I bookmarked GIMONS page[1] on this a long while back, as he's the original developer (I note there's also a link to it on this github).
I too am surrounded by SCSI drives, as I'm currently trying to get a bunch of Amiga 500 sidecar SCSI controllers up and running, so GIMONs work was of some interest.
However, for the moment, I've got enough working SCSI-2 drives to be getting on with, plus you can't beat that lovely sound of whirring and clicks during the spin up :)
One day I know I'll have to swap them out for solid state, but for now I'll stick with what I've got.
The best part about projects like these is that they’re 95% useless in the real world but they give a person the chance to connect the dots through so many different layers of hardware and software. I feel like every programmer should be grounded at the hardware/software interface so they truly understand how their code will execute as they move up through higher layers of abstraction. This is a cool project!
So all my old SCSI storage disk boxes and towers are hereby not only obsolete, but disrupted? Phew! :) The retro computing world is not only wide but can also be expansive.
"Interesting, this is a lot cheaper than a SCSI2SD."
SCSI2SD is stuck with a fairly expensive BOM (FPGA, 32 bit ARM SOC) and not enough volume for deep discounts.
Leveraging the Rpi's buyer power does make this a lot cheaper. The adapter board for this appears to be mostly just 4 bus transceivers (SN74LS641) and appropriate SCSI headers/connectors.
Still, it's a pretty weird niche. An SD card is not very reliable for computing use. If I spend this much I'd much rather have something a bit more reliable. Like a "SCSI2SATA" adapter which then connects to an SSD. It would have much better random write performance too I bet. But I haven't been able to find such a thing.
I already have SCSI2SD in my toolbox - but what I really want is the ability to write custom SCSI code for some of these devices, which is what the RaSCSI would provide, in a pretty nice environment ..
You're not paying for the raw components... given the incredibly niche audience for this product and the development effort involved it's really cheap (and it looks like they are even hand soldered).
Can confirm. Small-scale electronics projects (hundreds or thousands of sold boards) are really expensive compared to mass-manufactured stuff sold in the millions. Even if you just charge for the raw component prices, it will still be much more expensive than something that sells at a much higher volume.
Source: Personal experience from making and selling a few hundred units of self-made retrofit PCBs.
I'd happily pay less for an unpopulated board and do it my self, but I don't have the eyesight or steady hand for SMD soldering - a lot of these new 'retro' hardware releases are designed in such a way that a hobbyist has no hope building themselves, and nor can they afford the pre-builts. :(
--
Off-topic: Another couple of downvotes for simply expressing an opinion... HN's echo-chamber tendencies are in full-force these days (again... my opinion).
From my experience (see sister comment), I don't think the board will be much cheaper that way. Even at this low volume they will probably not hand-solder anything but rather rely on a company offering pick&place services, and that will not add much to the costs. If you want to try it yourself and prove that it can be done much more cheaply, you can maybe try building one yourself: https://github.com/fhgwright/SCSI2SD/tree/master/hardware
Yeah, Tindie just doesn't ship to my country, alas. I'll try to find a US address to use, because I'm not going to wait until Tindie catches up with Europe.
Tindie is a marketplace, they don't ship anything (and using them from Europe is not a problem). The specific seller of the product discussed here does ship to Europe too, contrary to your claim, and explicitly asks you to contact them if your country is not set up yet (because they do need to research shipping cost and set it up). What exactly is the problem with that?
I don't see where the seller states they'll ship to me here in Europe, but thanks for pointing out to me that I can contact them - I will do so, and see if it works out.
I do ship outside the US. Tindie doesn't have an option to allow all countries, so I need to add them as requested. Hit me up if your country isn't on the list.
They did compare to SCSI2SD version 5. Version 6 is supposedly much faster: "This board [V6] offers significantly better performance than the V5.x boards by adding support for synchronous SCSI transfer"
This is cool, but I am not really sure what they are doing for this 68k Mac project on top of what the original coder has already done when he created this for the X68000. If you look at the code, they translated his comments from Japanese to English and there are a lot of do-nothing/cleanup commits.
Just that they are testing it on real Macs probably has some value for other vintage Mac folks. Some of the english docs seem to have more than just Google translate applied, like: https://github.com/akuker/RASCSI/wiki/Setup-Instructions
Its not really specific to 68k Macs. This project was forked to add the following:
- The obvious translation of the code to English
- Github repo to allow for easier community contributions (pull requests)
- Create a community-driven Wiki for the documentation
We're definitely not trying "compete" with the original version or anything.
Side note, the hardware is compatible with both versions of the software. The main difference is that the "68k Mac" version uses a DB-25 connector, which is more typical for old Macs.
It surprises me how this kind of device could be emulated over Linux (I know it has the baremetal option), I always see RPi too complex to work like a MCU.
I see a 50 pin connector, so I would guess it's limited to SCSI-2, 5MBps/40Mbps. That seems to be right at the top of the range that you can bitbang GPIO reliably on an Rpi3 (50Mhz).
You're right in that the old rule was that bitbanging was limited to low bandwidth stuff in the past. The advent of cheap, high clock speed and small devices like the Rpi, ESP32, STM32Fx, etc, changed the game a lot. You can now bitbang VGA reliably on really cheap MCUs, for example.
??? complex? The interaction with a vanilla Linux kernel doesn't lend itself to real time applications, but if the SCSI protocol uses hand-shake signals like GPIB, then the timing isn't all that critical.
Sure, a Beagle Bone Black/Blue/Green with its PRUs seems like a better fit, but the RPi is undeniably much more popular and a little cheaper.
There are benchmark results using a X68000. Random i/o is comparatively good (as one would expect), sequential large reads is just about 1MB/s. Still much faster than floppy disks and not too far of the theoretical maximum (5MB/s for SCSI-1).
Because this is not the site for the project. This is the site for a Mac 68k version of the project, where they keep taking the author's code and translating the comments to English and really not much else so far. When the author releases new versions they merge the new code and translate the new comments to English. It's weird to me because SCSI isn't really OS specific, so other than writing some client app talk over SCSI and ask for images to be mounted, I can't make sense of why they are making a "Mac 68k" version.
Well, but the original ST's SCSI interface was special. They even called it ASCI. Which shouldn't surprise as SCSI was standardized only after the ST (and Mac) launched.
Was there every a "correct" connector? I have a stack of old SCSI gear and there are literally 5 different connectors in use. The closest you could get at the time to a standard would be the centronics plug, but I don't think that would even fit on the back of a Mac Plus, at least not without adding expense in the form of a daughterboard or extension cable.
Steve Jobs had left Apple by the time they added SCSI anyway.
This lets you connect a raspberry pi as a SCSI device. The Pi does emulation of the SCSI interface by bit-banging the GPIO, and exposes virtual drives that are mounted images from the Linux FS on the pi.
I too am surrounded by SCSI drives, as I'm currently trying to get a bunch of Amiga 500 sidecar SCSI controllers up and running, so GIMONs work was of some interest.
However, for the moment, I've got enough working SCSI-2 drives to be getting on with, plus you can't beat that lovely sound of whirring and clicks during the spin up :)
One day I know I'll have to swap them out for solid state, but for now I'll stick with what I've got.
---
[1] http://retropc.net/gimons/rascsi/index.html