>>8785 You'd need your own semiconductor industry, since you can't trust other countries to not backdoor your shit. Not just the final fab of boards them selves. There's only two countries on earth capable of that and it's the USA (albeit with the non profitable consumer parts outsourced to allies), and China (not as bleeding edge as US) Encryption in military is generally symmetric cipher I think. So unless quantum stops being a meme that means you will be using a symmetric stream cipher, where the the "key" is identical on both ends and changes constantly as you transmit, you take a very large (pseudorandom) hash and permute it based on the time of the device and then send small bits of the entire message at a time. This is a little different from say, and elliptic curve where you take a large random prime number and just MULTIPLY. That means you need: A functional mathematics/cryptography (educational) industry Semiconductor industry IT able to implement this as non retards (software industry) For semiconductors you need mining (raw materials), chemical (for the process to create wafers), material science (clean room) photolithography machines (laser), machinists (see before), design labs (for the instruction set architecture and the board itself), utilities. Metallic inputs are sillicon, copper and a bunch of rarer transition and metalloids.

>>8786 >you can't trust other countries to not backdoor your shit. Do backdoors mean anything if the only networking device a simple radio that is only supposed to receive a stream of digital data, and the manufacturer has no say in how you encrypt that? I just can't see how could they send instructions to some hardware backdoor when the program that handles the digital data should reject anything not encrypted the right way as garbage. >the "key" is identical on both ends and changes constantly as you transmit, you take a very large (pseudorandom) hash and permute it based on the time of the device and then send small bits of the entire message at a time. I'm not sure if I understand correctly, but does it mean that all devices share the same key, even if it is a constantly changing one?

>>8787 Strelok, a backdoor is a significant issue because for encryption there are two main attack vectors: A flaw in the hardware/software/algorythm which you can exploit to take the key A flaw in human cognition (read: retards) who get killed, or loose their key to an adversary. You can't really prevent no 2. but you can prevent no 1. The human factor is always the weakest but there's no changing that. A"simple" digital radio is in fact using a processor (albiet weaker than say your phone). They need that computational power to encrypt,decrypt, change the keys, and modulate the frequencies they go on (most digitals radios have frequency modulation). You can have multiple "keys" in most devices, or you can change the hash (the really large pseudorandom string, this is usually a military grade feature) used to generate the key but the algorithm remains the same usually. >I just can't see how could they send instructions to some hardware backdoor when the program that handles the digital data should reject anything not encrypted the right way as garbage. >Manufacturer has no say Stream ciphers generally require hardware implementation because it would be too computationally expensive software wise. Backdoors are extremely important as someone could just get the radio to send them your key. For example, a hidden command that you don't know which would give them full access to your radio (key included). Most likely the backdoor would be in the instruction set of the processor, and the spooks have been known to modify electronics of wiretapping victims to have hardware backdoors. x86 has a ton of "unlisted" commands for example, most ostensibly for "development" but there's definitely glow shit in there.

>>8785 Folks have been making encryption devices that plug into your headphone jack for over a decade now that the FBI struggles to crack since making encryption significantly difficult enough for someone to be unable to unencrypt it (without the use of incredibly expensive and limited equipment) isn't that hard. However, keeping the encryption keys secret is incredibly hard if you're relying on prebuilt computer architecture like Israeli-made CPUs as the most infamous example. >>8786 Several countries have semiconductor industries. Most of South-Central Europe like Croatia still has the infrastructure for it. It's just too expensive for those facilities to build consumer-grade semiconductors and compete with China/America so it's mostly national security-related items. The only area really incapable of producing semiconductor technologies for local purposes with the right investments are the majority of South America and Southern Asia because it's either too humid or too high-altitude (not that it can't be done small-batch, but never at export levels). >>8787 Some countries (China, Israel, America, Germany, allegedly Russia) have developed "smart hardware" that has built-in commands to hijack a NIC and transmit data abroad to some central intelligence server. This wasn't a problem when boards largely consisted of inductors, transistors, and capacitors, however anything at an encryption level that a laptop can't crack that doesn't cost six figures per unit is going to require microchips of one form or another which can be compromised quite easily. Even your standard Radio amplifier and tuner these days runs on a basic microchip instead of a dozen transistors acting as filters/OpAmps.

>>8789 >Folks have been making encryption devices that plug into your headphone jack for over a decade now that the FBI struggles to crack What are these devices and how they work?

>>8789 >Most Southern European countries still have infra for semis Strelok, forgive me for being ignorant but I thought NATO had bombed the absolute shit out of those in the former Yugoslavs during the wars in the 90s?

How many 14nm computers does one need to connect in order to make a big iron that can train neural networks in a timely manner?

>>8792 Depends on the size of the NN and the data set - how many parameters are you trying to train and how big is the training dataset? What does "timely" mean to you - an hour, a day, week, month, 6 months? 14nm computers can mean a lot of things, using general purpose CPUs versus GPU or TPU (basically GPUs optimized for machine learning workloads) will make a massive difference. There are plenty of examples/demo projects that will "train neural networks in a timely manner" (seconds or minutes) on whatever potato you used to make that post, but those NNs won't be that useful (e.g. classify pictures of dogs or cats with 90% accuracy).

>>8790 I don't remember the name of them (have a business card somewhere but can't find it at the moment), but they're basically little encryption devices that plug into the headphone jack on devices. The catch is both parties have to own one because it's encrypting the sound wave going out to the broadcasting tower so everything is encrypted the entire way through the process. They ran for like $50-$100 or at least that was their asking price back in 2016/2017ish and the sound quality was garbage/delayed by a couple seconds, but it was most definitely encrypted from glows. >>8791 Eh, we still get parts from Croatia from time to time at work so they've gotta be producing them somewhere. The Adriatic Sea has the perfect sand for doping facilities, the next best locations in the world being the Mekong Delta or the Sonoran desert. >>8792 I feel like this is one of those "how many apple branches do I need to graft to outpace the spread of an apple-killing fungus" questions. I'm sure there's a mathematical answer, but it's going about the problem wrong if I understand what you're saying correctly. It takes far too much processing power for even a supercomputer to actually run through all the processes it needs in real-time to handle a situation. As a news article once put it, it takes about 900,000 arithmetic operations to process a single hand-written digit on a piece of paper and several billion operations to recognize a 3-dimensional object. The size of the computer chip "matters" but when describing a 14nm chip we're describing the spacing between transistor nodes on the chip (on, off, and latching on/partial-on bits depending on architecture), E.G. the "speed" of the chip, not the capacity of the chip. "Smart" networks (such as neural networks) instead rely on building up massive databases of pre-obtained information and then looking for patterns in real-time. In theory you could do this with any old GPU and the better the GPU is the "faster" you can train the model and add more model data to your overall database/data tables to pull from. You can also accomplish the same goal by having a bunch of miniature microprocessors all processing tiny bits of information and sending it to a central server that matches collected data to appropriate databases. As the saying goes, "there are many ways to skin a cat" so I can't really give you a defined answer. Supercomputers used to work by brute-forcing through processing power, but these days most "supercomputers" are effectively just botnets that rely on a bunch of dumb really fast processors sending information to a central server that relays to a database to find patterns. Same thing that a crypto mining operation would use more or less. The downside to the server method is you are exchanging cost of stronger processing for cost of more electricity to power all the dumb devices.

>>8793 >>8794 I was just pretending to be retarded, so let me rephrase it: are there any real drawbacks for the military-industrial complex of a country if they cannot make chips smaller than 14nm? It seems like commercial products made with that level of tech are still plenty powerful for anything other than playing AAA shovelware games and training ”AI”. >whatever potato you used to make that post Don't bully X230-chan! She's cute and she's trying her best!

>>8795 >are there any real drawbacks for the military-industrial complex of a country if they cannot make chips smaller than 14nm? Yes, and no. If your military industrial complex is built around quality over quantity and you have the funding to afford it (America, Canada, Germany, Saudi Arabia, The UK, Israel, Armenia, Australia primarily), then there are real drawbacks to being unable to further improve the quality of the products you are using within your military. There is the drawback of brain drain with this process as you get two emerging "classes" of "the very few eggheads" and "the window lickers using egghead technology specialized in a single task" in your military. If your military consists of standardization to a certain level of quality that you can afford followed by quantity to make up for what you can't afford mixed military tactics (examples being Russia, France, the Emirates, Brazil, the Tatmadaw, Japan, Iran, Syria, etc.) then while there are some benefits to fast microprocessors, the price and availability drawbacks make it so that you can compromise with the understanding that you have to put a certain price tag on every soldier when going up against a quality-based military. In these cases you rely on economies of scale to offset the law of diminishing returns from faster processors and you focus more on experts/specialists instead of on honing technology to make mental expertise unnecessary. There is of course the third category of quantity as a quality of its own (employed by China. India, Vietnam, most of South America, Turkey, Yemen, Azerbaijan, etc.), but in those cases you're probably relying on guerilla tactics and counter-insurgency operations moreso than the quality of your equipment since lives are expendable so there is very little reason to improve your technological capabilities past a certain threshold to keep your military viable against foreign threats. tl;dr- The more you focus on quality over quantity, the more important faster microprocessors become to offset specialization of your military, but if you are willing to put a pricetag on your soldiers and/or have them in more generalist positions, than the law of diminishing returns mean that a 28nm microchip or potentially even a 90nm chip will get the job done just fine. Faster microprocessors only benefit defensive technologies past a certain point.

>>8795 >Don't bully X230-chan! No disrespect. t. Q6600 >>8796 >(America, Canada, Germany, Saudi Arabia, The UK, Israel, Armenia, Australia primarily) >Armenia Huh? >a 28nm microchip or potentially even a 90nm chip will get the job done just fine For what it's worth NASA mostly uses 150nm PowerPC chips in space, amazing how little it really takes to do real work when all the bloat is taken away. >>8795 >are there any real drawbacks for the military-industrial complex of a country if they cannot make chips smaller than 14nm? Not really. Quantum computing might become a game changer in the future for cryptography or data analytics but that's a speculative.

>>8797 >Huh? Armenia has managed to survive this long because they focus on quality of soldiers over quantity due to limited quantity of men. It's actually what fucked them in the Nagorno-Karabakh war because Azerbaijan which is a quantity-based military got drones from Turkey while Armenia did not have drones or drone-destroying tech. When you don't include drone kills the Armenians were killing about 20-30 Azeris for every Armenian lost, but drones balanced it down to about 3-7 Azeris per Armenian lost. The two have comparably sized professional military forces, but Azerbaijan has significantly larger conscript pools to pull from at any time while Armenia can only pull on conscripts in the event of a true existential threat like a full-on invasion.

And in reference to kikepedia numbers in case they get brought up, Azerbaijan completely lied about their casualties and because they are allied with Turkey nobody was actually allowed to go on the ground and investigate their losses. Footage and ground reports clearly show them taking significantly higher losses. The Armenian losses also factor in Artsakh conscripts who had much higher death rates and the Azeris intentionally fluffed their kill rates by counting murdered PoWs and genocided civilians as "combat casualties." In reality Armenia only lost about 500-600 professional soldiers while Azerbaijan lost closer to 7,000-8,000 professional soldiers but you can hide numbers using statistics and cooking the books.

>>8786 >since you can't trust other countries to not backdoor your shit. You can actually grind chips open and look with a microscope into it. Take some samples to check if they try to trick you. It might work with some fancy MRT-like devices as well, but I don't know.

>>8796 >then there are real drawbacks to being unable to further improve the quality of the products you are using within your military I disagree with this sentiment in regards to chip manufacturing capabilities. 28nm products are still 2010 tech, 14nm may be faster but I doubt it would lead to any tangible benefits other than less power/smaller die for a military. Especially considering how efficient these chips can be when you use basic instruction sets and cut out bloat from software. Plus, most modern process are really only used in high end compute and consumer products. Older nodes such as 28, 40 and even up to 90nm are still widely used, and TSMC is building new 28nm fabs and trying to get customers to move from older nodes to 28nm.

>>8799 Sorry but your figures are far beyond ridiculous, that is higher than 10:1 k:d ratio.

>>53015 >As of March of this year, China announced that they managed to create experimental tooling which is capable of producing 14nm chips, slated to go into some form of scale production sometime this year. Does it mean that in the future a country could buy some chinkonium tooling and happily produce 14nm RISC-V potechi? Of course, this is quite an oversimplification, because you need money and expertise, and raw materials, and so on, but it sounds like you don't need to be in the good graces of the US to buy stuff from that one Belgian company.

>>8803 >RISC-V The Chinese are already making RISC-V chips, but I don't expect the Chinese to sell the fab tools imho.

Assuming that a country can manufacture its own potachi, a good way of making a sustainable yet subsidized industry would be to give every student a computer made by that one domestic manufacturer. My idea would be a system-on-a-board in with an e-ink display in a tablet format, andit should be decent enough to play videos if an external monitor is connected. This way little kids could still use it for taking notes, but as they grow older they can switch to using it as a proper computer with a docking station. And even in a small country you (should) have tens of thousands of children entering school every year. And as it happens, the same SoB is what the military would use for nearly everything, even if in that context it's not a tablet.

>>8805 Seems a pretty rational plan. But you're not accounting for the globalist kikes in your scenario strelok. They certainly won't be pleased about you invading their turf with a cleaner, homegrown solution!

>>8805 >but as they grow older they can switch to using it as a proper computer with a docking station Nah, kids and adults alike will want nothing to do with it outside of government mandates or maybe the poorest of the poor. Foreign alternatives will be better/cheaper/sleeker because economies of scale, and inevitably "cooler". Another practical problem is brain drain, most of the homegrown talent in this field will emigrate at first opportunity, either because $$$, or to simply work on interesting/cutting edge problems (because let's be real these domestic programs will always be lagging behind and playing catchup rather than leading things forward). >>8806 >a cleaner, homegrown solution! >from a government-backed project Bless your innocence.

>>8807 >Bless your innocence. Kek. I freely admit I don't know every little thing across the entire universe, I'll strongly assert that even a banana republic is fundamentally more wholesome than (((Intel))), (((AMD))), or any other arm of the Globohomo. :^)

>>8807 >Nah, kids and adults alike will want nothing to do with it outside of government mandates or maybe the poorest of the poor. Foreign alternatives will be better/cheaper/sleeker because economies of scale, and inevitably "cooler". That doesn't matter if using these devices is mandatory during compulsory education, and then they get it for free*, and then they get to keep it, or they can give it back for recycling or reselling. With that many free* devices floating around, you'd inevitably have some hobbyists using them for various projects, and that could eventually lead to some cult following. *Free as in the taxpayers pay for it. >brain drain >lagging behind and playing catchup Neither of these is a problem if you set out realistic expectations and are honest about them. If people pay out of their pockets to learn the basics of anything related to this industry, then letting them go to learn even more is not a loss from a purely monetary perspective, and there is always a chance that one of them starts doing some industrial espionage out of pure patriotism. And the main goal of this whole program would be to make sure that the government doesn't have to rely on foreign tech companies to function, not developing bleeding edge experimental stuff in a l'art pour l'art fashion. Similarly, profitability is not the end goal, although selling stuff to niche markets could at least lessen the financial burden. On that note, other than education and military, I somehow completely forgot that the administration itself could use these computers. It's not like you need supercomputers to work with spreadsheets.

>>8805 >e-ink display in a tablet format >consumes little energy to run >keeps displaying the last image even when the device is turned off >does not need a backlight to work Isn't that perfect for the infantry?

>>8810 e-ink is so fucking cool. It needs to come back in a big way.

>>55992 > thought China (or it seems India) > Iran > Russia had a major deal recently to make a chunky straight railway going in that country order, Even if we disregard the geopolitics, such a railway would have to go through the Himalayas. An other route would be to go through the -stans, but then they'd have to play ball with all of them to make sure that the route works. Overall it is simpler to ship stuff between Iran and Shina through the sea, or send it through Russia. >I am now also realizing Russia's precarious position towards Azerbaturks despite being buddies with Armenia They can just ship stuff through the Caspian Sea. There is a break-of-gauge between Russia (including everything that was the Russian Empire) and Iran, so having to put everything on a boat first is not that great of a loss, because they'd have to put everything on different wagons anyway, and this way the boats can go through the Russian riverine routes, so that they can get quite close to their destination. Shina in theory is trying to build a route through the -stans as part of their belt and road program, but it doesn't seem to going anywhere anytime soon.

bump

Are tanks worth the logistical footprint?

>>8814 Yes. IFVs aren't though.

>>8814 Fair question in CY+8. In Ukraine both sides repurposed them as self propelled artillery for the most part. How many confirmed tank-vs-tank kills so far? In Palestine tanks are slow juicy targets, there's nothing for tanks to oppose that an IFV couldn't. Everyone else can't afford tanks

>Javelin Missile guidance computer - Part 1: teardown https://anontube.lvkaszus.pl/watch?v=11_5TB0-lNw >Javelin Missile guidance computer - Part 2: answers (in French with subtitles) https://anontube.lvkaszus.pl/watch?v=5JRbsv3TXt8 I think I can see why so many people claim that they could make an ATGM for the fraction of the price with off-the-shelf parts. Of course, it would be nice if one of then actually did it and stepped forward, but that is obviously not realistic.

Tanks are much less decisive than they were in WW2 or in the Cold War, mostly due to the ubiquity of Anti-tank weapons right now. Tanks will require another defensive innovation such as better ERA, armor tech, APS to counter Anti-tank weapons to be the decisive breakthrough tools they are. There is also a much more fundamental problem with tanks right now as well. As of right now, modern MBTs are suited to fight other tanks, not against entrenched and hidden infantry teams with ATGMs, Guided Artillery Shells, or FPV drones which are the main primary killers of Tanks right now. With all that said however, they still do their job surprisingly well since both Ukraine and Russia still wants more of them and combined arms tactics are still as valuable as they were in 1940. So, the answer is yes. >>8816 >In Ukraine both sides repurposed them as self propelled artillery for the most part. How many confirmed tank-vs-tank kills so far? More like they take a backseat to the forward infantry assaults, confirmed Tank-vs-Tank engagements are around dozens at most since 'confirming' it in the field is still hard even with all the drone footage. I'm surprised tanks don't seem to make use of personal drones to see what or where they're aiming at, seeing that it'll really help alleviate their weak situational awareness. Maybe because the commander is as pre-occupied as is that they're rather have a drone-equipped infantry be their spotter? >In Palestine tanks are slow juicy targets, there's nothing for tanks to oppose that an IFV couldn't. Tanks have the staying power than the IFV does not, and they can take more hits for the crew to escape than an IFV could. This is seen as a fair trade-off for the IDF seeing that they're so casualty-phobic and would rather waste dozens of tanks than having a squad getting wiped by a single RPG hit.

>>8818 meant for >>8814

>>8818 I think in this context anon means if a country with an economy like [insert country of choice] fighting [other country of choice] should bother with tanks if it plans to dig in for a war of attrition. All the resources spent on a tank fleet could be used to make your defences tougher and meatier, and you might just about be able to take the treeline on the other end of a field without any tanks once the enemy is sufficiently exhausted. I am not saying that this a surefire plan, but I can see why someone would came to this conclusion if we are envisioning a war between two relatively small countries that are on relatively equal footing.

>>8818 >I'm surprised tanks don't seem to make use of personal drones to see what or where they're aiming at, seeing that it'll really help alleviate their weak situational awareness. Maybe because the commander is as pre-occupied as is that they're rather have a drone-equipped infantry be their spotter? More or less. On the logistical side, a single vehicle would have a fairly small number of very vulnerable drones attached, so what to do when they are gone, steal/beg/borrow or do without? The sensible thing is to have a large pool for entire unit that moves together. But once you have it, what’s the point of cutting it into small parcels at any stage? On the functionality side, this: >>38807 Splitting it at a very low level would inevitably lead to various inefficiencies. From training to attention. It’s better to have slightly redundant set of specialists handling such a specialized task as their main job, than a massively redundant set of “amateurs” who have other things to do. There‘s no reason to have in each tank any of the functionality involved beyond request and end result. If the communications work well enough that drones are useful for real-time reconnaissance, they work more than well enough for remote specialists to take requests for what everyone needs to see and make sure they do.