What should we build UAV 3 or beyond the dark chip
Let me remind you that it all started with comments on the article Harmful misconceptions about nanometers. Or why having access to only 90nm in Russia is a disaster, where a lot of experts with fresh registration and a demonstrated lack of understanding of the industry as a whole manifested themselves. In my attempt to reveal the topic of production from the very beginning, there were fewer experts; in the second part, the experts begin to refer to Rosaty, although I don’t know what relation the RBHR of Rosaty have to the declared topic.
I’ll go further.
An interesting point – some commentators do not notice (or do not realize) the obvious – that there are different methods of producing “something”.
Theoretical. Is it possible to make A through B at all. Well, it’s probably possible – for example, in the production of microchips there are operations “to shoot the right atoms at the right place on the plate”. But if commentators ask questions of the level “is it possible to collect extra atoms from the plate with an atomic manipulator and add the missing ones directly to the crystal lattice” or why not shoot an arrow at the sun – then this question in the present form cannot be resolved, since the questioner operates with a false concept.
Laboratory – when we need to get 1 gram, or even a single atom of something once, for example, antimatter. Then you don’t have to think about how to store and transport it and invent traps per ton weight for antitransportation -0.(0..0)1 antigram of antimatter under antigravity conditions.
There is a semi-industrial method – when 90 mg (0.09 g) are obtained from a ton of uranium pitch (uraninite mineral, Yachymov deposit) radium chloride.
There is industrial method – when production begins with the organization of the logistics of several compositions of raw materials and a cascade of equipment in a cave the size of the entire metro ring line.
With the level of rejection during commissioning of pilot production of something complex, several dozens of feedstock compositions will be spent on experiments, and at each stage the product yield will be, at first, good if 0.1%.
There is an economy of the whole process – because any work costs at least the minimum wage, so that the worker does not die of hunger. At least four hundred grams of white bread, and forty grams of butter for professors (engineers can get twenty). And you still need to do something KIP / KIPiA, and finally figure it out with cookies.
There is still a methodology, quality control and other bureaucracy – when the details in each next batch are at least the same as in the previous one, and the parts have readable standard documentation and catalogs available. There are problems with this too.
I will return to the topic. Still a long way to go
Chips. You might think that apart from the CPU and GPU and DRAM and SSD and diodes and bridges and ballasts and drivers (at least for monitors) there is nothing else. However, if commentary experts at least once – during employment – opened the system unit, they would see such a healthy motherboard there. Her situation is also interesting. The 1975 Z80 processor was manufactured with 40 or 44 legs, using 0.004 mm (or 4000 nm) technology and contained 8.5 thousand transistors. Intel 80286 from 1982 – 68 feet, 134 thousand transistors and 1500 nm, respectively. Pentium (P5) Socket 4 from 1993 – 273 feet, 3.100 thousand transistors and 800 nm. Pentium 4 Socket 423 from 2000 – 423 feet (later 478 and 775) with 42.000 thousand transistors and 180 nm (how much is this, 0.000180 mm ? ).
Motherboards for Pentuim 4 were already 4 layered – MSI 850 Pro2 and 6-ply.
If you turn to literaturethen the more connectors – the more layers, so that MSI MEG Z690 Unify – 8 layers, Asrock Z590 – 12 layers. Server boards for 2-4 processors are even thicker in layers. The procedure for designing and debugging (and shielding pickups) is not the easiest. And we are not talking about the mathematics of a common bus. Why all this – there is a nuance that makes me think hard – not a single news about production in the Russian Federation (one,2, 3), there is no indication of where these 10-12 layers of textolite are rolled out and glued, and which production stages are localized – surface mounting, or textolite, and design, and testing. And why the Russian fee, even before all the February changes, cost 100-200 thousand rubles. Let’s skip the production itself and the CPU, and the design and production of chipsets, memory, and so on –
Let’s move on to the periphery.
Let’s forget for a moment that the world of modern microelectronics is somewhat wider than consumer electronics and general purpose chips. Let’s forget about such an interesting thing as the design of radio paths and chips for recalculating polar codes (polar codes2005 year, Erdal ArIkan), and the fact that such equipment is produced in the world – guess how many firms. And how quickly you will learn about the results of the manufacturer’s service failure to supply new equipment to your area and be able to write about it from your phone directly to the Internet. It’s too boring.
In addition to the fact that some data needs to be “calculated”, they also need to be “stored” somewhere, and here the question is the packing density of chips per square (and, taking into account the transition of production to gluing (or soldering) nanostructures – per cubic millimeter) is already of paramount importance. Or maybe experts believe that SD and SSD of the first generations were produced for 4-8-16 megabytes, until recently only out of greed? At this point, I will leave the opportunity for experts to tell in the comments about the production and packaging of SSD chips themselves in the Russian Federation.
But let’s get back to even simpler peripherals. The first HDD RAMAC 305 was released in 1956. Attention question – what is the production of hard drives in the Russian Federation in 2022? What volume? The point is not whether gray imports can bring a suitcase of disks across the border, and not even that a suitcase of disks will take up only 2U out of 42 in one rack.
The question is interesting as follows – for many years there was no alternative to foreign storage manufacturers in Russia “at all”. With the growth of CPU performance, it became possible to produce storage systems with processing exclusively on the central processor, and relatively recently – from 2016-2017. But there is nuance, which commenting experts forget about – a significant percentage of existing storage systems will not work with hard drives and SSDs from the consumer segment. It’s not even about SCSI command set, although in this, of course, too. And not even in 512/4k problem, but that a different microcode is written to a disk originally released at the same factory as the “non-vendor” one. Or something else is changing, quite a bit – for example, testing the level of marriage in the party. Different microcode means that the storage system will not work with this drive. But we’ll hack it, it can’t be hard. I had a program that issued a key for Windows back in 2001. And another one for Simcity. I can even search. Of course, all the codes of storage operating systems can be written off, disassembled, found in the code exactly how the OS interacts with the disk, write off the microcode from the disk and try to choose among the disks available on the market – suitable for replacing the microcode and installing in the storage system. This is sometimes done, and sometimes it even works until it breaks.
The next question is, will it be possible to do this in bulk? You can change one disk in about 2 minutes (and faster, if you count the time not from rebooting the system into update mode, but from the moment you press the “let’s go, darling” button). The procedure is not complicated if there is a ready-made microcode and an unequivocal confidence that this particular microcode is installed on exactly this model of disks of this particular batch with exactly this controller and precisely these servos. If a. And if not? Who in management is ready to sign an order to put a batch of bad disks into operation, knowing that a sudden complete failure can happen when not 1 disk fails in a disk group, but 4 at once? Well, somewhere wrong with the math. How much time and money will it cost to restore a failed group of 24 4TB SSDs with 1 disk failure and 60% occupancy on 4 10G NFS3 interfaces? Why 10G you ask? Because sooner or later you will not have 16/32G due to the policy of licensing and equipment replacement, and 25/100/200G often requires new capital investments. Here, of course, experts will say – but you can order the necessary components on ebay to a fake address and then see the link. suitcase across the border. Can. Only all pieces of equipment from large vendors have a serial number and a continuous history of shipments – to whom it was shipped, when and where. I won’t be surprised if in future versions of the software it turns out that the OS of the storage system itself, after rebooting, boots from an encrypted partition only after entering an individual (issued to a certified vendor engineer) one-time generated key (which is now already used to access the engineering menu of some storage systems), and the disks carry on itself individual (from batch to batch) encrypted non-removable certificates of conformity (which is implemented if you have your own SSD controller) – then such replacements will have even more interesting things.
Not less than uncomfortable questions will start at the next level as soon as I try to move from bare metal to people (this will be the 4th and last part). Competent people, and not only technically literate, but also able to apply this knowledge effectively (for the organization), and before February there were few (and, as we were told many times by HR, they were very very greedy for money – 182.5 grains per year.. and eat too much ), and after February and September they became even smaller. You can, of course, give a dozen more telegrams and send one hundred thousand couriers alone, or ban computer games, or say that sooner or later a commission will be convened to create a department to manage departments. That’s just .. today Georgia and the EU announced that this year alone 1.57 million people arrived to them, of which 600k through Georgia, (one, 2 ) – and there is also Kazakhstan, Turkey and the countries of Southeast Asia. Well, I will wait for HR articles about searching in new realities, the importance of soft skills and unsecured mortgages as a plus.
CERN will build containers for the accumulation and transportation of antimatter
Radium is a killer element that has long been considered a cure. History of Marie and Pierre Curie
Motherboard Layer Stack-Up; Six-Layer Motherboard Stack-Up; Figure 8. Six Layer Stack-Up – Intel Pentium 4 Design Manual
Methods for determining the required number of layers of a multilayer printed circuit board
5G NR Physical Layer Technical Principles and Simulations
Huawei, 5G, and the Man Who Conquered Noise
An overview of channel coding for 5G NR cellular communications
Design of Polar Codes in 5G New Radio