The digression 77 When AI eats the graphics cards

Format	size	Download
MP3 Audio	34 MB	Download
AAC m4a	18 MB	Download

transcript
00:00 - 0:00:02 Music
0:00:02 – 0:00:31 [Music]
0:00:31 – 0:00:39 Hello and welcome to a new release. I'm Sascha Markmann and I say
0:00:39 – 0:00:47 Hello there. First of all, please excuse my somewhat convoluted voice, but I have the
0:00:47 – 0:00:56 I've had a little bit to do with cold symptoms in the last few days or something like that. You know it
0:00:56 – 0:01:09 It's not. Yes, today's theme, yes, this is initialized by graphics cards. I have
0:01:09 – 0:01:17 in November the last time for me personally got a graphics card for my computer and
0:01:17 – 0:01:25 I paid, let's say, $440. And if I were to buy the same card now,
0:01:25 – 0:01:36 I would pay well over 600 or almost 700 euros for it. This is the current storage crisis.
0:01:36 – 0:01:41 thanks. It is just as good with memory, which has become up to 800 percent more expensive,
0:01:41 – 0:01:49 Especially if you want faster memory. Or SSDs, i.e. these electronic
0:01:49 – 0:01:56 Hard drives. They have all become more expensive. Even rotating hard drives have become more expensive.
0:01:56 – 0:02:08 And all because almost a million of the good euros worldwide are invested in AI data centers.
0:02:08 – 0:02:15 The market is almost empty. I had now read that Samsung, I think, for the year
0:02:15 – 0:02:24 26 the total annual production is sold. That's what they reported, and some have even
0:02:24 – 0:02:31 He had already sold the entire year's production for 27. You can imagine,
0:02:31 – 0:02:37 what this looks like for the poor little end consumers and not for you any AI
0:02:37 – 0:02:44 Hardware for where a graphics card costs beyond 10,000 euros. I would like an Nvidia.
0:02:44 – 0:02:53 H100. I'd really like to. This will give my projects a huge boost.
0:02:53 – 0:03:04 Not the electricity costs. Oh, my God. A little discussion about the current situation. This
0:03:04 – 0:03:09 Whole thing got me thinking about it. I went to my first computer.
0:03:09 – 0:03:25 So I had some before, so Schneider PCP or CPC, CPC 464, so it was called. A fantastic
0:03:25 – 0:03:30 Machine with a great processor Z80, you could really do much more than just
0:03:30 – 0:03:39 Playing the games. No one has. Then I also had a C64 from Commodore, which was then
0:03:39 – 0:03:43 Mainly to play. On the other I actually programmed and there I damned
0:03:43 – 0:03:52 Learned a lot. And that was so with, yes I started with ten and that was 87. I am a vintage
0:03:52 – 0:04:00 77 and thus the bow is struck. The 77 episode, the digression, year of birth 77. Hey,
0:04:00 – 0:04:08 great. Small side-fact for the data collectors among us all. In any case, I have,
0:04:08 – 0:04:20 That was Christmas 92, so just before I turned 15, I got from my parents,
0:04:20 – 0:04:26 by my aunts and all other relatives for Christmas and the birthday I later
0:04:26 – 0:04:36 I got a PC. A 386 SX25 megahertz with two megabytes of memory,
0:04:36 – 0:04:45 a 100 megabyte hard drive and now there is a 512 kilobyte graphics card. So one
0:04:45 – 0:05:00 Half a megabyte. Not gigabytes, megabytes. 1024 megabytes is a gigabyte. That's so tiny
0:05:00 – 0:05:09 Tiny small compared to the 8 gigabytes and 16 gigabytes monsters of today. Or stop when it
0:05:09 – 0:05:20 Hold an 80s or 90s is with 24 and even more gigabytes of memory. Yeah, and that thing was great.
0:05:20 – 0:05:33 You have to say that there were SX and DX back then. DX was a 32-bit processor with a 32-bit bus. The SX
0:05:33 – 0:05:40 had stop internally 32 bit but stop externally a 16 bit bus that has made a bit slower
0:05:40 – 0:05:50 The hardware is cheaper. Nevertheless, the whole thing at that time cost 1300 DM and was of course a
0:05:50 – 0:06:00 Damn great gift. That's why Christmas and birthday and I was so happy so happy
0:06:00 – 0:06:09 Of course, the set consisted of a keyboard, a mouse and a monitor with this beautiful
0:06:09 – 0:06:20 Midi tower case in grey, as usual at the time. And then I continued with that. These
0:06:20 – 0:06:27 100 megabytes of hard drive that was in there, that was incredibly huge. Even my aunt and
0:06:27 – 0:06:34 My uncle, who was at university and was writing her doctoral theses at the time
0:06:34 – 0:06:39 They had 40 or 80 megabytes and they were all jealous.
0:06:39 – 0:06:45 And then they heard sayings like, "This will never be full, you will live forever."
0:06:45 – 0:06:46 I'll deal with it.
0:06:46 – 0:06:57 Have you ever seen a young teenager who has a computer, access to all
0:06:57 – 0:07:04 possible programs, apart from now Works and then Harvard Graphics or something,
0:07:04 – 0:07:12 which he has installed, side games, also programming environments, so C++ and so on.
0:07:12 – 0:07:19 Have you ever seen this before? This thing was always full and I just went with it.
0:07:19 – 0:07:24 Had fun. And if I put this in relation to what I have today,
0:07:24 – 0:07:39 an Intel i9 13900K, 64 gigabytes of RAM, 8 terabytes of SSD, where three pieces are MVMEs
0:07:39 – 0:07:49 and a SATA, as a data tomb. You have to think about the fact that the processor is between 10,000 and 10,000.
0:07:49 – 0:08:00 and is 40,000 times faster and 32,000 times more memory or that's so incredible
0:08:00 – 0:08:09 84,000 times more disk capacity. You have to think about it and when you think about it,
0:08:09 – 0:08:15 that lasts from one hard drive to another hard drive in this 386 it up to 14 minutes
0:08:15 – 0:08:26 took a gigabyte to copy and two seconds at the sata ssd or 0.15
0:08:26 – 0:08:37 seconds at my samsung mvm ssd this is mindblowing this is 5800 times
0:08:37 – 0:08:44 Faster shit and all that in these few years
0:08:44 – 0:08:48 Let's see what 30 years have been like in this kind of technological development.
0:08:48 – 0:08:52 And that's the processor I have now, which is already a bit older.
0:08:52 – 0:08:54 There are newer generations.
0:08:54 – 0:08:56 I believe two.
0:08:56 – 0:09:02 I mean, I only deal with processors and motherboards when I need them new.
0:09:02 – 0:09:08 I mean, I often buy new graphics cards because I need the performance there, but not so with the processors.
0:09:08 – 0:09:14 Of course, I also started programming with GW Basic. It doesn't matter now,
0:09:14 – 0:09:20 This is a very simple programming language. Basic already existed on the C64 and on the tailor,
0:09:20 – 0:09:26 That's why I knew that. And there I have an almond bread, which is so the best known
0:09:26 – 0:09:34 Fractal at all, made then. And I had the SX back then and that's up to
0:09:34 – 0:09:39 90 minutes, full screen. Because it's always too long for me
0:09:39 – 0:09:44 It took me a long time to get a mathematical co-processor, because
0:09:44 – 0:09:51 the processor had at that time, that had all up to the 486,
0:09:51 – 0:09:57 All of them did not have higher mathematical functions installed in the
0:09:57 – 0:10:03 Processor. So a processor can usually only be a plus computer, so
0:10:03 – 0:10:09 Addition. And then there are so many different tricks and intermediate steps, with which you can stop by the
0:10:09 – 0:10:19 Addition of a complement, hahaha, nerd knowledge, a subtration, a division can make. Because
0:10:19 – 0:10:25 Multiplication is just addition and then just so many steps.
0:10:25 – 0:10:31 But it's just about this one complement, so that's by keeping the smallest possible
0:10:31 – 0:10:36 unit, which you put on the right hand and then you make such an intermediate step and
0:10:36 – 0:10:44 Then you can make a subtration out of it. This is only possible with binary numbers. Mind you,
0:10:44 – 0:10:53 Only with binary numbers, not with our numbers, i.e. with a decimal system. A long speech,
0:10:53 – 0:11:03 short sense. After the installation of this mathematical co-processor, the whole thing did not take 90 minutes
0:11:03 – 0:11:11 more, but a maximum of twelve. It was mind-blowing. Oh my God, it was so fast.
0:11:11 – 0:11:18 You could almost watch how line by line was painted and not so pixel by pixel.
0:11:18 – 0:11:29 Boah crass! Yeah, that's how it was back then. And if you think about it now, today, the hardware is in the
0:11:29 – 0:11:40 Compared to then but become cheaper and much more efficient with lower
0:11:40 – 0:11:45 Electricity consumption. And even if I think of such single-board computers as a Raspberry Pi or
0:11:45 – 0:11:55 So go get it. They're as powerful as my first computer. me
0:11:55 – 0:12:03 But I also have to say that at that time I always, even if I bought the co-processor
0:12:03 – 0:12:07 I don't have it right now, but once I've gone and I've got an update.
0:12:07 – 0:12:16 From a 386 SX to a 386 40 DX. Then I have the same for it.
0:12:16 – 0:12:21 mathematical co-processor in order to be able to make these calculations faster.
0:12:21 – 0:12:29 Because I was already a very strange young man at that time, who was next to the whole
0:12:29 – 0:12:37 Read and besides all the programming and next to the music and meet up with friends.
0:12:37 – 0:12:46 Yes, even outside I programmed hardcore. And even with one
0:12:46 – 0:12:54 Buddy together. We used to sit, sit, sometimes sit on the floor.
0:12:54 – 0:13:04 one has read the listings aloud, so 10, go to, here and there, address anyway 20,
0:13:04 – 0:13:09 go sub 10, if and so on. So totally the blatant instructions. I remember that,
0:13:09 – 0:13:14 we sat there again and then hacked the thing in, the program didn't work and
0:13:14 – 0:13:19 you had to go through line by line again, look where the code is, because at that time it was still
0:13:19 – 0:13:27 There was no debugging routine, but only in the higher languages for programming. And
0:13:27 – 0:13:34 It was such a beautiful world then. And you just have to say, there are these
0:13:34 – 0:13:38 the intermediate generation. So there's not only the baby boomers and then comes Generation X,
0:13:38 – 0:13:46 but it is said that at the end of the baby boomer and at the beginning of Generation X, there is an intermediate generation
0:13:46 – 0:13:59 There are between 76 and 83 and that would be the Sennials. And that's exactly where I'm in it, because I actually had
0:13:59 – 0:14:05 This analog childhood and digital adolescence, and I've seen it all. I remember,
0:14:05 – 0:14:13 How cool that was when I bought my first modem on such a computer exchange. Da
0:14:13 – 0:14:19 Since you could not always buy the latest hardware, but they are extremely cheap.
0:14:19 – 0:14:23 And that was a modem, you can't even think about it these days.
0:14:23 – 0:14:28 I get a higher data rate through my amateur radio with a modem
0:14:28 – 0:14:29 than then.
0:14:29 – 0:14:33 2,400 bits per second.
0:14:33 – 0:14:39 When you entered a mailbox, you could watch how
0:14:39 – 0:14:45 All these movies from the 80s, when there are characters for characters on the monitor.
0:14:45 – 0:14:49 It had been the same.
0:14:49 – 0:14:55 And how I dialed into a mailbox for the first time, which was just one place away,
0:14:55 – 0:15:01 because back then there was City Call and Region Call and you really had to be careful,
0:15:01 – 0:15:07 That you are not pushing the phone bill in an explosively high direction.
0:15:07 – 0:15:13 Would that have been cool, as people told me back then, that you had to go for a local call.
0:15:13 – 0:15:15 Only one unit was consumed.
0:15:15 – 0:15:18 I didn't get that pleasure.
0:15:18 – 0:15:20 No, no, I had to pay right.
0:15:20 – 0:15:25 And when I finally got my parents to turn on a second phone line
0:15:25 – 0:15:32 To leave with my own number, just for my computer, boah, I was so proud.
0:15:32 – 0:15:36 Yeah, yeah, you can't even imagine that these days.
0:15:36 – 0:15:43 I mean, with an internet connection these days, you have at least three lines.
0:15:43 – 0:15:46 So you can have two conversations out and one in.
0:15:46 – 0:15:48 This is the case with most providers.
0:15:48 – 0:15:51 But you can have up to ten phone numbers.
0:15:51 – 0:15:56 This is no longer a comparison to then.
0:15:56 – 0:16:01 In any case, I was allowed a telephone bill plus the connection fees of two D-Mark
0:16:01 – 0:16:02 Produced in the month.
0:16:02 – 0:16:06 And they took all that out of my pocket money.
0:16:06 – 0:16:10 Yeah, yeah. It's a tough time.
0:16:10 – 0:16:13 But still, if you did it in the evening,
0:16:13 – 0:16:16 You have already been able to achieve a lot.
0:16:16 – 0:16:19 And dial into a mailbox for the first time.
0:16:19 – 0:16:22 My God, what was that great.
0:16:22 – 0:16:27 These expanses. And then you could even do that in BBS,
0:16:27 – 0:16:35 BDS, i.e. Battling Board Services, you could go there and discuss with others and
0:16:35 – 0:16:38 Write and even get private messages.
0:16:38 – 0:16:43 And that was all so people on your level, your level, with your interests.
0:16:43 – 0:16:54 And I found myself in there, which also led me to stop
0:16:54 – 0:16:59 I was very interested in computer science and so on, because there is still
0:16:59 – 0:17:05 so many other things to discover and nowadays you really have problems,
0:17:05 – 0:17:10 to follow through on all the developments. Yes, that's how it was then,
0:17:10 – 0:17:15 but again to hit the arc now back to the graphics card.
0:17:15 – 0:17:26 AI applications run on graphics cards, i.e. on these chips. Why? Because it's not so complex
0:17:26 – 0:17:34 Processors are like the processor that drives the computer, no matter if it's an Intel processor.
0:17:34 – 0:17:40 or an AMD processor, but there are a lot of very simple computing units in it,
0:17:40 – 0:17:48 understand the few instructions, i.e. instructions, but these extremely quickly
0:17:48 – 0:17:55 It can be processed because it is all so reduced. And they're just good in numbers
0:17:55 – 0:18:02 push back and forth, because in the end, in the calculations for the graphics, no matter for a game
0:18:02 – 0:18:12 or for a 3D application, number groups with x digits behind the comma must stop and
0:18:12 – 0:18:18 pushed back, back and calculated, can be changed and that's what they can do, with Nvidia these are
0:18:18 – 0:18:28 CUDA cores, extremely good, very fast. There comes a processor here, like my Intel, at all
0:18:28 – 0:18:35 Then after. That's how fast it is. And that's what makes it so interesting that you
0:18:35 – 0:18:42 instead of just gaming with it, also renders his videos on it, because that then damn fast
0:18:42 – 0:18:48 and goes efficiently. So when I think about it, on my processor takes such an hour video,
0:18:48 – 0:18:57 Let's say 30 minutes, that's on the processor. On my graphics card
0:18:57 – 0:19:04 That's eight minutes and that's a huge difference or when I do the transcript,
0:19:04 – 0:19:11 which is by far better than that of YouTube, for example, then I have about ten times
0:19:11 – 0:19:16 speed. That is, if I put a minute in there, let's say ten minutes,
0:19:16 – 0:19:22 is easier to calculate, ten minutes, then it's done in a minute and smiles
0:19:22 – 0:19:27 then also the graphics card, because then not the calculation the
0:19:27 – 0:19:31 The real problem is, but the data bus, which is the problem, because my
0:19:31 – 0:19:36 Graphics card is connected with eight lanes, i.e. with eight lanes
0:19:36 – 0:19:41 data traces, where 16 traces would theoretically be possible, but the connection of the
0:19:41 – 0:19:45 Graphics card does not allow this, because I don't have a premium model, because I
0:19:45 – 0:19:50 I don't want to buy a graphics card for $2,000.
0:19:50 – 0:19:54 Because for this I can buy three pieces or almost four pieces
0:19:54 – 0:19:57 and can build me an AI rack,
0:19:57 – 0:20:01 where I can run different processes at the same time.
0:20:01 – 0:20:05 Of course, the expensive more graphics memory would have
0:20:05 – 0:20:08 And then I could run bigger models.
0:20:08 – 0:20:13 But since Nvidia abolished this MV link,
0:20:13 – 0:20:17 You can't connect graphics cards properly.
0:20:17 – 0:20:22 And so load even larger models.
0:20:22 – 0:20:28 In addition, electricity consumption will eventually become an astronomical unit,
0:20:28 – 0:20:34 if you let this run all the time and then just let the graphics card run at the limit,
0:20:34 – 0:20:38 This is the case with some applications.
0:20:38 – 0:20:52 That's why I don't necessarily have such a fat graphics monster at home, but just several small graphics cards that are cheaper.
0:20:52 – 0:20:57 Especially when they've been doing this for two years, they're done.
0:20:57 – 0:21:04 That's why you get so cheap graphics cards on ebay from the Asian region, because the
0:21:04 – 0:21:06 Graphics cards are ready.
0:21:06 – 0:21:14 They're so swamped, not the processor, but the electronic components around them.
0:21:14 – 0:21:19 I mean, the processor and memory don't matter, but the capacitors and
0:21:19 – 0:21:24 manufacture the resistors and some switching regulators for certain voltages which:
0:21:24 – 0:21:33 under this extreme load. First of all, because there are trade restrictions with China,
0:21:33 – 0:21:41 the graphics cards are then also overclocked and thus driven to the absolute limit for
0:21:41 – 0:21:46 the data centers and then the hardware suffers even if it is not perfectly cooled. This
0:21:46 – 0:21:51 You have to imagine if there are ten graphics cards in such a rack,
0:21:51 – 0:21:54 Ventilation is not necessarily the best.
0:21:54 – 0:22:02 And so you buy almost in time-lapse aged hardware
0:22:02 – 0:22:04 for still expensive money
0:22:04 – 0:22:08 I wonder why the graphics card
0:22:08 – 0:22:12 for demanding applications, especially gaming,
0:22:12 – 0:22:14 Someday it will go bad
0:22:14 – 0:22:20 and then simply dot the refresh rate, the FPS.
0:22:20 – 0:22:40 Well, you have to think about that, too. The next consideration I had was, you get a Raspberry Pi, 16 GB of RAM, there's an SSD on it, but now not via interface, but via USB, that's still fast enough,
0:22:40 – 0:22:49 and put on top of it on the pins for the GPIO, this is such a special data bus,
0:22:49 – 0:22:56 Technic nerd, technic stuff, just technic stuff. Then you put it on top of it, then stop like that
0:22:56 – 0:23:03 AI-Head, that's just an additional card that you put on top of it and there's just a new one,
0:23:03 – 0:23:12 Now I think it's coming out in April, it's got 40 terra operations. So Terra, you can
0:23:12 – 0:23:21 Open it up, it's a million. Mega is a billion, sorry, a billion operations.
0:23:21 – 0:23:26 That's 40 billion operations. There you can already simple small language models
0:23:26 – 0:23:33 Let it run. And for such specific tasks, since sometimes summarizing texts,
0:23:33 – 0:23:45 small search analyses, to look for abnormalities in data sets, such a small model would be good
0:23:45 – 0:23:53 If you look at the 8 gigabytes, which are now this add-on module, this
0:23:53 – 0:24:03 AI-Head plus 2, that's the name of the thing. If you buy it, it will cost about
0:24:03 – 0:24:08 so 180 euros, even has a heat sink with it, because the thing will then of course accordingly
0:24:08 – 0:24:15 Pigs are called and have a power consumption, which is special, so from two to three watts below
0:24:15 – 0:24:23 Full load. Of course, this is extremely little compared to a fat graphics card. And if
0:24:23 – 0:24:29 I'm still thinking about the 300 watts at least, if not rather 400 watts with a medium
0:24:29 – 0:24:38 graphics card and an entire PC and this is then still energy-optimized and the at least
0:24:38 – 0:24:47 800 to 900 watts, if I take a server hardware, then this is a huge difference to the
0:24:47 – 0:24:56 Let's say 45 watts maximum use the stop of the Raspberry Pi and this additional board. But there can
0:24:56 – 0:25:02 You're going to stop. So everything that fits in these 8 gigabytes, you can run on it.
0:25:02 – 0:25:15 Of course, now compared to a modern one, which you can get very cheaply, is 30, 60, which would create about 20 tokens.
0:25:15 – 0:25:17 What is a torch now?
0:25:17 – 0:25:25 A torch is about a syllable of a word to illustrate this again.
0:25:25 – 0:25:34 So that's about 20 torches, 15 to 20. The other thing would be so maximum
0:25:34 – 0:25:40 Bring 8 torches, more like 6. Now you can say, okay, but these may be
0:25:40 – 0:25:49 Just one word per minute. Yes, but if I have time and that's because now
0:25:49 – 0:25:55 The crucial point. The processor would grab the text to the thing,
0:25:55 – 0:26:02 would then pack the model and the text into the memory of these additional things. Then
0:26:02 – 0:26:06 If the processor had nothing to do again, it would only take over a few control tasks
0:26:06 – 0:26:13 and because there is no desktop running on it, so no graphical user interface or something,
0:26:13 – 0:26:20 can completely lie down and slumber so nicely a bit and then would stop
0:26:20 – 0:26:25 This add-on module would do that and then it would rapidly consume little power compared to the
0:26:25 – 0:26:31 graphic has but of course I would then say such an evaluation or the summary
0:26:31 – 0:26:37 It takes several hours, if not all day, but it doesn't matter.
0:26:37 – 0:26:42 The task would be stopped. And the other advantage is that there's no shit coming
0:26:42 – 0:26:47 Hot air out. When I look at my computer under full load, graphics card
0:26:47 – 0:26:52 full power, then the processor full power at work is what's back there for a warm
0:26:52 – 0:27:01 Air comes out of the power supply, out of the case fans, then I already know where my, well well 800
0:27:01 – 0:27:06 Up to 1000 watts, which the computer needs per hour. And with the current
0:27:06 – 0:27:11 Electricity prices. You have to think about it, and let's talk about it like this.
0:27:11 – 0:27:19 An average of 35 cents. Then the thing runs for an hour and then 35 cents is gone. This is why
0:27:19 – 0:27:23 I also have my PC, if I'm not rendering videos or podcasts now
0:27:23 – 0:27:29 render or make the transcripts, always keep a power saving mode, because the thing
0:27:29 – 0:27:35 firstly then is quieter and secondly consumes much less power and the power on my
0:27:35 – 0:27:44 What does this thing actually have? 32 core. It's enough and I need it.
0:27:44 – 0:27:50 Much less power, because the clock is almost only half as high. And that's still fast
0:27:50 – 0:27:55 Enough to work smoothly. There are no hangers, no jerks, nothing. Even so I can
0:27:55 – 0:28:02 stream because I have enough cores on which the processes can be distributed. And if
0:28:02 – 0:28:07 I would then take a more efficient operating system, such as Linux, where the tasks
0:28:07 – 0:28:14 Even better distributed as in Windows, then there are even fewer problems. But this
0:28:14 – 0:28:23 These are completely different topics and completely different areas. Yes, that was my thoughts and my experience
0:28:23 – 0:28:30 and also some stories from the past of a much, much younger Sascha, who
0:28:30 – 0:28:39 At 14, 15, he was so happy that he had an IBM-compatible
0:28:39 – 0:28:48 I got a computer. And my God, what has changed my life? Unbelievable,
0:28:48 – 0:28:56 unbelievable. I never thought I'd have my own AI agents like I do today.
0:28:56 – 0:29:00 They live on different devices.
0:29:00 – 0:29:05 There are communication channels where information goes back and forth
0:29:05 – 0:29:08 stung, pushed,
0:29:08 – 0:29:13 who observe current articles and hashtags,
0:29:13 – 0:29:17 Watch the comments on videos
0:29:17 – 0:29:20 And then make an assessment.
0:29:20 – 0:29:23 Pre-sort my emails.
0:29:25 – 0:29:29 Of course not from my main account, but for an extra account.
0:29:29 – 0:29:33 And so many other things. I never would have thought that possible.
0:29:33 – 0:29:37 That I can go there myself and with my phone and then say
0:29:37 – 0:29:43 "Hey, do this and that and then send me the result as an e-mail."
0:29:43 – 0:29:51 And that this is understood because a Whisper model is running
0:29:51 – 0:29:55 and then stop translating the voice message into text.
0:29:55 – 0:29:58 And then there is the processing.
0:29:58 – 0:30:00 And then there's another answer.
0:30:00 – 0:30:05 And I can even use text-to-speech as a language again.
0:30:05 – 0:30:11 I can set it up so I can just have someone call me
0:30:11 – 0:30:13 And then give a message.
0:30:13 – 0:30:15 Then he can answer.
0:30:15 – 0:30:17 And I'll get it as an email
0:30:17 – 0:30:24 or stop as a direct message in the communication channels I have set up as a message.
0:30:24 – 0:30:28 So, here, Peter said the appointment is okay.
0:30:28 – 0:30:39 And I could never have imagined that a person who is interested in
0:30:39 – 0:30:43 that everything at home can build and set up itself
0:30:43 – 0:30:48 You only need to invest a few dollars in electricity.
0:30:48 – 0:30:54 and a little time and then have such automations.
0:30:54 – 0:30:57 That's unbelievable, unbelievable.
0:30:57 – 0:31:02 And considering how full YouTube and other channels are at the moment
0:31:02 – 0:31:15 Even in the podcast world. The topic Open Cloud before Moldbot and Cloudbot is the topic,
0:31:15 – 0:31:27 Of course, it's huge to put together some AI agents. Yes, let's
0:31:27 – 0:31:35 see what the future brings and whether I will heat my apartment with gas or electricity in the future,
0:31:35 – 0:31:40 Because there is so much hardware around and you are counting on it. Yes then I thank you for it
0:31:40 – 0:31:47 Listen and if you liked it, I'd love to hear a comment and that was
0:31:47 – 0:31:54 Today again a quieter edition without any beef behavior. I hope it has you
0:31:54 – 0:32:01 still please and I say goodbye to next time your sasha
0:32:01 – 0:32:04 [Music]
0:32:04 – 0:32:10 [Music]
0:32:10 – 0:32:16 [Music]
0:32:16 – 0:32:22 [Music]
0:32:22 – 0:32:28 [Music]
0:32:28 – 0:32:31 [Music]
0:32:31 – 0:32:37 [Music]
0:32:37 – 0:32:43 [Music]
0:32:43 – 0:32:49 [Music]
0:32:49 – 0:32:56 [Music]
0:32:56 – 0:32:59 [Music]
0:32:59 – 0:33:05 [Music]
0:33:05 – 0:33:11 [Music]
0:33:11 – 0:33:17 [Music]
0:33:17 – 0:33:23 [Music]
0:33:23 – 0:33:26 [Music]
0:33:26 – 0:33:32 [Music]
0:33:32 – 0:33:38 [Music]
0:33:38 – 0:33:44 [Music]
0:33:44 – 0:33:50 [Music]
0:33:50 – 0:33:53 [Music]
0:33:53 – 0:33:59 [Music]
0:33:59 – 0:34:05 [Music]
0:34:05 – 0:34:11 [Music]
0:34:11 – 0:34:17 [Music]
0:34:17 – 0:34:20 [Music]
0:34:20 – 0:34:26 [Music]
0:34:26 – 0:34:32 [Music]
0:34:32 – 0:34:38 [Music]
0:34:38 – 0:34:44 [Music]
0:34:44 – 0:34:47 [Music]
0:34:47 – 0:34:54 [Music]
0:34:54 – 0:35:00 [Music]
0:35:00 – 0:35:07 [Music]
0:35:12 – 0:35:15 [Music]
0:35:15 – 0:35:21 [Music]
0:35:21 – 0:35:27 [Music]
0:35:27 – 0:35:33 [Music]
0:35:33 – 0:35:40 [Music]
0:35:40 – 0:35:43 [Music]
0:35:43 – 0:35:49 [Music]
0:35:49 – 0:35:55 [Music]
0:35:55 – 0:36:01 [Music]
0:36:01 – 0:36:07 [Music]
0:36:07 – 0:36:19 [Music]
0:36:19 – 0:36:22 [Music]

Christmas of 1992.
A 386SX with 25 MHz, 2 MB RAM and a 100 MB hard drive moves in – and changes everything.

This episode is about technological leaps in time. From the first almond bread in GW-BASIC, which took up to 90 minutes without a mathematical coprocessor, to the first 2,400-bit modem and mailbox nights, to today's system with i9-13900K, 64 GB of RAM and multiple NVMe SSDs, which shifts 1 GB in 0.15 seconds.

But it doesn't stay with nostalgia.

We talk about the current graphics card and storage crisis with AI data centres, CUDA cores, energy consumption, overclocked data centre hardware from the Far East – and whether small, efficient systems such as a Raspberry Pi with AI head can be the smarter alternative in the long term.

How does technological development feel when you have witnessed it from the beginning?
What does 30 years of computing power mean in real numbers?
And why isn't the move from the 387SX coprocessor to local AI agents as big as it seems?

A calm, personal sequence between retro computing, market analysis and vision of the future.

🎙 Chronological thematic structure

1️⃣ Entry - Current hardware reality

Price explosion for graphics cards, RAM and SSDs
AI data centers as drivers of scarcity
Personal experience with GPU price development
Nvidia H100 & professional AI hardware

2️⃣ Review - The First PC (1992)

Christmas of 1992: 386SX 25 MHz
2 MB RAM, 100 MB HDD, 512 kB graphics card
SX vs. DX Difference
Meaning of this moment

3️⃣ memory sizes then vs. today

100 MB "will never be full"
Software, games, programming environments
Compared to 8 TB SSD
Factor comparisons (capacity & speed)

4️⃣ First programming experience

GW-BASIC
Almond bread fractal
90 minutes render time
Mathematical Coprocessor (387SX)
Performance gain and mindblowing moment

5 ️⃣ The Mailbox era

2.400 Bit/s Modem
dial-in, telephone costs, second line
BBS & Community Feeling
Digital youth of an analogue generation

6 ️⃣ Technological evolutionary leap

From minutes to seconds (1 GB copy)
386SX vs. i9-13900K
2 MB vs. 64 GB RAM
Hardware exponential leap over 30 years

7️⃣ Modern GPU architecture & AI

Why AI runs on GPUs
CUDA cores vs. classic CPUs
Video rendering & transcription
Lane Limitations & Multi-GPU Approach
Wear and tear of data center hardware

8️⃣ Energy & Efficiency

800-1000 watts under full load
Electricity prices & real costs
Power saving mode & Process distribution
Windows vs. Linux Efficiency

9 ️⃣ Alternative concepts

Raspberry Pi + AI-Head
40 TOPS at 2-3 watts
Operate small language models locally
Slow but efficient processing

🔟 Future & personal reflection

Own AI agents in the home network
Automated processes & voice control
Development of teenager with 386SX
Where does the journey go?
Heating with hardware?

Advertising:

Leave feedback on the issue? ✍️ Then write a comment Audio feedback on the show? ☎️ 0231 53075818

Trustpilot

Contributors to this episode

Sascha

Number of downloads: 305
Playing time: 0:36:21
Date of admission: 18.02.26
This episode is from season 4, there are a total of 4 seasons.

One According to Funk Publication Cartel Production - Where stories are transmitted unmistakably.

Podcast license:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License

Note: This post contains affiliate links. If you shop through these links, I'll get a small commission at no extra cost to you.

The Digression 54 Digression into the Fediverse

Format	size	Download
MP3 Audio	34 MB	Download
AAC m4a	17 MB	Download

transcript
00:00 - 0:00:02 Music
0:00:02 – 0:00:05 [Music]
0:00:05 – 0:00:11 [Music]
0:00:11 – 0:00:17 [Music]
0:00:17 – 0:00:23 [Music]
0:00:23 – 0:00:29 [Music]
0:00:29 – 0:00:39 Hello and welcome to a new edition of the digression, where I am noisy about the topic
0:00:39 – 0:00:45 Frankstar doesn't digress as much as I should in this format.
0:00:45 – 0:00:56 I say hello. Yes, for all those who do not know me, I am Sascha Markmann and run a few
0:00:56 – 0:01:08 Podcast. Today's topic I want to talk about with you is the Feediverse or why I'm
0:01:08 – 0:01:20 Consider it important that there be alternatives to those led by crazy tech billionaires.
0:01:20 – 0:01:35 Networks exist and why I now use more of these networks such as the
0:01:35 – 0:01:44 Big tech companies. Why should there be alternatives? I once realized that
0:01:44 – 0:01:54 Social networks are no longer really social at goose feet, but that it is there
0:01:54 – 0:02:02 There is no longer a proper culture of discussion. I've done podcasts about it.
0:02:02 – 0:02:09 Somehow a culture of discussion is broken. I've had a YouTube video years ago.
0:02:09 – 0:02:15 I've done it and it just hasn't gotten any better. The first time it's me
0:02:15 – 0:02:24 At some point, when vaccination was compulsory for elementary school children and kindergarten children,
0:02:24 – 0:02:32 towards chickenpox, how then animal excited about it and mild infections
0:02:32 – 0:02:38 were sold as the most severe vaccine damage, if the child has just gotten a little fever.
0:02:38 – 0:02:47 So if it goes after that, I would have died several times if I had done so with statements of my own.
0:02:47 – 0:02:54 Mother hears how I reacted to a vaccination in places, which I believe more
0:02:54 – 0:03:01 as normal. And in the comment columns it was hot and there was no
0:03:01 – 0:03:06 Allowed other opinion. This has been going on over the years and completely
0:03:06 – 0:03:15 Things got worse during the coronavirus pandemic. I know, it's a very bad subject. How to stop
0:03:15 – 0:03:26 Anti-vaccination and vaccination advocates then just textually and verbally crushed each other
0:03:26 – 0:03:34 have. I realized that there was no room for me anymore. I am not posting
0:03:34 – 0:03:44 Not on Facebook and Instagram, not on Twitter. We don't want a totally crazy
0:03:44 – 0:03:52 Support tech billionaire who has harsh racist bias. The only thing
0:03:52 – 0:04:03 It just happens that I automatically post content to the networks. Speak on Facebook,
0:04:03 – 0:04:11 on Instagram, on LinkedIn and what they are all called. I just don't feel like it anymore
0:04:11 – 0:04:17 on it. While I'm on these networks every day to gather information,
0:04:17 – 0:04:26 Because some companies can only be found on certain networks. Journalistic content is
0:04:26 – 0:04:35 Find it here and there, too. That's how I look through my feed. And there is the
0:04:35 – 0:04:43 The next thing that bothers me a lot. Everything is algorithmic. So it is presumed,
0:04:43 – 0:04:52 What I want to see. Although these algorithms are quite good, I hardly have any
0:04:52 – 0:05:00 range. I see it in the statistics I keep as a content creator when I create an account.
0:05:00 – 0:05:12 As a content creator, you get more statistical data, insights like they always do.
0:05:12 – 0:05:19 It's so nice to see what the reach of the posts is. And that is mild
0:05:19 – 0:05:31 said lousy. Even this whole craving for attention is something I really …
0:05:31 – 0:05:43 That's against the grain. Then you have to say, I have posted less and less private things there
0:05:43 – 0:05:50 and for years I have been posting there except for the advertising of my music, my podcasts,
0:05:50 – 0:05:57 My blog posts are no more. It's possible that when there's something really good,
0:05:57 – 0:06:09 that I share the stop, so retweet, no idea what it's called somewhere else, to stop it
0:06:09 – 0:06:20 to show others who follow several. That is by and large. And what stops,
0:06:20 – 0:06:30 There are only lone fighters on the way. So this connection, which is so in the early years
0:06:30 – 0:06:36 Facebook has felt or something, it's no longer there, but everyone posts their shit and
0:06:36 – 0:06:43 That was it. And as I said, hold on to your own opinion and do not allow any other opinion,
0:06:43 – 0:06:51 This, of course, is one of the most serious points. So I looked for alternatives. Now
0:06:51 – 0:07:01 You have to say that there are alternatives and that's the "feediverse". So it is, to make it very rough
0:07:01 – 0:07:10 explain many small villages or islands on the Internet that are interconnected,
0:07:10 – 0:07:19 But each stands so for itself. There are protocols that are open. That, what is the name of a protocol,
0:07:19 – 0:07:36 Active Pub, I think most of them are based on that. And now when I post something that's so similar
0:07:36 – 0:07:46 Like Facebook, someone can also subscribe to Mastodon, which is very similar to Twitter. And
0:07:46 – 0:07:55 You don't have to have a Facebook account to see the content.
0:07:55 – 0:08:07 Or if I share a video on PearTube, then anyone can share it on these similar as Twitter,
0:08:07 – 0:08:15 Mastodon or similar to Facebook can subscribe and see this. And that's the great thing.
0:08:15 – 0:08:21 I don't have to sign up on x different platforms, but I have a place somewhere,
0:08:21 – 0:08:31 where I feel comfortable and then have access to all the content that is available in the Feli verse.
0:08:31 – 0:08:40 That's awesome. And if you will, I already said that Mastrodon is similar to Twitter,
0:08:40 – 0:08:50 without Elon Musk. Pixel Feed is similar to Instagram, but without any filters,
0:08:50 – 0:08:57 Because it's really just the high photos I'm uploading. PearTube is like YouTube,
0:08:57 – 0:09:05 Lemony is like Reddit. I think I said it right. Without this whole
0:09:05 – 0:09:11 The meme-coin drama. This has become a real plague on Reddit.
0:09:11 – 0:09:21 For lovers of 4chan and as the things were all called there is just MISC.E or Firefish. This
0:09:21 – 0:09:31 is really the colorful chaos, so GIFs, reaction and memes. It's a very interesting place.
0:09:31 – 0:09:42 Then there is "Mobileeezen", which is almost such a calendar of events, very interesting.
0:09:42 – 0:09:54 "Whitefrinkly" or "Plump", these are quasi blog offers, similar to bloggers or WordPress,
0:09:54 – 0:10:05 Hold on for longer texts. Then what else is there? Hubzilla. It's very similar to Facebook.
0:10:05 – 0:10:16 but even more rudimentary, but has the possibility to install plugins to RSS feeds
0:10:16 – 0:10:22 automatically pull in and then you can click on the posts and then stop in your
0:10:22 – 0:10:31 Share your own channel. I also like to use to quickly stop my content on Hubsiller
0:10:31 – 0:10:39 to make available and thus stop the whole Feediverse. Radio options, something similar to
0:10:39 – 0:10:51 Spotify and haven't you seen it? Yes, these are the alternatives. What are the advantages
0:10:51 – 0:11:02 The feediverse. Data protection and control. No corporations, no annoying user profiles,
0:11:02 – 0:11:10 So not annoying, glazing user profiles is the right word. It's really clear,
0:11:10 – 0:11:16 If I have subscribed to someone, then I also get what I have subscribed to. And not
0:11:16 – 0:11:22 That somehow, yes, you might be interested, but not that. But if I follow someone,
0:11:22 – 0:11:32 Subscribe, and so does it. It's not advertising. These systems are all free,
0:11:32 – 0:11:36 Except for Blue Sky. Blue Sky will be monetized someday.
0:11:36 – 0:11:47 and the Amen in the Church. Communities are diverse. There are also filtered
0:11:47 – 0:12:01 Areas where there is no FSK 18 and also other types of content that is criminally relevant
0:12:01 – 0:12:07 It can be filtered out or you don't have bridges to it. As I said, the filivers are small.
0:12:07 – 0:12:15 Islands or cities, villages that are connected and there is no need to connect
0:12:15 – 0:12:24 give. Not even in the back, but these instances are blocked. And it's easy
0:12:24 – 0:12:31 open. The source code is open, I can install an instance myself and run it,
0:12:31 – 0:12:41 but you have to think about it well, because by keeping this open standard there, you can
0:12:41 – 0:12:48 A lot of data through. So it's unbelievable what's holding a small instance with a single
0:12:48 – 0:12:54 User caused for a traffic. So if you do that, you should
0:12:54 – 0:13:01 a powerful server and best stop no data limits, so the
0:13:01 – 0:13:07 Traffic rate, there should be no limits, otherwise it will be expensive.
0:13:07 – 0:13:16 I said openly and extensibly. It is done by private individuals
0:13:16 – 0:13:25 developed and advanced. And so there are always updates, innovations, improvements.
0:13:25 – 0:13:31 Although it is not so fast or sometimes not so well documented at the beginning,
0:13:31 – 0:13:40 But with time it comes. So I'm amazed at what functions there are now in the Phillyverse.
0:13:40 – 0:13:50 There are other commercial providers who are really looking for it. The disadvantages, yes, there are.
0:13:50 – 0:13:59 fragmentation, as I said, these are all your own instances and you never know so precisely,
0:13:59 – 0:14:05 Where to register. I had the problem in the beginning too. I'm the first time in
0:14:05 – 0:14:11 Contact came with the 4Diverse about Mastodon, because after the first exodus,
0:14:11 – 0:14:15 which was there – I don’t know how long it has been – I have been there
0:14:15 – 0:14:21 notified. Or wanted to sign me up, but didn't know where. And then I saw sometime
0:14:21 – 0:14:29 I've had a well-known someone I follow on the podcast scene and know a little bit.
0:14:29 – 0:14:37 virtual on the servers of Chaos Computer Club and of which I have an invitation link
0:14:37 – 0:14:44 Get it. And so I ended up on the instance of Chaos Computer Club and that's a
0:14:44 – 0:14:53 Very exciting scene. A lot of agreement with my interests, even in terms of opinion.
0:14:53 – 0:15:02 And so I felt very comfortable very quickly. I can understand that there is someone,
0:15:02 – 0:15:09 who really doesn't know about it and doesn't have someone who's already there that it's overwhelming
0:15:09 – 0:15:16 can. And it is never so well documented for whom which instance in which
0:15:16 – 0:15:25 Part of the feediverse is, the same problem I had with an alternative for Facebook,
0:15:25 – 0:15:30 So when I decided on HubSeller, I didn't know which
0:15:30 – 0:15:37 Instance I'm leaving. Or in the instances of PairTube there are sometimes storage limits,
0:15:37 – 0:15:46 So what are 50 gigabytes for videos today, that's maybe 30 videos with me and then
0:15:46 – 0:15:52 The memory is gone. And I was looking for a long time, until I found something,
0:15:52 – 0:16:03 where I feel comfortable, which also has an in-house policy that is not so restrictive
0:16:03 – 0:16:10 is. Because that's what it is. instances, which then have such special rules of behavior,
0:16:10 – 0:16:21 or you have to be careful how you post, then stop, which of course can be an advantage,
0:16:21 – 0:16:30 A very sensitive handling of sensitive content is required. So that you can trigger warnings
0:16:30 – 0:16:40 I assume that you describe photos, that if they are content that could be burdensome,
0:16:40 – 0:16:50 that you have to make them actively visible. That's part of it
0:16:50 – 0:16:58 your compartments. Often the comfort suffers somehow, because the networks are still rudimentary. How
0:16:58 – 0:17:03 They say it's all done by private individuals.
0:17:03 – 0:17:06 There are no companies behind it.
0:17:06 – 0:17:10 Building reach is also essential
0:17:10 – 0:17:13 It's harder because there are no algorithms.
0:17:13 – 0:17:19 There is. Haha, who would have thought that one
0:17:19 – 0:17:21 Propose content. And you have to
0:17:21 – 0:17:24 Depending on a little
0:17:24 – 0:17:28 They have a technical understanding of what the
0:17:24 – 0:17:33 Integration of protocols or as with Hubsiller, if I have certain functions
0:17:33 – 0:17:41 I want to reinstall what's going on. There are, of course, alternatives that do not
0:17:41 – 0:17:55 There are a lot of divers. Speak instead of messengers like WhatsApp or something like Signal. Instead of discord like that.
0:17:55 – 0:18:06 Like the matrix. Nextcloud instead of Google Drive or Owen Cloud and as they are all called. Most of them
0:18:06 – 0:18:16 Wet systems can also be an alternative to Google Cloud. Open Street Maps instead of Google Map.
0:18:16 – 0:18:26 Yes, I already said radio dials, where radio dials already send into the feediverse and a
0:18:26 – 0:18:34 connection so has. Kuben as an alternative to Reddit, but stop in the Feediverse style. I have so far
0:18:34 – 0:18:41 I haven't used it yet and I haven't had any contact yet. And then there is this
0:18:41 – 0:18:54 Free, I can't pronounce it, FrindiK as Facebook, but also so in Facebook alternatives
0:18:54 – 0:19:01 Style. But they are still pretty much in development, there is still a lot missing, so yes,
0:19:01 – 0:19:09 Half ashamed of a Facebook. There is no real Facebook alternative.
0:19:09 – 0:19:17 Like I said, Hubzilla is the closest thing to me. So to me personally.
0:19:17 – 0:19:26 Yes, so we have now heard what there are advantages, for disadvantages, what offers there are,
0:19:26 – 0:19:32 What services and there are many, many more. Some are so small or even so in the
0:19:32 – 0:19:43 Development that you don't have to talk too much about it. The Fiddywörts is not a perfect place
0:19:43 – 0:19:52 or yes, but it's just a playground, a construction site and you can experience there,
0:19:52 – 0:19:59 How something like that grows. And I can say, in the years since I've been there, certainly since 2018,
0:19:59 – 0:20:11 Yes, it could be, maybe 17, 18, I've seen Mastodon continue to evolve
0:20:11 – 0:20:20 and other services. Pixel-Filt is one such thing. There are incredibly strong people,
0:20:20 – 0:20:29 who take incredibly horny photos and I get more reactions to my graphics,
0:20:29 – 0:20:39 Which I do now for the podcasts or also for my music projects, as elsewhere. himself
0:20:39 – 0:20:45 on Instagram, where I really have a lot of followers, there is hardly a reaction and get there
0:20:45 – 0:20:54 with zero or two or three followers so much reaction from other people, because it is also
0:20:54 – 0:21:02 this general feed returns to the instance or stops a feed for pixel feed. And
0:21:02 – 0:21:12 That's interesting. So there are people roaring at me and my content, which I never reach otherwise
0:21:12 – 0:21:21 I would have. And I believe myself, folks, I really post a lot of my productions on
0:21:21 – 0:21:36 Facebook in groups, on the Facebook page. I post it on Instagram and see that there
0:21:36 – 0:21:41 Hardly anyone ever comes. But from the feediverse I see again and again how people stop
0:21:41 – 0:21:51 And that's slowly getting more. And these are also pleasant comrades. The internet has otherwise
0:21:51 – 0:21:58 Always the habit, you write a review only when you are very dissatisfied
0:21:58 – 0:22:07 is. Rarely when you are satisfied. And that's where I see people reacting very positively.
0:22:07 – 0:22:11 Of course, there are times here and there someone who has a critical demand.
0:22:11 – 0:22:17 But then the true interest is a critical demand.
0:22:17 – 0:22:21 And I think that's nice.
0:22:21 – 0:22:27 That's a bit of what I saw at the beginning of Facebook.
0:22:27 – 0:22:30 I've also gone through different stages.
0:22:30 – 0:22:36 At the very beginning of my internet career, I must have been 16 years old.
0:22:36 – 0:22:43 Newsgroup. Then I slowly discovered the web, then came Web 2.0. Oh,
0:22:43 – 0:22:51 More than forums. At the time, there was a lot of discussion in forums.
0:22:51 – 0:23:01 They were nerdy places where people with a common interest stopped, moved. Then came
0:23:01 – 0:23:11 slowly so networks on, so not networks, it was just so platform, so chat systems with
0:23:11 – 0:23:18 Your own profile pages, where you could make games, where you also had something like a mini-blog.
0:23:18 – 0:23:26 And back then it was really like that, you had a few trolls, but the predominant
0:23:26 – 0:23:33 The number of people who responded was always positive. Then came the time of
0:23:33 – 0:23:46 Facebook and there was still such a mood and sometime in the late 2010s this is
0:23:46 – 0:23:51 more and more tilted. Maybe it even started in the middle, so around the middle,
0:23:51 – 0:23:58 That it is slowly tipped because access to the Internet has always been easier. You had to
0:23:58 – 0:24:03 Do not dial in with modem, as I did at the beginning. Then you had to DSL,
0:24:03 – 0:24:12 have an ISDN connection so that it could be halfway fast. It was all extra
0:24:12 – 0:24:19 Devices or cards that had to be installed. Then came the DSL. That was also a bit
0:24:19 – 0:24:27 It's tricky to do that in the beginning. You had to get yourself a router if you were
0:24:27 – 0:24:38 Several people wanted to go online. Or you had a card in your computer. Or they had an
0:24:38 – 0:24:46 Network card and then the modem from Deutsche Telekom or the other provider. That was always
0:24:46 – 0:24:57 Everything was a little bit fumbling and yes, there wasn't that much there. And with the triumph of smartphones
0:24:57 – 0:25:04 Suddenly everyone had the internet in their pocket and could then just start
0:25:04 – 0:25:14 Yes, let his love, in the positive as well as in the negative sense. And you could really see that,
0:25:14 – 0:25:25 So on the posts, what they had done, on my content. In the beginning, I have
0:25:25 – 0:25:35 I started using YouTube around 2009, so in 2007 I already had an account, but in 2009 I did.
0:25:35 – 0:25:41 I'm really starting to make YouTube. And in the beginning everything was mostly positive,
0:25:41 – 0:25:46 I would say 90% Positive and 10% There were trolls there.
0:25:46 – 0:25:55 And this has simply shifted over time to such an extent that just 70% trolls
0:25:55 – 0:26:08 and the other 30% Either do nothing at all or get little positive back.
0:26:08 – 0:26:10 So I get back.
0:26:10 – 0:26:18 I'm glad I made a video when there was no reaction. So besides thumbs up, thumbs down.
0:26:18 – 0:26:25 Because otherwise you always had to deal with some stupid comment.
0:26:25 – 0:26:36 And not every comment bounces off a how, yes, how, the film is not a word for one,
0:26:36 – 0:26:42 like a giant flummi, but some already scratch the ego a bit, because there are people,
0:26:42 – 0:26:48 they have their breastiness so well under control that the perfect tips can write.
0:26:48 – 0:26:56 So it's unbelievable. Now some would say, yes, you used to make videos,
0:26:56 – 0:27:03 Why don't you do it today? Because I don't see the meaning anymore. So if I
0:27:03 – 0:27:08 I sit there and I talk for 30 minutes, then I have a certain amount of effort,
0:27:08 – 0:27:14 To do that. So that means I have to record the sound well, I have to make sure
0:27:14 – 0:27:24 that the picture is good, I have to prepare myself, if necessary,
0:27:24 – 0:27:32 that I am still presentable, because I am not always so. Then thoughts about that,
0:27:32 – 0:27:36 What I want to say, make, if necessary, punctuation points or things completely.
0:27:36 – 0:27:44 It rarely happens, but I've done it before. Then this whole other thing, what else
0:27:44 – 0:27:54 with is. Rendering can also take a long time, although strong computer, good graphics card. Good,
0:27:54 – 0:28:01 It's been two generations since then, but it's still very fast. Yes, but
0:28:01 – 0:28:06 Then you sit there waiting for your video to finish after 40 minutes.
0:28:06 – 0:28:09 And then you have to upload it, do everything.
0:28:09 – 0:28:15 And then you have almost no reaction there or few views.
0:28:15 – 0:28:18 And that was the point where I said,
0:28:18 – 0:28:23 Okay, you're not doing the work anymore and you're just doing podcasts now.
0:28:23 – 0:28:28 In addition to blogging and all the other projects I do,
0:28:29 – 0:28:42 Because not only do I do this, but I also do music and get involved in other clubs and things like that.
0:28:42 – 0:28:51 Yes, that's just when you've been around for so long and there are people who've been around longer,
0:28:51 – 0:29:01 You have to set priorities and for me just the moving image and social media just doesn't make sense anymore.
0:29:01 – 0:29:09 That's why I'm just posting my ads and that's it.
0:29:09 – 0:29:19 Actively posting something or so at the big technology companies is simply no longer fun.
0:29:19 – 0:29:26 I don't know, I used to post all sorts of shit, so also thoughts or
0:29:26 – 0:29:33 Before posting it in this way, I also posted my thoughts, i.e. in text form.
0:29:33 – 0:29:40 Or quotes that I found inspiring, I don't do anymore.
0:29:40 – 0:29:47 Because at some point there was just nonsense as a reaction or nothing more.
0:29:47 – 0:29:55 And if you feel like you're doing something for free and you don't care, why keep doing it?
0:29:55 – 0:29:59 If you don't feel like you have to do something.
0:29:59 – 0:30:13 Yes, these are my thoughts on it and why I consider the Free Diverse to be a more perfect place for me.
0:30:13 – 0:30:23 Yeah, okay, thanks for listening. I hope I could give you a little insight
0:30:23 – 0:30:32 why it has become relatively silent in established social networks and
0:30:32 – 0:30:42 You can find me more in Feediverse. Where I'm to be found, everything is to be found in the show notes
0:30:42 – 0:30:51 or to be found on my website at Lautfunk.de with or without www does not matter.
0:30:51 – 0:30:58 And there is an "About me" page and everything is linked, everything is wonderful.
0:30:58 – 0:31:09 In the podcast there are links for the podcasts and for me as a person,
0:31:09 – 0:31:17 Some channels overlap, because I mainly act as a sound radio on the Internet.
0:31:17 – 0:31:25 Rarely as Sascha Markmann and rarely as a space wave, which is my artist name.
0:31:25 – 0:31:31 Exactly. For music projects, to have briefly explained.
0:31:31 – 0:31:36 Thanks for listening and until next time.
0:31:36 – 0:31:38 Farewell to Sasha.
0:31:38 – 0:31:41 [Music]
0:31:41 – 0:31:47 [Music]
0:31:47 – 0:31:53 [Music]
0:31:53 – 0:31:59 [Music]
0:31:59 – 0:32:05 [Music]
0:32:05 – 0:32:08 [Music]
0:32:08 – 0:32:14 [Music]
0:32:14 – 0:32:20 [Music]
0:32:20 – 0:32:26 [Music]
0:32:26 – 0:32:32 [Music]
0:32:32 – 0:32:35 [Music]
0:32:35 – 0:32:41 [Music]
0:32:41 – 0:32:47 [Music]
0:32:47 – 0:32:53 [Music]
0:32:53 – 0:32:59 [Music]
0:32:59 – 0:33:02 [Music]
0:33:02 – 0:33:08 [Music]
0:33:08 – 0:33:14 [Music]
0:33:14 – 0:33:20 [Music]
0:33:20 – 0:33:26 [Music]
0:33:26 – 0:33:29 [Music]
0:33:29 – 0:33:35 [Music]
0:33:35 – 0:33:41 [Music]
0:33:41 – 0:33:48 [Music]
0:33:48 – 0:33:54 [Music]
0:33:54 – 0:33:57 [Music]
0:33:57 – 0:34:04 [Music]
0:34:04 – 0:34:11 [Music]
0:34:11 – 0:34:19 [Music]
0:34:19 – 0:34:22 [Music]
0:34:22 – 0:34:29 [Music]
0:34:29 – 0:34:37 [Music]
0:34:37 – 0:34:44 [Music]
0:34:44 – 0:34:47 [Music]
0:34:47 – 0:34:53 [Music]
0:34:53 – 0:34:59 [Music]
0:34:59 – 0:35:05 [Music]
0:35:05 – 0:35:11 [Music]
0:35:11 – 0:35:14 [Music]
0:35:14 – 0:35:20 [Music]
0:35:20 – 0:35:26 [Music]
0:35:26 – 0:35:32 [Music]
0:35:32 – 0:35:38 [Music]
0:35:38 – 0:35:56 [Music]
0:35:56 – 0:35:58 [Music]

In this episode, I openly talk about why I turned my back on the big social media platforms like Facebook, Instagram and X (Twitter). It is about frustration, a lost culture of discussion, toxic algorithms – but also about hope. I have found alternatives: This Fediverse. What that actually is, how it works, what platforms are there – from Mastodon to PeerTube to Hubzilla – and why it's not just nerdy gimmicks, but a piece of digital self-determination.

I am talking about the Advantages and downsides this decentralised world, how to start there, why much is different – and perhaps better for that very reason.
And I'll tell you what this has to do with my history as a long-time Internet user. Between blogs, forums, YouTube, Facebook and a good portion of net nostalgia.

🧭 Overview of topics – digression into the Fediverse

🌀 Intro: Why I'm out

A personal look back at the frustration with Facebook, Instagram and X – and how the culture of discussion on social networks crumbled.

🧩 What is the Fediverse?

Explanation of the basic idea: Decentralization, open protocols like ActivityPub, many small platforms instead of a corporate monoculture.

🧱 Overview of the most important platforms

Mastodon – Microblogging as it used to be on Twitter
Pixelfed – Image network without filtering
PeerTube – Videos without YouTube advertising
Lemmy – Reddit-style discussion forums
Misskey / Firefish – Memes, reactions, colourful communities
Mobilizon – Events without Facebook
WriteFreely / Plume – For bloggers and long texts
Hubzilla – Facebook-like, modularly expandable
Funkwhale – Community music streaming

✅ Advantages of the Fediverse

Data protection & control over your own content
No algorithms, no advertising
Openness, self-hosting, diversity
A friendlier, more active community culture

⚠️ Disadvantages & Challenges

Fragmentation & Entry barriers
Loss of comfort compared to commercial platforms
Reduced reach without algorithm
Partial technical or organisational barriers

🔁 Alternatives outside the Fediverse

Signal Instead of WhatsApp
Matrix / Element instead of discord
Nextcloud Instead of Google Drive
OpenStreetMap Instead of Google Maps
Kbin Instead of Reddit
Friendica As a Facebook Attempt

🧠 Personal conclusion

Why the Fediverse feels right to me – even if it is not a perfect place, but a digital construction site with a lot of heart.

Links to the issue: