That's my wording, not theirs. It seemed close enough to be acceptable at the time. (Technically it is not even an SO-DIMM but a module in the form factor of an SO-DIMM.)
Yes true, sorry I’ve explained wrong: what I wanted to say is that is a very interesting system, but in order to be a full compact SoC like what we mean now, it should had only a GPU.
I suspect it’s your use of ‘GPU’ that’s causing the confusion. I would expect a GPU to have some form of 3D acceleration or compute capability and I don’t think that’s really applicable to the EGA / VGA video chips of the 386 era.
I could be incorrect though, I’m not sure there is a settled definition of GPU.
Yes my mistake because also now we call “GPU” some cores related to graphic processing inside a complete SoC, but is not really a GPU, only a “graphic core/s”.
What I wanted to say is that in order to “output something directly to a display like a contemporary SoC, this 386 system lacks only a graphic unit”. Maybe in this way is more correct.
The chip that would read RAM and generate a video signal was called a CRTC or RAMDAC back in the day - though RAMDAC, standing for RAM Digital to Analog Converter, only really applies to VGA which is analog. Some game consoles had an equivalent chip (Brooktree/Conexant) that was called an encoder.
This is separate from anything that blits textures, rasterizes triangles to video RAM, performs parallelized math operations on big vector lists, runs shader code, or does MPEG/etc. decoding. All of that stuff is what constitutes a GPU.
I'm not sure what the modern term for DAC is given everything is all digital.
The 3dfx Voodoo1 card _might_ be classifiable as the first GPU, I'd say. While SGI had display hardware with more processing before that, it wasn't (afaik) in a form that could be moved to a different system.
I understand by GPU you meant VGA, but either way a lot of SBCs come without a VGA/GPU. It’s x86/PC quirk that everything comes with a VGA compatible framebuffer.
For the ones among us who are not familiar with foone yet:
They only post on Twitter because blogging doesn't work for them [0]. The Twitter thread UI sucks and that's unfortunate but somewhat unavoidable in this case. I still prefer it to not getting any of foone's "weird hardware" discussions at all.
You might have to click "Show thread/more" if you do not see all of the posts. Or just run the post through a website like threadreaderapp, it will collect follow-up posts by the same user and compile it into a page similiar to a blog post.
It hurts how much that resonates with me, especially the constant editing. I'll probably be editing this post for two hours and attempt to tomorrow but then realize that it's too late.
Yeah, every post having an image doesn't help, but the same username/avatar repeated so much is a lot of noise too. If the point of a thread is 'microblogging' then it really needs third-party replies removed, and not such repetition.
It's not making you look at the thread through a three post window where if you don't change tweet every so often you end up seeing replies rather than the continuation?
You could use their timeline, but that presents different challenges, i.e. the order is reversed and other content is intermingled.
Tweetdeck works well if you use lists in Twitter. It allows you to create a column for each list. Each column is pretty information dense, and with lists I can have a local news column, a web dev column, etc.. whatever your interests are.
I wrote a little script, let's call it "twt", using stunnel and netcat to retrieve all tweets for a user and print a simple HTML blog page: https://news.ycombinator.com/item?id=28071491
(only first 50 lines shown for brevity; URLs are hyperlinks)
Fri Feb 15 03:51:09 +0000 2008
https://ko-fi.com/fooneturing
https://t.co/RWnHtUISyM
https://floppy.foone.org/w/Main_Page
https://t.co/0gnaVwARmR
Thu Aug 05 05:29:51 +0000 2021
1423154262276665349
https://twitter.com/Foone/status/1423154262276665349/photo/1
https://t.co/JiAIJaORFt
https://twitter.com/Foone/status/1423154262276665349/photo/1
https://t.co/JiAIJaORFt
I saw a sign that said not to take photos. So I took a photo of it
https://t.co/JiAIJaORFt
Fri Feb 15 03:51:09 +0000 2008
https://ko-fi.com/fooneturing
https://t.co/RWnHtUISyM
https://floppy.foone.org/w/Main_Page
https://t.co/0gnaVwARmR
Thu Aug 05 04:16:42 +0000 2021
1423135852612317184
https://twitter.com/ASovietOnion/status/1422672792860954627/photo/1
https://t.co/alyaBjyMfc
https://twitter.com/ASovietOnion/status/1422672792860954627/photo/1
https://t.co/alyaBjyMfc
RT @ASovietOnion: why did i make this https://t.co/alyaBjyMfc (RETWEET)
Fri Apr 19 09:27:10 +0000 2019
http://bing.com
https://t.co/nzJr38i0mG
Tue Aug 03 21:36:40 +0000 2021
1422672792860954627
https://twitter.com/ASovietOnion/status/1422672792860954627/photo/1
https://t.co/alyaBjyMfc
https://twitter.com/ASovietOnion/status/1422672792860954627/photo/1
https://t.co/alyaBjyMfc
why did i make this https://t.co/alyaBjyMfc
Fri Feb 15 03:51:09 +0000 2008
https://ko-fi.com/fooneturing
https://t.co/RWnHtUISyM
https://floppy.foone.org/w/Main_Page
https://t.co/0gnaVwARmR
Thu Aug 05 04:14:47 +0000 2021
1413694652822163459
https://twitter.com/Foone/status/1423135371139837954/photo/1
Wow, that is interesting. You can almost tell if he gets up between each Tweet and walks around to filter out his thoughts in order to push it into one single new tweet. There might come a new Tweet or not, each one expands the thought of the chain, but the chain by itself wouldn't be broken if he'd just had stopped tweeting at any point.
A very attractive "blogging model".
Up to this point the 1/4, 2/4, ... tweets were what defined my aversion to these chains.
It's like when you remember one thing after the fact and get annoyed that it's too late, but in his case it's never too late.
I do not mind these at all. Am I really auch a minority or are people who don‘t like threads just very loud?
I like the reading cadence, I like reading in the twitter mobile app and I like the ability to precisely reply to things or expand in a quote RT. Seems great to me, but I also have ADHD so maybe that‘s it.
Personally I'm not a huge fan because of how spaced out the text content is. So many images breaking the flow of thought. Id rather blocks of text and then a gallery of images. But you get what you get. A string of tweets is better than nothing.
It seems like it‘s really a "type of brain" thing. I think I would even appreciate an alternative browser reader mode that puts normal articles in this style, only on a phone though.
Personally I don't really mind the broken-up content but I do mind Twitter as a platform. It does tracking and it tries to constantly push me to make an account and/or use their app. I definitely won't use tools to make it more usable either. It's not about usability, it's about Twitter itself.
But it's fine, it doesn't have to be for everyone.
I have resorted to just using my own nitter install for all things twitter. Here's an example from a public instance [1]. Of course this is read-only but that's all I want, mostly to check in on @mntmn. I love the interface, it fixes everything I hate about twitter except the shitty dynamics and the fact that it is twitter.
It's dead simple to host as a docker instance by the way, and with privacy redirect [2] it's pretty seamless.
It does make me wonder if there is a need for personal microblogging software that presents to the world a bit better for reading. I notice we didn't get much uptake on our attempt at blogging for folks. Yet they can make quick Facebook posts easily. I do wonder how much friction a title has on sending out your thoughts.
Not w.r.t Fone but the input side of the threaded tweets are being addressed by app(web) like typefully, it splits long post into tweets and it does it well.
Reading side of threaded tweets require invoking bots like threaderapp, I'm not sure whether there's enough need for an app which renders threads as a post directly but I'm trying to find it out[2].
Or they could write short, tweet-style posts on a platform like micro.blog so they can still write in a way that aligns with the way their brain works while not having to deal with the way Twitter doesn't (and micro.blog will syndicate to Twitter for them so they needn't lose the Twitter audience).
I strongly suspect the real reason is they just doesn't care enough to spend time trying something different. And that's totally fine! They have absolutely no obligation to change just because a bunch of folks dislike reading tweet threads.
But there are absolutely alternatives out there if folks are interested in checking them out.
Please don't introduce this off-topic issue into the threads - it's a classic flamewar topic and extremely predictable, and we should just all avoid it.
This is an embedded/industrial PC; SODIMM is a convenient connector form factor when you need a really large number of pins. Old Pi Compute modules also used SODIMM: https://www.raspberrypi.org/products/compute-module-1/
A unique part of the product offer is the longevity commitment - they agree to provide identical replacements for years. These things end up in tills and machine control systems all over the place.
Yup, very good summary. I'll add: advantage of this design, is that you only need a SODIMM connector on your breakout board, which is very easy to source. Also, you can have a breakout board tailored to your needs, relatively cheap (e.g less PCB layers), with less care for radio design (most of it is done on the module), etc.
It's mostly advantageous for low-quantity, high margin embedded systems, with modules supported for years.
AFAIK PC104 is meant specifically for this. Re-using SO-DIMM (a connector meant for something entirely different) being a conventional practice seems slightly unusual however. Like if somebody was using NEMA 5-15 connectors for something like USB.
I suppose that the selling point may be that your custom DOS software from 30 years ago will still work. If said software is careful enough to only do video output via BIOS, it will even work without a video card, with the dot-matrix LED screen.
The thing can be retrofitted to various machines that were originally controlled by a PC board, but you can't get a replacement 80386 board any more. Control via the parallel port was pretty common.
If you work with embedded hardware, this isn't exactly strange or weird.
Even before the Raspberry Pi compute module was a thing, there were already many vendors around (e.g. Kontron, Phytec, Toradex, ...) who sell SO-DIMM form factor modules with the entire system on it and a breakout board to plug it in to. AFAIK the usual deal is that you get the design files for the breakout board, customize it for your application and simply plug in the vendor sourced module (semi-relevant XKCD: https://xkcd.com/1053/; given that the RPI compute module is a thing for some years now I considered this something that "people know by now").
To be honest, I've never seen one with an x86 SoC on it tough, instead of something ARM based. I would consider that strange or weird, but having bought a BifferBoard[1] way back when, I don't find it that surprising that x86 hardware of that size exists.
So, they only use the physical connector not the … uhm … transfer protocol associated with memory modules? Tbh, at least with memory, I always conflated both.
If there are multiple vendors around SO-DIMM SoCs, is there a common standard for how they are using the pins?
> If there are multiple vendors around SO-DIMM SoCs, is there a common standard for how they are using the pins?
Not really. If you want a common standard with multiple vendors, look at COMexpress and Qseven, with the former sporting everything up to Xeon CPUs and the latter low-power (12W TDP, iirc) only.
> is there a common standard for how they are using the pins?
None that I'm aware of. Considering some pinouts I've seen my guess is that different vendors simply assign them as they see fit. Also, especially ARM SoCs have widely varying capabilities, and different SoMs have all kinds of different hardware on module/off module, different bus systems they support, and so on. Also, some SoCs can change some of their pin assignments to a degree. Given all that and considering that the whole purpose is usually to cheaply develop semi-customized embedded hardware for specific applications, there might not be much demand for that either.
One thing about the memory bus is that it is a perfectly good bus for devices. You absolutely can put something like a computer into a memory socket and use it as a device. You simply latch onto a particular memory address that's not going to be used for memory and whenever you see that you know the memory bus is for your device.
I don't know if this is merely form factor compatible or an actual device for a SO-DIMM memory socket but there really isn't anything stopping you making a DIMM socket peripheral device except the need for external power.
Huge difference between memory bus in front of memory controller vs behind it. You cant just plug yourself into a DIMM slot and act like a device on a memory bus. You will have to perfectly emulate particular ram supported by your memory controller (protocol, access patterns, timings), and this will be strictly a slave affair.
reply