to make a single-chip solution, you would need at least two pins for power supply and another 44 pins just for connecting the signals. Just count the number of pins on that IDE'99 chip: It's 28, so it can't possibly be a buffered interface. It buffers a fraction of the lines, but not the ones that really MUST be buffered (the address and data lines). Even worse: The chip only has 10 pins that can be switched to output, two pins are power (doh!) and 4 of the 28 pins are not usable at all. How would you expect this to fully buffer the bus? From pictures in several shops you can see that the data bus is just wired straight through to the 3.5" ports, clearly saying "not buffered" without even buying the thing.


    While the technical side is leaving no doubt, it may actually be OK on the legal side, as there is no law that says "buffered" needs to apply to the whole bus. Buffering a single signal may be OK in order to legally call it "buffered". It is still misleading customers, believing that it may solve a problem that is caused by the IDE port being completely unbuffered in A600/A1200 computers.


    There you have it: You have been tricked. It's like being promised a portable light-weight 40-piece survival toolkit, and getting a box of matches.

    I have a one or two questions/suggestions:

    1. I would still suggest to make CF card accesible from outside (as provided by new A1200 cases)

    I have decided against that, as PCMCIA-CF adapters are widely available and the PCMCIA port is already hot-plug compatible. So it would not solve any problem, but only add complexity, which you don't want in such a simple product.


    2. For people who replace RF modulator with Indivision HDMI/VGA port it might be tight fit as is (because of connecting cable).

    Maybe - the next iteration of this prototype will look slightly different, making it mass-producable, easier to install (="harder to install the wrong way"), and with that, the whole board will be raised about 6mm higher than this prototype. Further, the part of the board that is so very close to the TV modulator will be shorter by almost 2mm. As a result, you'll have more space in two dimensions - though I need to check myself where the PCB of the MK3 output board would be.


    3. On the side note: would it be possible to get separate VGA and HDMI ports for Indivisio? These days it is very easy to get replacement back plate with VGA in place of RF.

    If "separate" means that you want a board like the one that's included, minus the "other" connector, then this is possible, yes. However, from a "distance to be bridged" point of view, and of course the soldering effort, it may be better to design something that uses the space of the two RCA-audio connectors for the HDM! port. That's a whole different discussion, though, and I tend to think about this more like you do about people not wanting to use network storage - soldering on the old, precious machine is not an everyday-thing, most likely only for a small fraction of users.

  • Anyone reading here who actually has problems with a CF card after the IDE speeder has been switched on?

    I'm reading, but my interest it this product is not related to the IDE speeder and your accelerator. I actually have Blizzard 1230. But I sometimes experience some instabilities on the IDE port in some specific scenarios that might or might not be related to the fact that IDE port on A1200 is not buffered. I want to buy (once when it will be available) this buffered CF IDE interface regardless if it will help me or not. At least I will know that this part is OK and rock solid (hopefully). I can discuss those scenarios that I have a problem with if necesary, but will not go to details right now since you are probably looking for people that have exact problem with IDE speeder and your cards.

  • Yup, I do. I get random read/write errors and even occasional reboots if the IDE speeder is active, only after an hour or two, but they don't happen at all if I have it switched off, so I think it's my current not-so-great cf->ide adapter causing it, or at least contributing to the issue.

  • Yes, as soon as I switch it on it starts to create read errors. Than CF is not visible anymore. After power cycling it comes online again and I let file system fix it. Any speed (PIO mode) creates such problem. It works with buffered interface(s) and HDD (sata to pata adapter).

  • IDE interfaces are from AMIGA kit (one that you said is not properly buffered because it has one chip only) and one from e-bay (with 5 chips). HDD was working on both on short cable. However, on the first one not proprly on long cable. I am not sure about CF card it is been rebranded 32GB saying Amiga Forever.

  • I'd rather not include that in any of my tests - that's at the user's risk.

    Now, I have done some more testing:

    1. I have acquired SanDisk Extreme 64GB CF card.

    2. Under OS 3.2 it works w/ speeder enabled w/o buffered interface

    3. Speeds with PIO 4- and 4+ are the same (around 2.8)

    4. Speeds are around 3.3 w/ buffered interface and w/ speeder enabled


    Also, I have acquired M. type SSD and put it in 2.5" IDE adapter. It works via buffered interface BUT only in PIO 4- mode. When 4+ ia activated it reports errors but NOT read/write errors, just file execution. Speed achieved is same as CF card.


    QUESTION: during disk partitioning there is a field with max transfer rate. By default it is 0xffffff but only one value is mentioned on internet as safe: 0x1fe00. Does this string have actual impact when modern hardware is used (e.g. CF cards, SSD's, moder 2.5" SATA HDD's etc.)? I have been seeking proper information for ages but to no avail.

  • 3. Speeds with PIO 4- and 4+ are the same (around 2.8)

    4. Speeds are around 3.3 w/ buffered interface and w/ speeder enabled

    Does not compute. The interface - buffered or not - does not change the access speed. Yes, the IDE-fix express (with the express-adapter installed and activated) does change the access speed, but that should NOT be combined with the IDE speeder of the ACA1233n. My guess is that there was a difference in the measurement setup, such as a different MapROM setting or burst being toggled between measurements.


    QUESTION: during disk partitioning there is a field with max transfer rate. By default it is 0xffffff but only one value is mentioned on internet as safe: 0x1fe00. Does this string have actual impact when modern hardware is used (e.g. CF cards, SSD's, moder 2.5" SATA HDD's etc.)? I have been seeking proper information for ages but to no avail.

    MaxTransfer is the maximum size that the driver may request from the physical device in one go. 64k is the magic value, which has been misunderstood as "64k bytes" during the first Amiga years. Since the goal is to avoid an overflow of a 16-bit counter in the harddisk, the value chosen was always "minus one", so you will see a lot of references to 0xffff or 0xfffe. The latter following the idea that the value should be even to not cause unnecessary splits of words.


    Later on, people discovered that it's a 16-bit counter for words, not bytes. This increases the magical value to 0x1fffe.


    The next iteration is to make sure you're not ripping a file system block apart. Most Amiga file systems use 512-byte blocks, so 64k words minus 256 words for one block would be a max transfer value of 0x1fe00. With the above information, this is finally, correctly, accepted as common knowledge.


    With this going really deep into the inner workings of drivers and harddisks, it was Oliver Kastl (of IDE-fix fame) who was the first to completely ignore the MaxTransfer value, and made sure that 64k words would not be exceeded in any low-level transfer. This made it a no-worry-setting, reducing support work or data loss due to a false setting down to zero; that was about 25 years ago. Not sure if the latest versions of scsi.device are implementing the same idea - IMHO, they should.

  • What is the current status about this project?

    Small delay, as my back was killing me since about a week b4 Xmas (now fixed). I know the first few units are complete and ready to be picked up at the assembly company, just need to find time...

  • The last reply was more than 365 days ago, this thread is most likely obsolete. It is recommended to create a new thread instead.